Enhancing The Reading of Dyslexic Children by Reading Acceleration and Auditory Masking
Enhancing The Reading of Dyslexic Children by Reading Acceleration and Auditory Masking
This research attempted to improve the reading performance of dyslexic children through two
different methods: reading acceleration and auditory masking. Participants were 52 dyslexic
children and 52 reading-level matched normal, novice readers. Results indicated that whereas
acceleration improved reading performance in both groups, auditory masking was beneficial
to dyslexic children only. Furthermore, a combined condition of both acceleration and
masking was the most effective in enhancing dyslexic children's comprehension. It is argued
that because normal readers use the phonological route quite effectively, its masking is
detrimental to performance. On the other hand, auditory masking reduces the impact of the
presumed phonological impairment of dyslexic children. Analysis of decoding mistakes
suggested that both manipulations might have resulted in a more effective utilization of
orthographic information and enhanced top-down context effects for dyslexic and novice
readers.
Voluminous research implicates a basic phonological def- ner. The phonological processing system operates via an
icit at the core of dyslexic reading difficulties in word auditory, speech-based route and provides phonemic infor-
recognition. This study attempts to explore the possibility of mation about printed words through the articulatory loop
ameliorating some of this deficit by two independent and (Baddeley, 1986). Information processing via the phonolog-
mutually supporting methods: reading acceleration and au- ical route is sequential and relatively slow. The orthographic
ditory masking. system is a swifter processing mechanism, providing infor-
Word recognition and comprehension are the fundamen- mation about the visual patterns of words or parts of words.
tal component skills of reading (Perfetti, 1985, 1991; Increasingly efficient processing within each system, and
Stanovich, 1991). The connectionist approach to reading among the three (as measured by decoding accuracy and
postulates continuous, distributed interaction of phonologi- rate), facilitates lexical access and improves reading perfor-
cal, orthographic, and semantic processing systems during mance (Seidenberg, 1990).
word recognition and comprehension (Adams, 1990; Bar- Impairment in one (or more) of the three processing
ron, 1991; Reitsma, 1989; Seidenberg, 1990). The amount systems may increase the processing time both within and
and level of activation in each system (Adams, 1990; Sei- between systems. Interference may consequently occur in
denberg, 1990) and the speed of processing within and the timing, coordination, and integration of information
among these systems are products of factors both intrinsic arriving from each system. When connections among the
and extrinsic to the reader (Berninger, 1987, 1990; Bowers systems are inefficient, mismatches may occur among pho-
& Wolf, 1993; Breznitz, 1991). Reading rate and quality of nological, orthographic, and semantic codes.
decoding and comprehension are outcomes of the interac-
One indication of poor interaction among the three sys-
tions between these factors (Breznitz, 1991).
tems during reading is an increase in reading time. When the
Connectionist models portray the process of reading as reader experiences uncertainty or difficulty in correct letter-
beginning with the reception of an image of a printed-letter sound translations, processing within each system and
string by the visual-information processing mechanism. among the three is slowed and impaired (Breznitz, 1991;
This string is then recognized through activation of the Ehri, 1991). Limitations of human information processing
appropriate codes in the phonological and orthographic (Anderson, 1990) dictate that a slow reading rate may affect
systems. These systems operate in tandem to activate the the functioning of various cognitive subsystems, thereby
semantic system (Adams, 1990; Seidenberg, 1990). impeding reading effectiveness (see Breznitz, 1991, for a
Although the phonological and orthographic systems op- review).
erate interactively, they are separate. Each involves differ- Dyslexic children have considerably less proficient word
ent modalities and processes information in a distinct man-
recognition skills than normal readers (Vellutino & Scan-
Ion, 1987). A prodigious amount of research has conclu-
I am greatly indebted to Ann Meyler for her help at all stages of sively implicated a phonological core deficit as the primary
this study. source of dyslexic children's impaired word recognition
Correspondence concerning this article should be addressed to skills. This deficit is held to be the outcome of an inability
Zvia Breznitz, School of Education, University of Haifa, Mt. to process the sound segments of printed materials inter-
Carmel, Haifa, Israel. Electronic mail may be sent via Internet to nally (see Liberman & Shankweiler, 1991, for a review).
redc401 @ uvm. naifa. ac .il. Dyslexic readers* phonological difficulties are both per-
103
104 BREZNITZ
vasive and persistent (Ben-Dror, Pollatsek, & Scarpati, resulted in the greatest improvement in decoding accuracy
1991; Bruck, 1990a, 1990b; Gross-Glen, Jallad, Novoa, but the lowest level of comprehension.
Helgen-Lempesis, & Lubs, 1990; Lefly, 1991). Hence, It is important to note that Carver's (1990) systematic
many researchers and practitioners are concerned with al- studies of reading rate present results that appear contradic-
leviating these problems. Many remediational techniques tory. Specifically, he reports negative correlations between
focus on improving phonological skills, but they are not rate and comprehension. However, a closer look at the
always effective (Lovett, 1992). Despite the abundance of parameters involved may resolve the issue. He found that
research in this area there remains a need to examine alter- the average optimal rate for college students consists of 300
nate approaches. In this study I sought to explore ways in words per minute, with a range of 13% (see Carver, 1990,
which phonological processing might be manipulated p. 146), depending on the difficulty of the reading material.
among dyslexic readers. Two different means were used to Because Breznitz's method of forcing participants to read at
achieve this end: reading acceleration and auditory the fastest pace demonstrated in the self-paced condition
masking. invariably uses the individual's reading rate. The manipu-
lation that was found effective operated within the 13%
range reported by Carver. In fact, the average speed increase
Reading Acceleration of college students (Breznitz et al., 1993, 1994) turned out
to be about 10%.
