ArticlePDF Available

Rhyme, Language, and Children's Reading

Authors:

Abstract

It has been shown that there is a strong relation between children's phonological skills and the progress that they make in reading. But there is some uncertainty whether this is a specific connection or whether it is just a byproduct of variations in general language ability. We report evidence from a longitudinal study showing that the relation between children's sensitivity to rhyme and alliteration and their success in reading is highly specific and cannot be accounted for in terms of general language ability. In this study measures were taken of a group of children's linguistic and metalinguistic skills when they were 3 and 4 years old. The linguistic measures were of the children's vocabulary, their receptive and expressive use of grammar, and their ability to imitate sentences. The metalinguistic measures were of their ability to detect rhyme and alliteration and of their awareness of syntax. Two to three years later, when the children were 6;7, we measured their progress in reading and spelling. The children's rhyme and alliteration scores were related to their reading two years later even after controls for differences in linguistic skills and also for differences in intelligence and in social background. The other metalinguistic task – syntax awareness – did not predict reading after these controls. Awareness of rhyme, we argue, makes a distinctive contribution to reading by helping children to form spelling categories.
Applied Psycholinguistics
11
(1990), 237-252
Printed in the United States of America
Rhyme, language, and children's
reading
PETER BRYANT, MORAG MACLEAN, and LYNETTE BRADLEY
University
of Oxford
ADDRESS FOR CORRESPONDENCE
Peter Bryant, Department of Experimental Psychology, University of Oxford,
South Parks Road, Oxford 0X1 3UD, England
ABSTRACT
It has been shown that there is a strong relation between children's phonological skills and the
progress that they make in reading. But there is some uncertainty whether this is a specific
connection or whether it
is
just a byproduct of variations in general language ability.
We
report
evidence from a longitudinal study showing that the relation between children's sensitivity to
rhyme and alliteration and their success in reading is highly specific and cannot be accounted
for in terms of general language ability. In this study measures were taken of a group of
children's linguistic and metalinguistic skills when they were 3 and 4 years old. The linguistic
measures were of the children's vocabulary, their receptive and expressive use of grammar, and
their ability to imitate sentences. The metalinguistic measures were of their ability to detect
rhyme and alliteration and of their awareness of syntax. Two to three years later, when the
children were 6;7,
we
measured their progress in reading and spelling. The children's rhyme and
alliteration scores were related to their reading two years later even after controls for differences
in linguistic skills and also for differences in intelligence and in social background. The other
metalinguistic task - syntax awareness - did not predict reading after these controls. Awareness
of rhyme, we argue, makes a distinctive contribution to reading by helping children to form
spelling categories.
Preschool children are reasonably good at producing rhymes and at detect-
ing whether words rhyme or not (Chukovsky, 1963; Dowker, 1989; Lenel &
Cantor, 1981; MacLean, Bryant, & Bradley, 1987). Their ability to do so is
interesting for two reasons. First, it demonstrates that preschool children
can analyze the constituent sounds in words. If they understand that cat and
hat rhyme, they recognize that these two words have a speech segment - the
rime /at/ - in common.
The second reason why young children's rhyming abilities deserve atten-
tion is that there is a powerful connection between these abilities and the
progress that children make later on when they learn to read and write. The
performance of
3-,
4-, and 5-year-old children, who cannot yet read, in tests
of sensitivity to rhyme and alliteration predicts their success in learning to
read over the next three to four years even after controls for differences in
©1990 Cambridge University Press 0142-7164/90 $5.00 + .00
Applied Psycholinguistics 11:3 238
Bryant et al.: Rhyme, language, & reading
intelligence, vocabulary, and social background (Bradley & Bryant, 1983,
1985;
Bryant, MacLean, Bradley, & Crossland, 1990; Ellis & Large, 1987;
MacLean et al., 1987). Furthermore, this connection is a specific one. The
children's early rhyme scores predict their progress in reading and spelling
later on but not in arithmetic (Bradley & Bryant, 1983; MacLean et al.,
1987).
One possible reason for a connection between rhyme and reading is that
children's experiences with rhyme allow them to form categories of words
that share common sounds, and that later on they make a connection be-
tween these categories and words that share common spelling patterns. The
knowledge that
light,
fight, and might rhyme might help children grasp the
fact that the words share a common spelling sequence as well as a common
rhyme. There is evidence that children do link rhyming categories with
spelling categories as soon as they begin to learn to read and spell
(Goswami, 1986, 1988).
The fact that measures of sensitivity to rhyme, taken before children learn
to read, predict their progress in reading suggests a causal connection. The
idea of a causal link is supported to some extent by the discovery that
training in rhyme does have a beneficial effect on reading (Bradley & Bry-
ant, 1983). However, it is always possible that a predictive relationship might
not be a causal one: the relationship might be due to the fact that both the
predicting and the predicted variables are determined by some other, un-
known, and therefore unmeasured, factor.
Recently Bowey and Patel (1988) made the interesting suggestion that
children's general language abilities might determine both children's sensi-
tivity to rhyme and also their reading performance. They ran a correlational
study with 60 children whose average age was 6 years. They gave these
children an oddity rhyme and alliteration test based on the one used by
Bradley and Bryant (1983). They also gave the children a syntactic awareness
task, in which they had to "detect and correct" mistakes in ungrammatical
sentences that were read to them. Bowey and Patel refer to the rhyme and
alliteration tests and the syntax awareness task as "metalinguistic" tasks,
since both involve a degree of awareness to an aspect of
language.
As well as
this,
the children were given two standardized linguistic tests, one a test of
vocabulary (the well-known PPVT) and the other a sentence imitation test
(a subtest of TOLD-P) in which they had to repeat a set of sentences of
increasing semantic difficulty. Bowey and Patel also gave the children a
standardized reading test. They did not measure the children's IQ.
The experimenters found a significant relationship between the two meta-
linguistic tests, rhyme and syntactic awareness, and the children's reading
levels.
But they also showed that this relationship was no longer significant
when the effects of differences in the children's performance in the two
linguistic tasks, vocabulary and sentence imitation, were partialled out. On
the other hand, these two linguistic tests were significantly related to the
children's reading comprehension even after differences in the two metalin-
guistic tasks were controlled. Bowey and Patel concluded that the most
important variable here was the children's general language ability, as mea-
Applied Psycholinguistics 11:3 239
Bryant et al.: Rhyme, language, & reading
sured by the two linguistic tests, and that this factor controlled the children's
rhyming and syntactic abilities as well as their success in reading.