Typically, phonological decoding occurs relatively The beneficial effects of acceleration in both novice and
slowly in unskilled readers. The slower the decoding pro- older readers have been found to be at least partially attrib-
cess, the less efficient the interactions among the phonolog- utable to improved working memory functioning (Breznitz
ical, orthographic, and semantic systems and the slower the & Share, 1992), reduced visual distractibility (Breznitz,
overall processing rate. 1988), and more effective, context-based information pro-
Reading rate is crucial to reading performance. Not only cessing (Breznitz, 1987a).
does a faster reading speed characterize better readers but Recently, evidence was obtained indicating changes in
also increasing reading speed can improve reading profi- brain activity during reading acceleration (Breznitz et al.,
ciency. Although reading rate is usually conceived of as a 1993, 1994). Electrophysiological measures revealed that
dependent variable, an outcome of the effectiveness of text acceleration significantly decreased the latencies of two
decoding and comprehension (Carver, 1990; Gough & Tun- evoked potential components: an early component (N100)
mer, 1986), extensive research has demonstrated that read- thought to be related to perception, and a later component
ing rate can function as an independent variable that influ- (N400) presumed to be related to semantic processing (Rea-
ences the quality of reading performance (Breznitz, 1987a, gen, 1989). These latencies correlated negatively with com-
1987b, 1988, 1990, 1991; Breznitz, DeMarco, & Hakerem, prehension: The shorter the latency, the higher the compre-
1993; Breznitz, DeMarco, Sbammi, & Hakerem, 1994; hension score. One interpretation of these results was that
Breznitz & Share, 1992). In this research, individual reading reading acceleration might increase the efficiency of cogni-
rates were controlled and manipulated to examine the ef- tive information processing. Conceivably, reading rate ac-
fects of different reading rates on decoding and comprehen- celeration could alter the amount of processing occurring in
sion performance. Participants were prompted to accelerate phonological, orthographic, and semantic systems during
and decelerate their reading rates in a computer-controlled reading. The extent to which this occurs may depend on the
reading situation according to the highest and lowest rates reader's processing resources in each system.
demonstrated by each participant in a preliminary test. The
tester asked participants to read a series of items aloud at
their routine, self-paced reading rate. Using this method,
acceleration and deceleration rates differed for each partic- Auditory Masking
ipant. Because individual ability was taken into account
Phonological processing is essentially a form of auditory
(adjusted reading rates were not equivalent across partici-
processing. Thus, interference in the auditory channel might
pants), the research paradigm was unlike those usually used
obstruct the use of phonological codes during reading and
in this context (e.g., Carver, 1990). Decoding accuracy and
encourage dyslexic readers to use alternative, less impaired
comprehension were measured in each condition.
information. Because the phonological system is presum-
Several experiments have shown that accelerated reading ably intact in control readers, its masking might actually
rate improves the quality of reading by enhancing decoding reduce their reading effectiveness.
and comprehension among novice readers (see Breznitz, Interference might be achieved through the introduction
1991, for a review). Results indicated the following: (a) of a speech-based auditory distractor during reading. Previ-
Readers in the initial stages of reading acquisition can read ous research on the effects of articulatory suppression sup-
faster than they normally do (by about 20%). (b) When ports this suggestion (Brandimonte, Hitch, & Bishop, 1992;
induced to read at the fastest rate demonstrated in the Salame & Baddeley, 1982, 1987). Salarae and Baddeley
self-paced condition, people's decoding accuracy and com- (1982, 1987) demonstrated that unattended speech, unlike
prehension increased significantly. This effect was particu- white noise, reduced performance on a visual digit span
larly pronounced for poor readers, (c) Reducing reading task. The Salame and Baddeley (1987) study further showed
pace to the slowest rate obtained in the self-paced condition that articulatory suppression abolished the unattended
ENHANCING THE READING OF DYSLEXIC CHILDREN 105
speech effect. Later research by Brandimonte et al. (1992) with the accepted definition of dyslexia DSM-/V(DSM; American
indicated that articulatory suppression significantly im- Psychiatric Association, 1994), children were free of "hard" neu-
proved performance in a visual imagery task. The authors of rological signs and did not suffer from primary or severe behav-
ioral or psychiatric disturbances or lack of educational
both studies explained their results within the framework of
opportunities.
the working memory model (Baddeley, 1986; Baddeley &
The 52 dyslexic children were then tested on measures of
Hitch, 1974), suggesting that articulatory suppression acts phonemic awareness (Phonemic Awareness Test: Share, 1988a),
to disrupt phonological codes in the articulatory loop system word recognition (Word Recognition Test: Share, 1988b) and
of working memory. Brandimonte et al. (1992) explained comprehension (one version of the Comprehension Test for
the effect of articulatory suppression in visual tasks as Grades 1 and 2, Israel Ministry of Education and Culture, 1980;
resulting from blockage of the tendency to recode visual Breznitz, 1987a, 1987b). Data are presented in Table 1.
stimuli into verbal short-term memory (STM). They suggest The normal reader group. As the dyslexic children were in the
that the suppression of verbal recitals of visual stimuli has third grade, their matched controls were novice first-grade readers.
the effect of emphasizing the establishment of a visual code. For each dyslexic reader, the first control reader that matched on
Hence, the reduced availability of verbal codes is assumed gender, IQ, handedness, word recognition, and comprehension
to improve the learning of the visual characteristics of the scores was designated as his or her matched pair. Out of a large
stimuli. These authors interpreted their findings in terms of sample of children (iV = 367) who participated in several other
studies, 52 right-handed, normal readers (48 boys and 4 girls) were
a tradeoff between verbal and visual coding that is mediated
matched with the dyslexic readers on IQ and on the measures of
by phonological STM. Such a tradeoff can occur because word recognition and comprehension described. The mean age of
the visual and phonological subsystems of working memory the normal reader group was 6 years and 9 months (SD = 2
compete for limited central processing capacity. months) and ranged from 6 years and 4 months to 7 years and 5
In this study, I hypothesized that reading acceleration months. The mean estimated IQ score (abridged Israeli-normed
would improve the reading performance of both dyslexic WISC-R; Sattler, 1982) was 107.0 (SD = 1.1) and ranged from
105.6 to 113.8. The average score on the Vocabulary subtest was
and control readers. Although auditory masking might ac-
M = 10.5 (SD = 0.5) and ranged from 10.2 to 13.1. The mean
tually disrupt the reading of normal controls, it was ex- score on the Block Design subtest was 11.9 (SD = 1.0) and ranged
pected to have a facilitating effect on the dyslexic readers* from 10.7 to 13.8. Following selection, normally reading children
performance. Finally, dyslexic readers were expected to
exhibit the highest level of performance in the condition
combining both manipulations. Table 1
Means and Standard Deviations for Phonological, Word
Recognition, and Comprehension Tests Among
Method Dyslexic and Control Reader Groups
Participants Dyslexic Control
Test readers readers r(102)
One hundred and four children, 52 dyslexic and 52 normal Phonology
readers, participated in this study. All children were of middle-
Synthesis errors
class background and attended public elementary schools in an M 5.77 2.48 6.97*
urban area of northern Israel. Hebrew was their primary language. SD 4.39 2.74
Children in both groups were matched on IQ scores, reading level, Analysis errors
gender, and handedness. M 6.28 3.55 5.25**
The dyslexic reader group. Dyslexic readers were recruited SD 6.04 4.26
from the afterschool Learning Disability Clinic run by the Depart- Reversal errors
ment of Education. A total of 78 children were referred to the M 7.11 5.67 4.12**
clinic by their teachers or school psychologists. At the time of their SD 4.89 3.77
Total errors
referral, all children were diagnosed by an intake battery that M 19.16 11.70 2.68*
included an evaluative test of their reading skills (Reading Skills SD 5.66 4.02
Test; Nitzan Learning Disability Center, 1988). Only 52 of the Word recognition
children met the Israeli criterion for dyslexia, which requires at Words correct (out of 45)
least 1 year and 8 months of reading retardation (Nitzan, 1988). M 21.59 23.48 0.34
Forty-eight of these children were boys and 4 were girls. The SD 6.87 5.40
children's ages ranged from 8 years and 7 months to 9 years and Nonwords correct (out of 45)
10 months. The mean age was 9 years and 1 month (SD — 1 year M 15.57 19.61 2.49*
and 4 months). The average number of years of reading retardation SD 6.01 5.23
was 1 year and 11 months (SD = 9 months) and ranged from 1 Words time (in s)
M 51.59 37.48 10.20**
year and 8 months to 2 years and 2 months. The children's SD 19.53 10.32
estimated IQ scores (based on the Vocabulary and Block Design Nonwords time (in s)
subtests of the Israeli-normed WISC-R; Sattler, 1982; Wechsler, M 74.44 50.32 12.49**
1977) yielded a mean score of M — 106.0 (SD = 1.3) and a range SD 26.51 24.26
of 102.2 to 114.1. The mean score on the Vocabulary subtest was Comprehension 3.09 3.10 0.52
M - 10.3 (SD = 2.1) and ranged from 10.5 to 12.8. The mean Correct (out of 6) 1.14 1.11
score on the Block Design subtest was M = 11.9 (SD = 1.2) and Note. Absolute t values are presented.
ranged from 10.9 to 14.1. All children were right-handed. In line *p < .05. * * p < .001.
106 BREZNTTZ
were administered the phonological test. The data are presented in whether spare capacity was available in the auditory channel
Table 1. during the reading process that created an increase in the signal-
The results of t tests indicated significant differences between to-noise ratio. This question was particularly important as children
the dyslexic and the normal reader groups on the phonological were at a reading stage in which the auditory route is used
awareness test. Dyslexic children made more errors overall and on extensively. The pretest data indicated that only the familiar tune
each of the individual subtests, and their performance on decoding influenced the reading performance of the two groups of children.
tasks was worse. Dyslexic children read both the real word and the Therefore, in the present study the familiar song was adapted to the
nonword lists in the word recognition battery slower than the computer and activated simultaneously with the appearance of
normal controls. Although they made significantly more errors in each item on the computer screen.
reading nonwords, they performed as well as the normal group
when decoding real words. Their comprehension scores did not
differ from those of the normal readers.
Procedure
Materials After selection of the 52 dyslexic children and their normal
controls, all children took the phonemic awareness test (see Table
Word recognition test. This test (Share, 1988b) comprises two 1). Following this, I conducted manipulation of reading rate. The
subtests. The first measures recognition of real words and the experiment was carried out in two stages, with a 1-week interval
second assesses recognition of pronounceable nonwords. Both between them.
subtests are composed of 40 items of increasing difficulty and Part 1 of the experiment was designed to measure the effect of
length arranged in groups of 10 (i.e., beginning with 10 items of 1 accelerated reading rate on the decoding and comprehension per-
syllable each and concluding with 10 items of 4 syllables). Mea- formance of dyslexic children as compared with their normal
sures included the number of errors and mean reading time per 10 controls. All children read three parallel versions of the Compre-
items. hension Test under 3 reading-rate conditions: self-paced, fast-
Phonological test. The phonemic awareness test (Share, paced, and again self-paced. The order of the conditions was fixed,
1988a) was designed to evaluate the children's phonemic aware- as the fast-paced manipulation was necessarily based on the self-
ness skills. (As the children were unskilled readers, the focus was paced reading rate (Breznitz & Share, 1992).