Bowey and Patel's data are extremely valuable and interesting, but they
need to be extended. The children in Bowey and Patel's study were relatively
old, and the study was not a longitudinal one. In studies that have shown a
connection between rhyme and reading, the rhyme measures were given
when the children were 3, 4, or
5
years old. There is also evidence from two
studies (Lundberg, Frost, & Petersen, 1988; Stanovich, Cunningham, &
Cramer, 1984) that by
6
years the relationship between scores in a rhyme task
and in reading tests tends to be
low.
So it is quite possible that tests of rhyme
given to 4-year-old children would continue to predict children's reading
later on, even after the stringent controls for general language ability which
are quite rightly advocated by Bowey and Patel.
We needed a study that measured children's sensitivity to rhyme (and also
their ability to correct and detect mistakes in ungrammatical sentences) at an
earlier age (say 4 years), which included measures of IQ and of social
background, and which, pace Bowey and Patel, introduced a wide range of
language measures including standardized tests of the children's receptive
and expressive language, of their ability to understand sentences of varying
grammatical complexity, and of their ability to imitate sentences.
We now report a longitudinal study of 65 children, which included all
these tests as well as measures of the children's progress in reading and
spelling over the following two years.
METHODS
Subjects
Ages. There were 66 children in this project, but we shall report data on 65.
One child left the country halfway through the project. All but one of the
children came from native English-speaking backgrounds. The exception
was a boy whose mother is Swedish; although English is the language spo-
ken in his home, he knew a certain amount of Swedish as well.
At the time of the first tests that we shall report, the average age of the 65
children (31 boys, 34 girls) was 3;4 (SD = 2.8 months; range = 2;10-3;9).
We shall report data over a period of 3 years; when the last measure was
taken, the average age of the 64 children was 6;7 (range 5;9-6;10).
All the children were at school by the time the project ended. In all, the
children were distributed over 28 different schools. The practices involved in
teaching reading varied between schools and even between different class-
rooms in the same school. However, on the whole the teachers adopted a
mixture of "phonic" and "whole word" approaches in teaching children how
to read words.
Intelligence
levels.
At 4;3 the children were given the full WPPSI. The mean
IQ was 111.1 (SD = 12.31). At 6;7 (range =
6;2-7;l)
the children were
Applied Psycholinguistics 11:3 240
Bryant et al.: Rhyme, language, & reading
given the short version of the WISC/R either just before or just after the
final session. The four subtests given were Similarities, Vocabulary, Block
Design, and Object Assembly. The mean IQ was 112.2 (SD = 16.49).
Social
background.
The children came from a wide range of backgrounds.
Our measures of the home background included social class and the educa-
tional level of the parents. We decided to use mothers' educational level as
our main measure of the children's background. We could not use social
class because the project included several single-parent (mother) families to
whom we decided not to apply the social class index because it is based on
the father's occupation.
PROCEDURE
The data for this article come from a longitudinal study of the 65 children,
which started when their mean age was 3;4 and finished when it was 6;7.
Parts of this study have been reported in other papers (Bryant et al., 1990;
Kirtley, Bryant, MacLean, & Bradley, 1989; MacLean et al., 1987). We shall
report data from six sessions, when the children were 3;4, 3;5, 4;5, 4;7, 4;11,
and 6;7.
In the first five of these sessions, we measured the children's linguistic
abilities, and also their sensitivity to rhyme and alliteration and their aware-
ness of syntax. In the last session we measured their reading and spelling
levels.
We shall describe the three kinds of measures - of linguistic ability, of
metalinguistic skills (rhyme, alliteration, and syntax awareness), and of
reading and spelling levels - separately.
Linguistic measures
Vocabulary.
When the children were 3;4,
we
administered the British Picture
Vocabulary Scale (BPVS; the most modern British version of the Peabody
Picture Vocabulary test, it is standardized in Britain). The mean ratio score
for the group on the BPVS was 104.7 (SD = 12.68). The average for the
population was 100.
Expressive
and
receptive
language.
One month later we administered the
Reynell Developmental Language Scale (Reynell, 1983). This test assesses
expressive language and receptive language (verbal comprehension)
separately.
In the expressive part of the test, the child has to talk about scenarios
acted out with toys or about scenes in pictures. For example, the child sees a
boy who is sitting by a table being knocked off his chair and has to say what
happened. The child's responses are scored in terms of their grammatical
sophistication (the use of pronouns, prepositions, and tenses).
The receptive scale requires no speech from the child. The child is given
Applied Psycholinguistics 11:3 241
Bryant et al.: Rhyme, language, & reading
some toys and has to respond to sentences of increasing difficulty about
them. An example of a simple sentence is "Where is the chair?" In a slightly
more complex question, "Which one do we cut with?", the child has to
choose from a bed, chair, pot, pen, and knife. A difficult sentence is "Put
one of the little pigs beside the man."
The mean standardized Expressive score was 0.937 (SD = 0.797) and the
mean Receptive score was 0.988 (SD = 0.863). The average standardized
score for the population was 0 in both cases.
Sentence
imitation.
When the children were
4;
11
we gave them a sentence
imitation task.
We
read out a series of
12
sentences of increasing complexity,
beginning with "Dave likes cars" and ending with "Tom drank his milk
because he likes to play with Bart when he's home." Our instructions to the
child were: "I want you to say just what I say." This task was based on one
devised by Jorm, Share, Maclean, and Matthews (1984) and differed from
their task in one detail only. We substituted the sentence "He drank a glass
of cold milk" for "He drank a glass of cold beer" because our children were
younger than those tested by Jorm and colleagues and might have been less
familiar with beer. The mean score in this task was 7.0 (SD = 2.12).
Metalinguistic measures
Rhyme and alliteration oddity
tasks.
In the session when the children were 4;7
we gave them a version of the rhyme oddity task that was used in a previous
study (Bradley & Bryant, 1983). Before the trials began, we asked the chil-
dren if they knew the nursery rhyme "Jack and Jill" and recited the first two
lines:
"Jack and Jill went up the ... yes
hill,
Jill, hill, they sound the same,
they rhyme, can you tell me another word that sounds like
hilP.
. . . fill.
Now we're going to play a game about words that sound the same, about
words that rhyme."
In the test proper we used pictures in order to remove the memory load.