on the assessment of this form of phonological ability.) The test Self-paced 1. In this condition, each item (i.e., passages con-
consists of a series of nonsense syllables and is composed of three sisting of 1-3 sentences, followed by a multiple-choice question)
subtests comprising 15 items each. The first measures phoneme was presented individually on the screen, and children were asked
synthesis, the second phoneme analysis, and the third phoneme to begin reading orally immediately on appearance of the item.
reversal. Each subtest includes a two-phoneme practice section One training trial was given before the set of six test items. To
containing 4 items. Each practice section is followed by three test reduce the possibility of looking back, the experimenter, who
sections containing 5 items each, arranged in order of increasing controlled both item presentation and item erasure, removed the
complexity (from two-phoneme units to four-phoneme units). The item immediately after reading was completed. The computer
maximum possible score for each subtest is 15 and the maximum clock was activated when the child began reading and stopped
overall score is 45. when pronunciation of each item was completed. Per-letter reading
Comprehension test. Seven parallel versions (Breznitz, 1987a, rates were calculated for each item, yielding six separate per-letter
1987b, 1988, 1989, 1990, 1991; Breznitz & Share, 1992) of the reading rates. For each child, the fastest reading rate of the six was
Comprehension Test for Grades 1 and 2 (Israel, 1980) were ad- used as his or her individual acceleration rate in the fast-paced
ministered. The seven versions (i.e., Versions 1-7) are equivalent condition.
in terms of comprehension level. Each version was composed of a Fast-paced. In the second condition, each child read another
sample item, followed by six test items of increasing difficulty in version of the test (six additional items) individually accelerated
terms of length and complexity. The items in each version con- according to his or her fastest self-paced reading rate. To minimize
sisted of 1-3 declarative sentences followed by an inferential disruption of natural eye movements, the following procedure was
question and 4 short multiple-choice answers. To illustrate, 2 adopted: The experimenter introduced the entire item on the screen
sample items were as follows: Easy—A child came home in the and the child was asked to begin reading aloud immediately. As
afternoon and asked his mother for food. What did the child ask soon as reading commenced, the passage was automatically erased
for? (a) salt, (b) a plate, (c) water, (d) bread. Hard—The oranges letter by letter at the individual's maximum per-letter rate. One
are taken from the grove to the packing station by cart. From there training trial was given before the set of six test items.
they are sent by train to the port, and from the port they are sent Self-paced 2. To control for a possible warm-up effect, a third
overseas by boat. How do the oranges reach the port? (a) by boat, version of the test was read again at a self-paced reading rate after
(b) by cart, (c) by train, (d) by car. the fast-paced condition was completed.
One test form of the Reading Comprehension Test (Version 1) Immediately following the reading of each item in each condi-
was used for evaluation of the dyslexic children's comprehension tion, the item was removed from the screen and the multiple-
level and for the selection of the reading-level matched control choice answers appeared on the screen. As with the preceding text,
group. In each experiment (acceleration and auditory masking), the the children read the answers orally. The children were asked to
six additional versions (Versions 2-7) were randomly distributed press the letter on the computer keyboard corresponding to the
among children according to a fully counterbalanced design. In correct answer.
each of the two experiments reported here, children read the test Part 2 of the experiment attempted to determine the effect of
items from an IBM personal computer screen. loading the phonological route with auditory stimuli irrelevant to
Auditory masking. A pretest on the effects of auditory inter- the central reading task. As with the procedure for Part 1 of the
ference during the reading process was conducted. Three types of experiment, all children in Part 2 of the experiment read three
distractors were compared: white noise, an unfamiliar tune, and the (different) parallel versions of the Comprehension Test, using the
music to a well-known children's song. The aim was to determine three reading rate conditions: self-paced, fast-paced, and again
ENHANCING THE READING OF DYSLEXIC CHILDREN 107
self-paced. However, in this part of the experiment an auditory Table 2 for a comparison of means and standard deviations
distractor was introduced during each reading. between the reading groups.
The music of a well-known children's song was played by the
computer during all three reading rate conditions. As each item
appeared, the melody started playing and then stopped when the Overall Performance
item disappeared from the screen. Hence, presentation of the
melody and the text was simultaneous. In the two self-paced To determine whether dyslexic children, in a similar
conditions, items were manually erased by the experimenter when manner to normal beginning readers, would benefit from the
the child completed reading. Cessation of the melody occurred acceleration of their reading rate, I estimated the overall
immediately with erasure of the text. In the fast-paced condition,
effects of accelerated reading rate by using a two-way
both the items and the music disappeared automatically on the
basis of the fastest reading rate achieved by the individual in the repeated measures multivariate analysis of variance
self-paced condition.
Measures Table 2
Means and Standard Deviations for Reading
Scores on the phonemic awareness test were based on the Comprehension, Time, and Number of Errors:
number of correct responses obtained. A maximum score of 15 Self- and Fast-Paced Conditions, With and
was possible on each of the subtests (phoneme synthesis, analysis,
Without Auditory Masking
and reversal, respectively), making the maximum overall test score
45 (Share, 1988a). Pace Self 1 Self 2 Fast '(51)
In each of the three reading manipulation conditions, reading
times, comprehension, and oral reading errors were assessed as Without auditory masking
follows: Comprehension
Dyslexic readers
1. Reading times in seconds were obtained for each of the M 3.11 3.11 3.90 6.22**
six items in each test version. Times were measured by SD 0.98 0.84 0.91
the computer during the reading of each item. Control readers
2. Comprehension scores were based on the number of cor- M 3.09 3.09 4.73 13.30**
rect responses achieved by each child, with a maximum SD 0.97 0.91 0.88
score of 6. Time (in s)
3. Two types of scores were procured for oral reading errors: Dyslexic readers
an overall total number of errors for the six combined M 34.13 32.10 22.50 4.63**
SD 15.37 17.56 8.70
items in each condition, and scores for the different types Control readers
of errors in each condition. To analyze the oral reading M 30.23 29.12 21.63 7.60**
errors, all of the tests were recorded on an audiocassette SD 14.28 13.02 9.89
recorder. Children were asked to read all items aloud as Errors
accurately as possible. Two coders recorded the oral Dyslexic readers
reading errors from the audiocassettes. In addition, a pair M 27.01 26.30 29.30 0.97
of trained coders independently marked the number of SD 13.45 11.03 11.71
errors on transcripts of the tests. Based on Adams's Control readers
(1990) miscues analysis of oral reading, errors were clas- M 14.00 13.99 8.69 4.93**
sified according to four mutually exclusive categories: (a) SD 6.12 7.34 6.02
Substitution errors (i.e., a letter, vowel, syllable, word, or With auditory masking
part of a sentence is read instead of the original); (b) Comprehension
Repetition errors (i.e., repetition of a letter, vowel, sylla- Dyslexic readers
ble, word, or part of a sentence); (c) Deletion errors (i.e., M 3.19 3.12 4.30 17.08**
the omission of a letter, vowel, syllable, word, or part of SD 0.93 0.88 0.96
a sentence); (d) Addition errors (i.e., the insertion of a Control readers
letter, vowel, syllable, word, or part of a sentence not M 3.15 3.21 3.63 4.45**
present in the original text). SD 0.95 0.71 0.89
Time (in s)
Intercoder reliability was assessed by using intraclass correla- Dyslexic readers
tions (reliability of single ratings; Bartko, 1966; Shrout & Fleiss, M 27.20 29.11 18.77 7.71**
1979) based on each coder's evaluation of the oral reading errors SD 11.19 13.33 6.99
of 12 children (6 dyslexic and 6 normal readers). Reliability ranged Control readers
between 0.85 and 0.96 (M = 0.93). M 31.15 32.17 21.02 10.15**
SD 10.42 11.03 5.95
Errors
Results Dyslexic readers
M 23.09 22.01 17.78 6.91**
SD 10.88 9.64 9.35
The data indicated that there were no significant differ- Control readers
ences in the performance between the first and second M 13.25 13.89 12.84 1.78
self-paced conditions of either group of children. Therefore, SD 4.95 4.01 5.10
only the data from the first self-paced condition were re- Note. Absolute t values are presented.
tained for comparison with the fast-paced condition. See **p < .001.
108 BREZNITZ
Dyslexic
100
90
80
'•
70
60 - Dyslexic Self
o 50-
• Dyslexic Fast
a 40
30
20
10
0
1 3 4
Difficulty Level
B
Control
•Control Self
u
•Control Fast
o.
3 4
Difficulty Level
Figure 1. Reading comprehension gains across test items for dyslexic and control groups during
fast-paced reading.
was registered for the control group. Reading time was hypothesis that diminishing dyslexic readers' reliance on
related only to the pace variable, F(l, 102) = 53.02, p < their problematic phonological systems may enhance their
.001. (Means and standard deviations for the three reading reading efficiency. In the control group, auditory masking
measures under Auditory Masking appear in Table 2.) produced a small, nonsignificant increase in reading time,
Also note that in this part of the experiment, self- and and the interaction between condition and group was found
fast-paced reading times were highly correlated for all of the to be significant, F(l, 102) = 11.96, p < .001. Thus, it
children: r = .73, p < .001, for the dyslexic readers, and r = appears that auditory masking produced effects similar to
.74, p < .001, for the normal readers. the fast-paced manipulation among the dyslexic children.
In the self-paced condition, neither comprehension nor To determine the relative impact of fast-pacing with au-
errors were significantly altered by auditory masking (see ditory masking on reading performance as opposed to fast-
Table 2). At the same time, the dyslexic children's reading pacing alone among both groups of readers, a three-way
speed significantly improved in the auditory masking con- MANOVA was conducted (Group X Pace X Masking).
dition, t(51) - 3.99, p < .001. This is in line with our Errors, comprehension, and reading time were the depen-
110 BREZNITZ
Table 3
Means and Standard Deviation for Types of Error: Self and Fast-Paced Conditions,
With and Without Auditory Masking
Without auditory masking With auditory masking
Pace Self Fast *51) Self Fast '(51)
Repetition
Dyslexic readers
M 9.44 7.62 4.70** 5.79 2.48 6.97**
SD 6.05 4.24 4.40 2.75
Control readers
M 5.60 4.10 5.18** 6.29 3.56 5.25**
SD 2.73 1.50 6.41 4.27
Substitution
Dyslexic readers
M 11.08 8.12 2.59** 7.12 5.67 4.12**
SD 7.47 5.24 4.89 3.77
Control readers
M 7.58 3.58 7.12** 8.96 8.40 2.86*
SD 3.99 2.80 4.79 4.89
Deletion
Dyslexic readers
M 3.56 6.33 3.97** 2.10 2.92 2.44*
SD 2.97 4.18 2.90 2.40
Control readers
M 0.10 0.60 0.90 0.67 1.67 3.41**
SD 0.41 1.39 1.06 2.39
Addition
Dyslexic readers
M 2.92 7.48 7.10** 1.58 0.75 5.52**
SD 2.35 3.51 1.24 1.12
Control readers
M 0.73 0.77 0.88 0.19 0.12 1.66
SD 1.30 1.53 0.40 0.38
Note. Absolute t values are presented.
*p < .05. **p < .001.
dent variables. The results indicated a significant Group X effects of order are similar in both groups, this remains a
Pace X Masking interaction (multivariate F), F(3, 101) - question for further research.