This was necessary because each trial involved three words. The rhyme test
consisted of two practice trials and then ten experimental trials. In each trial
the child was given three words with pictures, two of which rhymed while the
third did not
(e.g.,
peg,
cot, leg;
fish,
dish, book). The children's task was to
tell us the one that did not rhyme. Their mean scores were 6.22 (SD = 2.63)
out of
10
in this test (chance level = 3.33).
We also measured the child's sensitivity to alliteration using the same
methods. The children had to judge which of three words began with a
different sound (e.g., pin, pig,
tree).
The children's mean scores were 6.53
(SD = 2.44) in this alliteration oddity test.
Syntax
awareness.
Our task was a slight adaptation of one developed by
Tunmer, Nesdale, and Wright (1987), and it was very similar to the task used
by Bowey and Patel. The children were told "I've got some puppets here who
haven't quite learned how to talk properly. What they say is wrong. I want
you to show them what to say." They were then introduced to one puppet
who produced eight sentences (two of which were practice items) in which
Applied Psycholinguistics 11:3 242
Bryant et al.: Rhyme, language, & reading
there was a missing morpheme. Two such sentences were "Andrew drink
juice every day" and "Sandra is paint a picture." After that they were intro-
duced to another puppet who produced eight sentences in which the word
order was incorrect (e.g., "Patted Bill the dog" and "Susan the bike rode").
Each trial was scored as correct or incorrect. We adopted Tunmer and
colleagues' "strict" procedure. For example, for "Sandra is paint a picture"
we accepted "Sandra is painting a picture" as correct, but judged "Sandra
paints a picture" as incorrect because in this case the child did not produce
the correct missing morpheme.
The missing morpheme questions were a great deal easier than the
jumbled order questions. The mean scores for the two sets of questions were
respectively 47.25% correct (SD = 26.86) and 31.57% correct (SD =
25.50).
One possible reason for this difference is that the jumbled order
questions make particularly difficult demands on the children's memory. We
decided to combine the scores for the missing morpheme and jumbled order
sentences for two reasons. One was that the distribution of the combined
score was much nearer normality than the two separate scores. The second
was that each of the separate scores was only based on six trials, and we felt
that this was not a sufficient base for a predictive test. The mean combined
score was 37.05% correct (SD = 25.60). Bowey and Patel combined missing
morpheme questions with other questions in their test and used a combined
score in their analyses.
Reading
When the children had a mean age of 4;5 (i.e., 2 months before we gave our
first metalinguistic tests), we gave them some words to read to check whether
any of them were precocious readers. The standardized reading tests were
plainly too difficult to be appropriate, and so we gave the children twelve
simple, highly frequent words to read. The words were: on, the, car, dog,
and, my, girl, was, boy, you, put. Only 13 children could read any words at
all.
Of these, 5 could read one word only, and 8 could read three words or
more.
In the final session (age 6;7) we gave the following three standardized
tests:
1.
France Primary Reading Test: a multiple choice test with 48 items ar-
ranged in ascending difficulty to assess the understanding of words and
simple sentences. The items that involve sentences depend on the child
understanding the meaning of these sentences, and so this test measures
the child's comprehension as well as his or her ability to read single words.
The group's mean reading age on the test was 7;6 (SD in months =
17.24).
2.
Schonell Graded Word Reading Test: involved reading single words. The
group's mean reading age on the test was 7;2 (SD in months = 15.27).
3.
Schonell Spelling: tested spelling, using Form A. The group's mean spell-
ing age on the test was 6;4 (SD in months = 14.4).
Applied Psycholinguistics 11:3 243
Bryant et al.: Rhyme, language, & reading
RESULTS
Correlations
Our first step was to check that the individual phonological and linguistic
tests that we gave the children in the first half of the project were related to
their reading and spelling level at the project's end and to look at the other
correlations between the different variables. Table 1 gives the correlation
matrix. The table shows several interesting patterns.
First, the relations of the language measures, and particularly of the
Expressive and Receptive (Reynell) scores, to the metalinguistic measures are
quite high. This supports Bowey and Patel's suggestion that the metalinguis-
tic tests may be no more than a measure of language proficiency.
The second interesting point is that the correlations between the metalin-
guistic variables and the reading/spelling measures were appreciably higher
than the relationships between the language measures and reading and spell-
ing. The correlations between the metalinguistic measures and reading and
spelling varied in size from .63 to .80. The correlations of the language
measures with reading and spelling ranged from .31 to .59. The relative
strength of these correlations suggests that rhyme and alliteration should be
related to reading and spelling even after controls for differences in the
children's scores in the other linguistic tests.
Third, there is the question of whether the linguistic and the metalinguistic
scores measure the same thing as intelligence
tests.
It should be noted that the
two IQ scores correlate well with both kinds of measure, and with the reading
and spelling scores as well. Thus, any relationship between the linguistic or
metalinguistic measures and reading could be attributed to the variance that
the linguistic and metalinguistic variables share with intelligence. As a result,
it is essential to carry out fixed order multiple regressions that partial out the
influence of differences in intelligence before charting the relationship be-
tween the linguistic and metalinguistic scores and reading.
Multiple regressions
Our next question was whether the children's early rhyme and alliteration
scores and their performance in the syntax awareness test would predict their
reading and spelling levels later on, even after the effects of differences in
general linguistic ability had been partialled out.
We
ran two sets of multiple
regressions to see whether this relation was significant.
Regressions without the social background and
IQ
scores. The regressions in
the first set were very similar to the ones run by Bowey and Patel. They did
not enter IQ or social background into their regressions and nor did we in
this first set, which is presented in Table 2. The first five steps in all of these
regressions were the same. We simply entered the children's age as the first
step,
and then we entered our four measures of general linguistic ability as
the next four steps. The sixth and final step in each regression was one of the
three metalinguistic measures - rhyme, alliteration, or syntax awareness.