25.4, p < .001. The dyslexic children improved their read-
ing performance to the greatest extent in the combined
fast-paced reading with auditory masking condition. A uni- Types of Errors
variate analysis indicated that improvement in the combined
condition was due to more dramatic comprehension in- A repeated measures MANOVA was conducted on the
creases, F(i, 103) = 46.7, p < .001, and a further reduction four types of errors—repetition, substitution, deletion, and
in reading errors, F(l, 103) = 27.1, p < .001, as compared addition—under auditory masking. A significant multivar-
with the fast-paced condition alone. In contrast, the most iate Group X Pace effect was found, F(4, 99) = 6.11, p <
significant improvement in reading performance among .001. A univariate analysis indicated that during fast-paced
normal controls occurred in the fast-paced alone condition reading with auditory masking, both groups increased dele-
(in the absence of auditory masking). tion errors, F(l, 102) = 7.90, p < .006, and reduced
Although the fast-paced alone condition always preceded repetition, F(l, 102) = 4.71, p < .03, and substitution
the auditory masking conditions, so that possible order errors, F(l, 102) = 22.92,/? < .001. A significant Group X
effects were not controlled, the results suggest that order of Pace interaction was found only for substitution errors, F(2,
presentation may not have been a source of confounding. 101) = 13.12, p < .04. The dyslexic children exhibited a
Among normal control readers, fast-pacing alone proved greater reduction in the frequency of substitution errors than
more effective in terms of reading performance than did the normal controls. Addition errors were significantly re-
fast-pacing in combination with auditory masking. Had duced only in the dyslexic group (See Table 3).
effects of order been present, a similar improvement in the The relative impact of fast-pacing with auditory masking
last combined condition would have been expected. The on reading performance as opposed to fast-pacing alone
possibility that order effects were present among the dys- among both groups of readers on the four types of errors
lexic readers cannot be eliminated on the basis of the was measured by a three-way MANOVA. Results indicated
present procedures. As it cannot been assumed that the a significant Group X Pace X Masking interaction, F(4,
ENHANCING THE READING OF DYSLEXIC CHILDREN 111
99) = 13.2, p < .001. A univariate analysis indicated that dyslexic readers appear to be less impaired (Manis, Szeszul-
the dyslexic children reduced their repetition errors, F(l, ski, Holt, & Graves, 1990; Olson, Davidson, Kliegal, &
102) = 15.1, p < .001, and addition errors, F(l, 102) = Foltz, 1985; Olson, Wise, Conners, Rack, & Fulker, 1989;
21.3, p < .001, to a greater extent than did normal controls Pennington, Lefly, Van-Orden, Bookman, & Smith, 1987;
in the combined condition as compared with the fast-paced Rack, 1985). The results of the first experiment indicate
alone condition. that, although to a lesser degree than normal controls, dys-
Among the dyslexic readers, better reading comprehen- lexic children benefited from acceleration, particularly in
sion was associated with fewer repetition and substitution terms of improved comprehension.
errors in the self-paced condition (r = —.38, p < .03, and In the absence of direct measures of auditory interference
r — -.54, p < .001, respectively) but not in the fast-paced in the fast-paced condition, it is naturally impossible to rule
condition (r = -.10, p < .46, and r = - . 2 1 , p < .13). out alternative explanations. Thus, further research should
Comprehension was associated with more deletion and ad- attempt to rule out the possibility that the increased effec-
dition errors only in the fast-paced condition (r = .42, p < tiveness was not caused by a more optimal arousal.
.001, and r = .62, p < .001). For the normal readers these The manipulation tested in the second part of the exper-
types of errors were correlated with reading comprehension iment consisted of direct interference with phonological
in both experimental conditions (repetition: r = - .64, p < processing through auditory masking. The rationale here
.001, and r — - . 5 7 , p < .001, respectively, for self- and was based on the notion that one method of reducing dys-
fast-paced; substitution: r = -.45,/? < .001, and r = —.68, lexic children's reliance on their impaired phonological
p < .001). No significant correlations were found for the skills was to overload it with task-irrelevant information. As
normal readers between reading comprehension and dele- expected, auditory masking was found to be somewhat
tion or addition errors in either experimental condition. detrimental for normal readers, who were deprived of an
effective information-processing route. In the case of dys-
lexic readers, however, even in the self-paced condition,
Discussion auditory masking significantly decreased all types of oral
reading errors, increased their reading speed and, indirectly,
The research reported here attempted to enhance the the beneficial affects of acceleration.
reading performance of dyslexic readers by using two dif- The most dramatic gain in dyslexic children's reading
ferent experimental manipulations: forced acceleration of performance was obtained in combining both acceleration
reading rate and auditory masking. The results indicate that and auditory masking. Note that in this condition, the dys-
among dyslexic children the two manipulations were suc- lexic children's comprehension was statistically better than
cessful, with the strongest improvement occurring when that of the normal control group, t(102) = 3.71, p < .001.
both were combined. In contrast, normal controls improved Auditory masking during fast-paced reading interferes with
their reading performance to the greatest degree during the performance of normal beginning readers who depend
fast-paced reading in the absence of auditory masking. The primarily on their effective phonological processing during
underlying rationale for the specific treatments explored in word recognition. It appears that although they have suffi-
this investigation relates to the well documented difficulties cient spare capacity in the self-paced condition to overcome
of dyslexic readers with phonologically based information the extra load on the auditory channel, the combination of
processing. masking and reading acceleration impedes their capabilities.