Applied Psycholinguistics 11:3
Bryant et al.: Rhyme, language, & reading
Table 1. Correlations between the main variables in the study
244
Mother's education
WPPSI (WP)
WISC (WS)
Vocabulary (Voc)
Expressive language (EL)
Receptive language (RL)
Sentence imitation (SI)
Syntax awareness (SA)
Alliteration (Al)
Rhyme (Rh)
Reading (France) (R-F)
Reading (Schonell) (R-S)
Spelling (Sp)
Wp Ws Voc
.62 .47 .43
.71 .35
.33
EL
.58
.64
.47
.38
RL
.50
.57
.40
.31
.70
SI SA Al Rh R-F
.44 .65 .53 .59 .63
.43 .66 .66 .58 .71
.34 .59 .60 .59 .67
.34 .45 .29 .31 .45
.48 .64 .55 .52 .59
.50 .59 .36 .44 .46
.57 .51 .45 .50
.59 .53 .63
.75 .77
.69
R-S
.58
.67
.68
.41
.55
.40
.51
.63
.80
.69
.95
Sp
.61
.66
.67
.39
.47
.34
.45
.63
.73
.64
.84
.89
Table 2. R2 change in nine fixed order multiple regressions
Steps in
regression
Extraneous variables
1.
Age at test of reading
Linguistic variables
2.
Vocabulary (BPVS)
3.
Language (Reynell) expressive
4.
Language (Reynell) receptive
5.
Sentence imitation
Final step
6a. Final step rhyme
6b.
Final step alliteration
6c.
Final step syntax awareness
Reading
France test
.003
.201***
.122**
.021
.144***
.219***
.046*
Outcome measure
Reading
Schonell test
.031
.156***
.115**
.140***
.030
.152***
.251***
.062*
Spelling
Schonell test
.039
.141**
.079*
.102**
.019
.149***
.231***
.120***
Note: In each regression the children's general language scores are entered before the
metalinguistic score. Each of the three outcome measures is analyzed three times; in
each analysis one of the three metalinguistic measures represents the final step.
*p < .05; **p < .01; ***/? < .001.
Applied Psycholinguistics 11:3 245
Bryant et al.: Rhyme, language, & reading
There were three different outcome measures, or dependent variables,
which were the children's scores in the France reading test, in the Schonell
reading test, and the Schonell spelling test when they were 6;7. This meant
that there were nine fixed order multiple regressions in all (three different
final steps with three different outcome measures).
Table 2 shows that our results are similar to Bowey and Patel's in one way
but different in another. The results of the two studies are similar in that the
linguistic measures in both predict reading well. Bowey and Patel reported
that their linguistic measures accounted for 41% of the variance in one
reading test and 29% in another. Our linguistic measures were even more
powerfully connected to reading. As Table 2 shows, they accounted for
45.8%
of the variance in the France reading test and
43.1%
in the Schonell
reading test. Ours probably accounted for more variance because they cov-
ered a wider range of linguistic behavior. All the linguistic tests were reason-
able predictors of reading. The fact that the Sentence Imitation scores did
not account for a significant portion of the variance in reading is merely a
result of our entering it as the last linguistic measure in the regression. In
other regressions we entered it at an earlier stage and then it became
significant.
However, when we consider the final step in each of the multiple regres-
sions,
we can see that our results were quite different from Bowey and
Patel's. Table 2 shows that all three metalinguistic scores did predict reading
even after differences in the children's general linguistic abilities had been
partialled out. There is a strong connection between 4-year-old children's
sensitivity to rhyme, to alliteration, and to syntax and their reading two
years later, which cannot be explained away as a mere symptom of a more
general linguistic ability.
Regressions with the social background and
IQ
scores. However, this first set
of regressions did not include IQ scores or measures of differences in social
background. So it is possible that the connections between some of our
measures and reading could simply have been a byproduct of differences in
these important but extraneous variables. We needed multiple regressions in
which these variables were entered before either the linguistic or the metalin-
guistic measures. The second set of nine multiple regressions, which are
presented in Table 3, took this form.
The regressions had two extra steps. The children's social background
(measured by their mothers' educational level) was entered as the second
step.
In these regressions we treated mothers' educational level as a categori-
cal and not a continuous variable. The children's IQ was entered as the third
step.
We had two IQ scores for the group - the WPPSI scores when they
were 4;3 and the WISC scores when they were 6;7. We decided to use the
WPPSI scores because the test was given in the early part of the project at
the time of the other predictive measures. (Table 1 shows that the WPPSI
and the WISC scores were equally strongly related to reading.) Otherwise,
the multiple regressions were the same as before.
It can be seen from Table
3
that the presence of the social background and
Applied Psycholinguistics 11:3
246
Bryant
et
al.: Rhyme, language, & reading
Table 3. R2 change in fixed order multiple regressions in which social background
and IQ are controlled as well as general language abilities
Outcome measure
Steps in
regression
Extraneous variables
1. Age at test of reading
2.
Mothers' educational level
3.
IQ(WPPSI)
Linguistic variables
4.
Vocabulary (BPVS)
5.
Language (Reynell) expressive
6. Language (Reynell) receptive
7.
Sentence imitation
Final step
8a. Final step rhyme
8b.
Final step alliteration
8c.
Final step syntax awareness
Reading
France test
.003
444***
.178**
.008
.000
.016
.004
.048**
.085***
.001
Reading
Schonell test
.031
.392***
.198***
.002
.000
.026
.009
.056**
.110***
.000
Spelling
Schonell test
.039
.413***
.158***
.000
.000
.007
.002
.034*
.069***
.000
*p < .05; **p < .01; ***p < .001.
the IQ measures
in
these regressions makes
a
striking difference
in
two ways.
First, these
two
variables account
for a
large proportion
of the
variance.
Second, the linguistic variables
are no
longer good predictors
of
reading
and
spelling.
In
fact,
the
vocabulary
and the
expressive
and
receptive language
measures, which were significantly related
to
reading
and
spelling
in the
first
set
of
regressions,
now
account
for a
tiny proportion
of the
variance
in
reading
and
spelling
and
fall
far
short
of
significance.
So it
seems that
the
previous connection between these linguistic measures
and
reading could
simply have reflected differences
in
social background
or in
intelligence,
since these variables were
not
controlled
in the
first
set of
regressions.
Exactly the same point
can
be made about one
of
the metalinguistic scores
- syntax awareness. This
was an
excellent predictor
of
reading
and
spelling
in
the
first
set of
regressions.
In the
second
set it
accounted
for
hardly
any
variance
at all in the
children's reading
or in
their spelling.
But
our
third point about these regressions
is
that
the
other metalinguistic
measures
-
rhyme
and
alliteration
-
withstood
the
effect
of
partialling
out
differences
in
social background
and IQ, as
well
as in
general language
abilities, extremely well. Even after controls
for all
these variables, both
rhyme
and
alliteration accounted
for a
significant proportion
of the
vari-
ance
in all
three outcome measures.