Previous research has shown that reading acceleration As the melody was that of a familiar song, it is highly likely
improves reading effectiveness in normal readers (Breznitz, that they translated it into speech-based code. Speech-like
1987a, 1987b, 1988, 1989, 1990, 1991; Breznitz & Share, material gains automatic access to the phonological system
1992). The same phenomenon was demonstrated in both (Salame & Baddeley, 1987). Therefore, the familiar tune
sets of results reported here. Although the specific mecha- may have filled phonological channel capacity, leaving few
nisms that account for the benefits of acceleration are still processing resources for the reading task. Consequently, the
an open issue, the ones that have been suggested so far (i.e., potential benefit of acceleration was canceled out by mask-
enhanced operation of working memory, greater reliance on ing. In the case of dyslexic readers, the situation is different
orthographic information, and higher involvement of top- because both manipulations work in the same direction.
down context mediated processes) all share a reduced reli- Note, however, that in this study no attempt was made to
ance on phonologically based decoding. Consequently, ac- distinguish between pre-lexical and post-lexical phonolog-
celeration could be seen as one possible way to shift the ical interference. Because post-lexical phonological recod-
emphasis away from the slow phonological route to other, ing plays a role in maintaining information in working
possibly compensating, routes of information processing memory, this ought to be clarified in future studies.
during reading. It is argued that both acceleration and auditory masking
This reasoning is in line with several studies that suggest effectively altered the distribution of processing resources
that an increased exposure to print forces dyslexic readers to among phonological, orthographic, and semantic processing
compensate for their phonological impairment by increasing systems. This may have engendered some changes in pro-
their reliance on orthographic and contextual cues (Stano- cessing both within and between the systems and between
vich, 1980, 1986) thereby reducing their decoding errors. the two groups of children.
Compared with phonology, the orthographic skills of most An analysis of decoding errors may be viewed as lending
112 BREZNITZ
further support to the notion that both auditory masking and effect of pictorial distractors. Journal of Research in Education,
acceleration, separately and in combination, reduced the 83, 47-53.
dyslexic children's reliance on the phonological route. Dys- Breznitz, Z. (1989). Teacher evaluation of pupils' reading perfor-
lexic children experienced a reduction in all types of oral mance: Criteria for assessment. Reading Performance, 26,
reading errors in the presence of auditory masking in the 174-181.
self-paced condition. Although the reduction in errors was Breznitz, Z. (1990). Vocalization and pauses in fast-paced reading.
Journal of General Psychology, 117, 153-161.
dramatic for dyslexic children during auditory masking
Breznitz, Z. (1991). The beneficial effect of accelerating reading
alone, it was even more so when combined with reading rate on dyslexic readers' reading comprehension. In M. Snow-
acceleration. ling & M. Thomson (Eds.), Dyslexia: Integrating theory and
The potential implications of these results to the theory of practice (pp. 236-245). London: Whurr.
dyslexia must be tempered with the reservations they call Breznitz, Z., DeMarco, A , & Hakerem, G. (1993). Topographic
for. At the same time, it is conceivable that whatever ulti- measures of cerebral activity during reading of text at fast- and
mate explanations are offered for the mechanisms involved, slow-paced rates. Brain Topography, 6(2), 117-121.
evidence of significant reading improvement may eventu- Breznitz, Z., DeMarco, T , Shammi, P., & Hakerem, G. (1994).
ally pave the way to successful therapeutic intervention. Self-paced versus fast-paced reading rates and their effect upon
Effective interventions would, of course, .be based on the comprehension and event-related potentials. Journal of Genetic
broadest possible approach, using both phonological train- Psychology, 155, 397-409.
ing and some of the ideas suggested in this article. Breznitz, Z., & Share, D. L. (1992). The effect of accelerated
reading rate on memory of text. Journal of Educational Psy-
chology, 4, 193-199.
Bruck, M. (1990a). Word-recognition skills of adults with child-
References hood diagnoses of dyslexia. Developmental Psychology, 26,
439-454.
Adams, M. J, (1990). Beginning to read: Thinking and learning Bruck, M. (1990b). Persistence of dyslexics' phonological aware-
about print. Cambridge, MA: MIT Press. ness deficits. Developmental Psychology, 28, 874-886.
American Psychiatric Association. Diagnostic and statistical man-
Carver, R. P. (1990). Reading rate: A review of research and
ual of mental disorders (4th ed., 1994). Washington, DC:
theory. San Diego, CA: Academic Press.
Author.
Ehri, L. (1991). The development of reading and spelling in
Anderson, J. R. (1990). Cognitive psychology and its implications.
children: An overview. In M. Snowling & M. Thomson (Eds.),
New York: W. H. Freeman.
Dyslexia: Integrating theory and practice (pp. 63-79). London:
Baddeley, A. D. (1986). Working memory. Oxford: Clarendon
Whurr.
Press.
Gough, P. B., & Tunmer, W. E. (1986). Decoding, reading, and
Baddeley, A. D., & Hitch, J. G. (1974). Working memory. In G. H.
reading disability. Remedial and Special Education, 7, 6-10.
Bower (Ed.), The psychology of learning and motivation (Vol. 8,
pp. 47-90). New York: Academic Press. Gross-Glen, K., Jallad, B., Novoa, L., Helgen-Lempesis, V., &
Lubs, H. A. (1990). Nonsense passage reading as a diagnostic
Barron, R. W. (1991). Proto-literacy, literacy and the acquisition
aid in the study of adult familial dyslexia. Reading and Writing:
of phonological awareness. Learning and Individual Differ-
An Interdisciplinary Journal, 2, 161-173-
ences, 31, 243-255.
Bartko, J. J. (1966). The intraclass correlation coefficient as a Israel Ministry of Education and Culture. (1980). Reading com-
measure of reliability. Psychological Reports, 19, 3-11. prehension test for grades one and two. Jerusalem: Author.
Ben-Dror, I., Pollatsek, A., & Scarpati, S. (1991). Word identifi- Lefly, D. L. (1991). Spelling errors and reading fluency in com-
cation in isolation and in context by college dyslexic students. pensated adult dyslexics. Annals of Dyslexia, 41, 143-162.