The
children's alliteration scores were
particularly powerful predictors
of
reading
and
spelling.
We
should like
to
make
a
final remark about
the
second
set of
regressions.
Applied Psycholinguistics 11:3 247
Bryant et al.: Rhyme, language, & reading
The combination of the "extraneous variables" and the linguistic and meta-
linguistic measures accounts for an impressive amount of the variance in the
children's reading and spelling. For example, the multiple regressions in
which the final step was the children's alliteration scores accounted for
73 %
of the variance in the France reading scores, 78% in the Schonell reading
scores, and
68%
in the Schonell spelling scores.
Precocious readers
and
outliers.
Our main concern in this article has been
with the possible importance of the age at which the children's sensitivity to
rhyme is tested. Our argument is that tests given in the preschool years are
likely to be more powerful predictors of reading than tests given after chil-
dren have reached school. However, this raises the question whether tests of
rhyme and alliteration are a powerful predictor of reading when they are
given only to children who cannot yet read. It is certainly important to check
whether the powerful relationships that we have reported between the rhyme
and alliteration measures and reading two years later would be the same if
we confined our analysis to children who could not read at the time of the
original rhyme and alliteration tests. So we ran further analyses in which we
excluded children who, on the basis of the results of the simple reading test
that we gave to the children when they were 4;5, appeared to be precocious
readers at the time of the original tests of rhyme and alliteration.
As we reported earlier, 8 children could read three or more words in our
first reading test. We decided to exclude these 8 children from our analysis
on the grounds that they must have made definite progress in reading in
order to be able to read that many words. However we felt we could not
count the 5 children who only managed to read one word in our test as
precocious readers, because being able to read only one word of the simple
and very common words in the test did not seem to us enough to warrant
that description. So we did not exclude those 5 children from our further
analyses, which were designed to establish whether the relationships between
the metalinguistic measures and reading would be the same when the sample
consisted only of children who had made no substantial progress in reading
at the time of the metalinguistic tests. The new sample consisted of 57
children.
Table 4 gives the correlation matrix, and Table 5 gives the equivalent
multiple regressions to those already presented in Table 3, for this reduced
sample. The pattern of correlations in Table 4 was very similar to the corre-
lations obtained from the total sample. The correlations between rhyme and
alliteration were slightly smaller in the smaller sample, but still very strong.
Table 5 shows that rhyme and alliteration continued to predict reading and
spelling after the effects of differences in age, social background, IQ, and of
the various linguistic measures were removed. Thus, the relationships that
we found in our total sample are not the product of differences in reading
skills in the preschool period.
Another possibility is that the very strong relation between our rhyme and
alliteration scores and reading could be explained partly by the existence of
"outliers," i.e., children not particularly representative of the rest of the
Applied Psycholinguistics 11:3 248
Bryant et al.: Rhyme, language,
&
reading
Tkble 4. Correlations between the main variables in the study after exclusion of
precocious readers (N = 57)
Wp Ws Voc EL RL SI SA Al Rh R-F R-S Sp
Mother's education
WPPSI (WP)
WISC (WS)
Vocabulary (Voc)
Expressive language (EL)
Receptive language (RL)
Sentence imitation (SI)
Syntax awareness (SA)
Alliteration (Al)
Rhyme (Rh)
Reading (France) (R-F)
Reading (Schonell) (R-S)
Spelling (Sp)
.58 .61 .35 .55 .53 .36.61 .40 .50 .50
.67 .31 .62 .56 .38 .63 .61 .53 .69
.26 .42 .38 .26 .56 .49 .51 .56
.32 .38 .33 .43 .16 .23 .36
.74 .47 .66 .51 .49 .59
.49 .59 .36 .44 .49
.54 .46 .40 .45
.55 .49 .60
.70 .71
.62
.43 .50
.63 .67
.57 .60
.31 .34
.55 .51
.40 .33
.46 .41
.58 .59
.74 .70
.62 .58
.93 .82
.87
Table 5. R2 change infixed order multiple regressions of the 57 children who were not
precocious readers
Steps in
regression
Extraneous variables
1.
Age at test of reading
2.
Mothers' educational level
3.
IQ (WPPSI)
Linguistic variables
4.
Vocabulary (BPVS)
5.
Language (Reynell) expressive
6. Language (Reynell) receptive
7.
Sentence imitation
Final step
8a. Final step rhyme
8b.
Final step alliteration
8c.
Final step syntax awareness
Reading
France test
.000
.284***
.272**
.005
.030
.001
.003
.041**
.090***
.000
Outcome measure
Reading
Schonell test
.011
.216***
.287***
.001
.030
.002
.009
.060*
.121***
.000
Spelling
.001
.258***
.275***
.003
.017
.029
.004
.036*
.076**
.000
Note: Social background and IQ are controlled as well as general language abilities.
*p < .05; **p < .01; ***p < .001.
Applied Psycholinguistics 11:3 249
Bryant et al.: Rhyme, language, & reading
sample. To find who could be counted as outliers, we used standardized
residuals from the regression of the alliteration score and the Schonell read-
ing level (the strongest relation of all in our total sample). This analysis
produced 5 children who were clear outliers. We ran further analyses in
which these
5
children were excluded.
The exclusion of these outliers led to even stronger correlations than
before between rhyme and alliteration on the one hand and reading and
spelling on the other. The correlations of rhyme to the France reading test,
to the Schonell reading test, and to the spelling test were, respectively, .71,
.73,
and .68. The equivalent figures of alliteration were .79, .83, and .76.
Not surprisingly, given these correlations, the exclusion of the five outliers in
further multiple regressions made no difference to the relationships of
rhyme and alliteration to reading or to spelling that we reported in Tables 2
and 3. These relationships cannot be attributed to the presence of outliers.
DISCUSSION
Our study produced four main results. First, the measures that we gave the
children when they were 3 and 4 years old accounted for an extremely high
proportion of the variance in their reading single words, their comprehen-
sion, and their spelling at the age of 6;7.
Our second discovery was that the children's scores in rhyme and allitera-
tion tests given to them at 4 years predicted their reading and spelling levels
at 6 years after stringent controls for general language ability as well as for
social background and intelligence. The connection between children's sen-
sitivity to rhyme and alliteration at 4 years is not a mere byproduct of some
broader linguistic ability.
Third, the children's scores in our other metalinguistic test - the syntax
awareness task - were also related to their reading after the differences in
general language ability had been partialled out, but this relationship
dropped out when social background and IQ were entered into the equation.
So we conclude that the connection between awareness of syntax and read-
ing merely reflected differences in these extraneous variables.
Our fourth result concerns the linguistic measures (vocabulary, expressive
and receptive language, and sentence imitation). The children's scores in
these tests were related to their reading levels, but this relationship also
became non-significant after the two extraneous variables, social back-
ground and intelligence, had been included in the regression. It seems that
the connection between the linguistic measures and reading was simply due
to the fact that both were determined by differences in intelligence and in
background. Of course we cannot be sure whether the same pattern would
have applied to the children tested by Bowey and Patel, because they gave
their language measures to 6-year-olds, whereas we gave ours to 4-year-olds.
There is a clear need to repeat Bowey and Patel's study with 6-year-olds, but
this time to include measures of IQ and social background.
The clearest difference between our results and those of Bowey and Patel
is in the power of the rhyme and alliteration tests as predictors of reading.
Applied Psycholinguistics 11:3 250
Bryant et al.: Rhyme, language, & reading
Their rhyme and alliteration measures failed to predict reading after controls
for general linguistic ability. In complete contrast, there was a highly signifi-
cant relationship between our measures of rhyme and alliteration after simi-
lar controls. There seem to be two reasonable explanations for this differ-
ence.
One, which we favor (mentioned in the introduction), is that our
rhyme and alliteration measures were given to 4-year-olds, while Bowey and
Patel's were given to 6-year-olds.
It may be much more important, from the point of reading, to know how
aware a child is of rhyme at 4 years than at 6 years. Why should this be so?
Our answer is that at 4 years one gets a relatively pure measure of the child's
ability to categorize words by sounds, because he has not begun to read. At
6 years, on the other hand, the child will have had quite a lot of experience
with reading, and his judgments about rhyme might now be affected by his
knowledge about spelling. So, children might use their knowledge that two
words do or do not have a common spelling pattern to help them judge
whether the words rhyme. Thus, when the test is given to preschool children,
one is measuring their ability to form categories that will help them to read
later on, but when it is given to older children one might be picking up
effects of the experience of reading as well.
The second possibility is that the difference in results is due to the fact
that our tests of rhyme and alliteration took a different form than Bowey
and Patel's. In our test the odd word had no phonemes in common with the
other two words
{peg,
cot,
leg;
pin, pig,
tree).
In theirs, the odd word had
one phoneme in common with the other words (sun, gun, rub; hat, cot, pot;
bun,
rug, bus). There is evidence that our version is easier than theirs (Lenel
& Cantor, 1981), and our reason for adopting the easier version was that we
were working with young children and wanted to be sure that the task would
not be too difficult for them. The fact that we adopted an easy test of rhyme
and alliteration, while Bowey and Patel adopted a hard one could explain the
difference in results, but it is hard to see why. Neither study suffered from
floor or ceiling effects, and there seems no reason why an easy test should be
a better predictor of reading than a difficult one. In fact, the overall error
rate in the rhyme and alliteration tasks seems to have been much the same in
the two studies, presumably because the children in the Bowey and Patel
study who were given the harder task were also two years older than the
children in our study. So we prefer the first possible explanation to the
second.
The fact that the odd word shared a phoneme with the other two in Bowey
and Patel's task raises another issue. Throughout their paper, they refer to
their rhyme and alliteration tests as "phonemic awareness" tests. They use
this term because one phoneme told the odd word apart from the other two
in their tests. For example with sun, gun, and rub, the first two words share
a final phoneme /n/ which the other words does not possess. But, in our
view, this does not mean that the children were responding to the presence or
absence of a single phoneme. There is good evidence that children of this
age are aware of a speech unit called the "rime" (Kirtley et al., 1989;
Applied Psycholinguistics 11:3 251
Bryant et al.: Rhyme, language, & reading
Trieman, 1985). In single syllable words like sun, the rime consists of the
vowel and the following consonant - un. Kirtley and colleagues showed that
6-year-old children find it extremely difficult to say which is the odd word in
mop,
lead,
and whip, and yet quite easy in top,
rail,
and hop. In the former
case,
the odd word can be detected only on the basis of the fact that it lacks
a single phoneme which is present in the other
two.
In
sun,
gun, and rub (the
Bowey and Patel example) two words share a common rime that the odd
word out does not possess. So it is likely that the children who produced the
correct answer in Bowey and Patel's rhyme tasks did so on the basis of the
words' rimes, which contained two phonemes, rather than on the basis of a
single phoneme.
The point seems an important one to us, because we (Bryant et al., 1990)
hold the view that awareness of rhyme makes a distinctive contribution to
reading. Words that rhyme often share spelling sequences as well (e.g., light,
might, fight). By forming categories of words that rhyme, the child might be
preparing himself for learning spelling categories later on. Note that the unit
here is not the single grapheme-phoneme relation. The connection is be-
tween sounds with more than one phoneme (rimes) and strings of letters.
There is some evidence that this kind of association plays an important part
in learning to read. We have shown (Bryant et al., 1990) that children's
rhyme scores do make a separate contribution to children's reading even
after their scores in phoneme detection tasks have been partialled out, which
suggests that the connection between rhyme and reading concerns more than
the learning of grapheme-phoneme correspondences. In addition, Goswami
(1986,
1988) found that beginning readers make inferences about spelling
patterns on the basis of
rhyme.
If they know how to read the work
beak,
for
example, they will often use this knowledge to work out what a new word
like
peak means.
This idea is speculative. We still need to know a lot more about the
relationship between the children's knowledge of spelling patterns and their
judgments about rhyme. But our data certainly demonstrate that there is a
powerful and specific connection between 4-year-old children's sensitivity to
rhyme and the progress that they eventually make in reading and spelling.
This connection is not a byproduct of differences in general linguistic
ability.
ACKNOWLEDGMENTS
This research was supported by a grant from the Medical Research Council. We are
very grateful to the teachers and staff of several local schools for letting us visit the
schools and see the children there. The schools are Bernwood, Botley, Church Cow-
ley, East Oxford, Greycoates, Headington Quarry, Kennington, Larkrise, New Mar-
ston, Meadow, Our Lady's, Pegasus, St. Andrew's, St. Ebbe's, St. Francis, St. John
Fisher, St. Joseph's, St. Mary and St. John, St. Nicholas, St. Thomas More, Sand-
hills,
Speedwell, Speenhamland, The Crescent, The Queen's Dyke, Windmill,
Wolvercote, and Wood Farm.