Brain and Language, 40t 471-490. Liberman, I. Y., & Shankweiler, D. (1991). Phonology and begin-
Berninger, V. W. (1987). Global, component, and serial processing ning to read: A tutorial. In L. Reiben & C. A. Perfetti (Eds.),
of printed words in beginning reading. Journal of Experimental Learning to read (pp. 3-17). Hillsdale, NJ: Erlbaum.
Child Psychology, 43, 387-418. Lovett, M. W. (1992). Developmental dyslexia. In S. J. Seglowitz
Berninger, V. W. (1990). Multiple orthographic codes: Key & I. Rapin (Eds.), Handbook of neuropsychology: Vol. 7: Child
to alternative instructional methodologies for developing neuropsychology. New York: Elsevier.
orthographic-phonological connections underlying word identi- Manis, F. R., Szeszulski, P. A., Holt, L. K., & Graves, L. (1990).
fication. School Psychology Review, 19, 518-533. Variation in component word recognition and spelling skills
Bowers, P. G., & Wolf, M. (1993). Theoretical links among nam- among dyslexic children and normal readers. In T. H. Carr &
ing speed, precise timing mechanisms and orthographic skill in B. A. Levy (Eds.), Reading and its development: Component
dyslexia. Reading and Writing: An Interdisciplinary Journal, 5, skills approaches (pp. 207-259). New York: Academic Press.
69-85. Nitzan Learning Disability Center. (1988). Reading skills test. Tel
Brandimonte, M. A , Hitch, G. J., & Bishop, D. V. M. (1992). Aviv: Author.
Verbal receding of visual stimuli impairs mental image trans- Olson, R.K., Davidson, B. J., Kliegal, R., & Foltz, G. (1985).
formations. Memory and Cognition, 20, 449-455. Individual and developmental differences in reading disability.
Breznitz, Z. (1987a). Increasing first-graders' reading accuracy In G. E. Mackinnon & T. G. Waller (Eds.), Reading research:
and comprehension by accelerating their reading rates. Journal Advances in theory and practice, (Vol 4, pp. 1-64). New York:
of Educational Psychology, 79, 236-242. Academic Press.
Breznitz, Z. (1987b). Reducing the gap in reading performance Olson, R., Wise, B., Conners, F., Rack, J., & Fulker, D. (1989).
between Israeli lower- and middle-class first grade pupils. Jour- Specific deficits in component reading and language skills:
nal of Psychology, 121, 491-501. Genetic and environmental influences. Journal of Learning Dis-
Breznitz, Z. (1988). Reading performance of first graders: The abilities, 22, 339-348.
ENHANCING THE READING OF DYSLEXIC CHILDREN 113
Pennington, B.F., Lefly, D. L., Van-Orden, G.C., Bookman, Share, D. (1988a). Phonemic awareness test. School of Education,
M. O., & Smith, S. D. (1987). Is phonology bypassed in normal University of Haifa, unpublished manuscript.
or dyslexic development? Annals of Dyslexia, 37, 62-89. Share, D. (1988b). Word recognition test. School of Education,
Perfetti, C. A. (1985). Reading ability. New York: Oxford Univer- University of Haifa, unpublished manuscript.
sity Press. Shrout, P. E., & Fleiss, J. L. (1979). Intraclass correlations: Uses in
Perfetti, C. A. (1991). Representations and awareness in the ac- assessing rater reliability. Psychological Bulletin, 86, 420-428.
quisition of reading competence. In L. Reiben & C. A. Perfetti Stanovich, K. E. (1980). Toward an interactive-compensatory
(Eds.), Learning to read (pp. 33-46). Hillsdale, NJ: Erlbaum. model of individual differences in the development of reading
Rack, J. P. (1985). Orthographic and phonetic coding in develop- fluency. Reading Research Quarterly, 16, 182-201.
mental dyslexia. British Journal of Psychology, 76, 325-340. Stanovich, K. E. (1986). Matthew's effects in reading: Some con-
Reagen, D. (1989). Human brain electrophysiology. New York: sequences of individual differences in the acquisition of literacy.
Elsevier.
Reading Research Quarterly, 21, 360-406.
Reitsma, P. (1989). Orthographic memory and learning to read. In
Stanovich, K. E. (1991). Word recognition: Changing perspec-
P. G. Aaron & R. M. Joshi (Eds.), Reading and writing in
different orthographic systems (pp. 51-73). Dordrecht, The tives. In R. Barr, M. L. Kamil, P. Mosenthal, & P. D. Pearson
Netherlands: Kluwer Academic. (Eds.), Handbook of reading research (pp. 418-452). New
Salame, P., & Baddeley, A. (1982). Disruption of short-term York: Longman.
memory by unattended speech: Implications for the structure of Vellutino, F. R., & Scanlon, D. (1987). Phonological coding, pho-
working memory. Journal of Verbal Learning and Verbal Be- nological awareness, and reading ability: Evidence from a lon-
havior, 21, 150-164. gitudinal and experimental study. Merrill-Palmer Quarterly, 33,
Salame, P., & Baddeley, A. (1987). Noise, unattended speech and 312-363.
short-term memory. Ergonomics, 30, 1185-1194. Wechsler, D. (1977). Wechsler Intelligence Scale for Children-
Sattler, J. M. (1982). Assessment of childrens' intelligence and Revised. New York: Psychological Corporation.
special abilities (2nd ed.). Boston: Ailyn & Bacon.
Seidenberg, M. S. (1990). Dyslexia in a computational model of
word recognition in reading. In P. B. Gough, L. C. Ehri, & R.
Received August 15, 1994
Treiman (Eds.), Reading acquisition (pp. 243-273). Hillsdale, Revision received June 12, 1996
NJ: Erlbaum. Accepted June 13, 1996 •