Applied Psycholinguistics 11:3 252
Bryant et al.: Rhyme, language, & reading
REFERENCES
Bowey, J. A., & Patel, R. K. (1988). Metalinguistic ability and early reading achievement.
Applied Psycholinguistics, 9, 367-383.
Bradley, L., & Bryant, P. E. (1983). Categorising sounds and learning to read - A causal
connection. Nature, 301, 419-521.
(1985). Rhyme and reason in reading and spelling. I.A.R.L.D. Monographs, 1. Ann Arbor:
University of Michigan Press.
Bryant, P. E., MacLean, M., Bradley, L., & Crossland, J. (1990). Rhyme and alliteration,
phoneme detection and learning to read. Developmental Psychology (in press).
Chukovsky, K. (1963). From two to five. Berkeley: University of California Press.
Dowker, A. (1989). Rhymes and alliteration in poems elicited from young children. Journal of
Child Language, 16, 181-202.
Ellis,
N., & Large, B. (1987). The development of reading: As you seek so shall you find.
British Journal of Psychology, 78, 1-28.
Goswami, U. (1986). Children's use of analogy in learning to read: A developmental study.
Journal of Experimental Child Psychology, 42, 73-83.
(1988). Children's use of analogy in learning to spell. British Journal of Developmental
Psychology, 6, 21-33.
Jorm, A. R, Share, D. L., Maclean, R., & Matthews, R. (1984). Phonological confusability in
short term memory for sentences as a predictor of reading ability. British Journal of
Psychology, 75, 393-400.
Kirtley, C, Bryant, P., MacLean, M., & Bradley, L. (1989). Rhyme, rime and the onset of
reading. Journal of Experimental Child Psychology, 48, 224-245.
Lenel, J. C, & Cantor, J. H. (1981). Rhyme recognition and phonemic perception in young
children. Journal of Psycholinguistic Research, 10, 57-68.
Lundberg, I., Frost, J., & Petersen, O. (1988). Effects of an extensive program for stimulating
phonological awareness in preschool children. Reading Research Quarterly, 23, 263-284.
MacLean, M., Bryant, P. E., & Bradley, L. (1987). Rhymes, nursery rhymes and reading in
early childhood. Merrill-Palmer Quarterly, 33, 255-282.
Reynell, J. K. (1983). Developmental Language Scales. East Windsor: NFER-Nelson.
Stanovich, K. E., Cunningham, A. E., & Cramer, B. R. (1984). Assessing phonological aware-
ness of kindergarten children: Issues of task comparability. Journal of Experimental Child
Psychology, 38, 175-190.
Treiman, R. (1985). Onsets and rimes as units of spoken syllables: Evidence from children.
Journal of Experimental Child Psychology, 39, 161-181.
TUnmer, W. E., Nesdale, A. R., & Wright, A. D. (1987). Syntactic awareness and reading
acquisition. British Journal of Developmental Psychology, 5, 25-34.
Wise, B. W., Olson, D. K., & Treiman, R. (1990). Subsyllabic units as aids in beginning readers'
word learning: Onset-rime versus post-vowel segmentation. Journal of Experimental Child
Psychology (in press).
... While this finding may seem surprising, it may be an artifact of the tasks chosen to assess PA, i.e., Rhyming and Elision. Rhyming abilities are among the first PA skills to emerge and have long been associated with generally strong language skills, particularly with vocabulary (Bryant, MacLean & Bradley 1990). The decreased English vocabulary skills of the ELLs in the present study, therefore, may have had a direct effect on rhyme detection and production skills. ...
Article
Full-text available
Early spelling development is an important area of research as it presents an opportunity for our understanding of how children begin to represent sounds within words through application of letter-sound mapping. The development of spelling is often characterized by stage-like theories, although it has been suggested that children may draw on different patterns of representations ( Integration of Multiple Patterns ). This study examined early spelling acquisition among monolingual English-speaking and ELL kindergartners with Spanish as L1. ELLs receive all academic instruction, including written language, in English (L2) while continuously using oral L1 for all other communications. As languages differ in their linguistic and structural composition, our aim was to identify possible influences of oral L1 on spelling abilities in L2. We were particularly interested in representation of sublexical units within the words, as English and Spanish differ in their intrasyllabic awareness. Our results showed expected differences in psycholinguistic profiles of language groups as well as differences in allocation of resources in spelling attempts. While there were no statistically significant differences in representation of sublexical units between language groups, there were notable differences in percentage of correct responses pointing to language specific influences. These findings support the Dual System Model of phonological representations development among bilingual individual and suggests that in the early stages of spelling acquisition in L2, L1 Spanish kindergartners are still influenced by their native language although they use it in their oral modality only.
... Firstly, we cannot rule out the possibility that our participants (children with at least nearly two years of reading and spelling experience) used -besides phonological information -orthographic representations when solving the phonological awareness tasks. This is a general potential confound when measuring phonological awareness in children with reading and spelling experience (see also Bryant et al., 1990;Treiman & Cassar, 1997). The fact that in the current study, the items used in the phonological awareness tasks consisted solely of real German words might, however, have aggravated potential tendencies for referring to the items' spelling, as an orthographic strategy for solving phonological awareness tasks is more often applied with words than with pseudowords (Stuart, 1990). ...
Article
Full-text available
Studies investigating relations between morphological awareness and literacy in German, a language with a rather transparent but asymmetric orthography, are sparse. Little is known about the role of grade level for these relationships and of their relative strength compared to those between other language-related variables and literacy skills. This cross-sectional study was conducted with German-speaking second-, third- and fourth-graders ( n of final sample ≥ 85 per grade). Morphological awareness tasks required the production of inflections, derivations and compounds. Additionally, phonological processing, vocabulary, reading fluency, reading comprehension and spelling were measured. Factor analyses revealed two facets of morphological awareness: morphological fluency and morphological awareness for pseudowords. These were correlated with both reading and spelling skills in all grades. More literacy variables were related to morphological fluency in Grade 4 than in Grades 2 and 3. In regression analyses, variance in literacy skills was predominantly explained by phonological awareness. Morphological awareness did not explain additional variance. The results reveal that different facets of morphological awareness are related to literacy skills in German primary school children. Despite the asymmetry of German orthography, no evidence was found for differences in the association of morphological awareness with spelling versus reading. Phonological processing shows stronger relations with literacy than morphological awareness does. This might indicate that in the transparent German orthography, alphabetic reading and spelling strategies are particularly relevant until the end of Grade 4. Yet, morphological fluency might start to unfold its relevance for reading and spelling near the end of fourth grade in German.
... Detecting phonological patterns such as rhymes is of great importance to children over the course of linguistic development. Being aware of the rhyme relationship between words has been associated with various linguistic abilities: rhythmic awareness (Wood, 2006), articulation (Mann & Foy, 2007), phonological perception (Foy & Mann, 2001) and acquisition of literacy in general (Bryant, Maclean, & Bradley, 1990;Wood & Terrell, 1998). Child songs and nursery rhymes might be the earliest contexts in which infants encounter rhymes in their linguistic input. ...
... Navedeno tumačenje može nam pomoći u objašnjenju donekle kontradiktornih nalaza istraživanja na osnovu kojih se tumače bolja ili lošija postignuća dece u čitanju (npr. Bryant et al., 1990;Roth & Spekman, 1986). U skladu sa ovim činjenicama, pretpostavljamo da će se razumevanje pokazatelja čitanja poboljšati ispitivanjem razvoja leksičke obrade i fonoloških procesa povezanih sa čitanjem. ...
Article
Full-text available
Introduction. According to the lexical restructuring model, the development of vocabulary initiates the development of phonological representations. Therefore, it can be predicted that children with developed vocabulary will possess very specific phonological representations of words. Aim. The aim of this research was to determine the significance of phonological awareness, verbal working memory, and rapid automatic naming as indicators of the development of expressive and receptive vocabulary in preschool children. Method. The research included 86 children of typical development. For the assessment of phonological awareness, verbal working memory, rapid automatic naming, expressive and receptive vocabulary, the subtests of Clinical Evaluation of Language Fundamentals -Fourth Edition battery were used. Results. The results showed that the model containing age, phonological awareness, verbal working memory, and rapid automatic naming predicted 51% of the achievement variance on expressive vocabulary tasks (F = 7.73, df1 = 2, df2 = 75, p < .001) and 38% of the achievement variance on receptive vocabulary tasks (F = 4.65, df1 = 2, df2 = 75, p < .001). In the final model, only the tasks of phoneme identification, phoneme substitution, and rapid automatic naming (p < .05) stood out as statistically significant indicators of expressive vocabulary. On the other hand, statistically significant indicators of receptive vocabulary included analysis and synthesis of phonemes, verbal working memory, and rapid automatic naming (p < .05). Conclusion. The results showed that age, phonological awareness, verbal working memory, and rapid automatic naming significantly contributed to vocabulary development in children before reading acquisition.
... Moreover, less motivation in reading is akin to poor PA (Shen 2014). Researchers demonstrate that the mastery of PA underpins the students' performance of learning to read in English (Bryant et al., 1990). As such, students with weak PA are assumed to encounter barriers in reaching objectives highlighted in the New Standard (MoE PRC, 2011), such as reading instructions, stories and short passages. ...
... Phonological awareness. Phonological awareness, the conscious knowledge of the phonological structure of words (e.g., syllables, onsets, rimes, codas), helps new readers to match sub-lexical phonological units with graphemic ones and is therefore a key predictor of reading success (Blachman, 2000;Bryant et al., 1990;Ehri et al., 2001;Stanovich, 1992;Vellutino and Scanlon, 2001). However, there are age-related differences in the predictive value of types of phonological awareness tests. ...
Article
This longitudinal study documents a trilingual child’s struggle with decoding and word recognition, the remedies sought to help him start reading in his second language (English) while he was in French immersion, and his performance after the intervention on tests of phonological awareness in L1 Romanian, L2 English, and L3 French. The study commenced at age 5;6, when the child, Alex, was in English kindergarten and diagnosed with a reading deficit. The initial diagnostic assessment uncovered his near-complete lack of phonological awareness, a key ingredient of emergent reading. An intervention using a multisensory approach to reading was used twice a week until the child was 7;9, at which point he was completing grade 2 in French immersion. Alex’s phonological processing abilities were assessed in all three languages immediately after remediation in order to determine: (i) whether his phonological processing skills improved in English, the language of the intervention; (ii) whether there were similar effects in the two non-remediation languages (Romanian and French); and, finally, (iii) whether children at-risk for reading difficulties are able to continue their education in an L3, such as French in an immersion context.
Article
Full-text available
This case study aims to explore students’ perceptions of their metalinguistic awareness in L2 writing mediated by Grammarly as online feedback tool. Data garnered from twenty participants’ questionnaire and reflection were analyzed. The findings reveal that the participants perceive the online feedback tool facilitates their drafting process in terms of grammatical awareness. In other words, the online grammar checker could be help to raise their metalinguistic awareness in L2 writing. This study implies that language teacher and teacher educator could utilize this online tool in learning and teaching process in order to raise learners’ metalinguistic awareness as well as learner autonomy. Keywords: Grammarly; L2 writing; metalinguistic awareness; online feedback tool
Article
Background For young children experiencing phonological awareness (PA) difficulties, the need for early and targeted intervention to prevent reading disability is unequivocal. There are very few studies, however, on the efficacy of PA interventions delivered at school. Aims This study examined the impact of an early PA intervention embedded within an oral language program designed for at-risk kindergartners. Methods Using a quasi-experimental pretest/posttest design, at-risk readers from four schools received either the 10-week intervention in small groups, three times a week for 30 min as a supplement to the regular classroom curriculum or served as controls not participating in the intervention and receiving the usual classroom instruction. Results Children in the intervention group demonstrated a greater use of phonological awareness at posttest on overall composites of phonological processing, and on several individual accuracy and fluency measures targeting skills at the phoneme level. Conclusions The results add to accumulating evidence on the efficacy and effectiveness of teacher-delivered school-based early literacy interventions.
Chapter
This chapter provides a theoretical introduction to Online Sex Talk and the Social World, arguing that in order to analyse sexual practices and behaviours in an online community, it can be useful to first frame these within an understanding of the everyday world in relation to social norms, social policing, and transgression. Building on arguments about social norms from sociological theory as well as from diverse branches of feminist and queer theories, this chapter provides a theoretical foundation to the data-driven analysis that follows in the book, contextualising the analysis of covert language use patterns within a system of broader social practices and structures.