SLI Fonološki I Čitanje Bishop 1990

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 25

f. Child Psychol. Psychiat. Vol. 31, No. 7, pp.

1027-1050, 1990 0021-9630/90 $3 00 + 0 00


Printed in Great Britain. Pergamon'Press pic
© 1990 Association for Child Psychology and Psychiatry

A Prospective Study of the Relationship


between Specific Language Impairment,
Phonological Disorders and
Reading Retardation
D. V. M. Bishop* and C. Adams^

Abstract—Language and literacy skills were assessed in 83 8 54 -year olds whose language
development had been impaired at 4 years of age. Provided that language problems had
resolved by age 5 V2 years, literacy development was normal, but many of the children
who still had verbal deficits at 5 V2 years of age did have reading difficulties and persisting
oral language impairments later on. In these children, reading comprehension tended to
be poor relative to reading accuracy. Syntactic competence in the preschool period accounted
for a substantial proportion of the variance in literacy attainments, after allowing for the
effects of non-verbal ability. There were only weak links between expressive phonological
disorders and later ability to read either meaningful text or non-words.

Keywords: Language disorder, reading disability, phonology, prediction

Introduction

Part I. Reading and Spelling Proficiency in Children with a History of


Language Delay
There is good evidence from several sources for a relationship between reading
difficulties and impaired language development. First, longitudinal studies of children
who are identified as having delayed language milestones find relatively low attainments
in reading and spelling when these children are followed up at 7 or 8 years of age
(Fundudis, Kolvin & Garside, 1979; Richman, Stevenson & Graham, 1982; Silva,

Accepted manuscript received 20 March 1990

*Department of Psychology, University of Manchester, Manchester M13 9PL, U.K.


^Present address: Centre of Audiology, Education of the Deaf and Speech Pathology, University of
Manchester, Manchester, U.K.
Requests for reprints to: Dr D. V. M. Bishop, Department of Psychology, University of Manchester,
Manchester M13 9PL, U.K.

1027
1028 D. V. M. Bishop and C. Adams

Williams & McGee, 1987). Second, retrospective studies report that an unusually
high proportion of children with specific reading disability were late in starting to
talk (Kinsbourne & Warrington, 1963; Naidoo, 1972; Rutter, Tizard & Whitmore,
1970). Furthermore, experimental studies of children with specific reading disability
have found evidence of continuing verbal deficits, especially in tasks involving
phonological processing, even in children who have no immediately obvious signs
of language disorder (see reviews by Jorm & Share, 1983; Snowling, 1987).
On the basis of these studies one might predict that a child who has a specific
language impairment in the preschool period will be at high risk for specific reading
disability later on, even if the obvious problems with spoken language have resolved.
In 1987, Bishop and Edmundson reported results from a longitudinal study of
children who were identified as having language impairment at the age of 4 years,
who were followed up to 5 ^ years of age. We report here results from a further follow-
up of this sample carried out between 1986 and 1988, when these children were
8 ^ -years old. The follow-up assessment included tests of reading and spelling, making
it possible to investigate further the nature of the link between early language delay
and later literacy problems.

Method
Subfects

Language-delayed sample
The study sample was described in detail by Bishop and Edmundson (1987a). Between 1982 and
1984, paediatricians and speech therapists were asked to refer children aged between the ages of 3:9
and 4:2 who had any impairment of language development that could not be attributed to low intelligence,
hearing loss, physical defect or bilingual background, and which was not associated with a recognized
syndrome such as infantile autism. Eighty-eight children were recruited to the study, of whom 19 (the
'general delay' subgroup) were identified as having delayed non-verbal development on the basis of a
shortened version of the Leiter International Performance Scale (Leiter, 1948). The remaining children
were designated as having 'specific language impairment' (SLI). All but one of these children were
followed up at the age of 4i/4 years and again at 5 ^ years, by which time 32 of them no longer had
any evidence of language impairment. In line with the terminology adopted by Bishop and Edmundson,
these children will be referred to here as the group with good outcome at 5 years. The remaining children
are classified as having poor outcome at 5 years. In classifying children into these subgroups, we do
not wish to imply that there are sharp divisions between those with good and poor outcomes, or those
with specific and non-specific developmental delays. Indeed, our own analyses support the idea that
differences between these subgroups are quantitative rather than qualitative (Bishop & Edmundson,
1987b). However, the distinction can be of use when considering clinical implications of the study,
where one wishes to ask such questions as whether language-delayed children who appear to have
recovered by 5 J^ years will go on to have literacy problems.
Eighty-three (94.3%) of the children studied by Bishop and Edmundson were seen again at the age
of 8/2 years (mean age 100 months, S.D. 1.2 months). One child had been discovered to have a high
frequency hearing loss at the age of 6 years, and her data are not included here. Parents of the four
remaining children were unwilling for their child to participate further in the study.

Control group
A sample of 30 control 8-year olds, matched with the experimental group in terms of sex ratio (22
boys and 8 girls), was given a shortened version of the test battery to provide normative standards
Language impairment and reading retardation 1029

for some of the unstandardized measures. Twenty of these children came from a primary school in
inner-city Manchester and 10 from a rural town in Lancashire.
In addition, we also made use of WISC-R and reading data gathered by Bishop and Butterworth
(1979, 1980) on an unselected sample of 168 8>/2-year olds. These data were used to compute the
regressions of reading ability on WISC-R subtests, and to predict reading comprehension on the basis
of reading accuracy (see below).

Materials

Non-verbal ability
At the first assessment, when children were aged 4 years, a shortened version of the Leiter International
Performance Scale was administered, scores being transformed to scaled scores on the basis of normative
data from a control group of 4-year olds.
At 8 J4 years of age, two WISC-R performance subtests, picture completion and block design (Wechsler,
1974), were administered.

Receptive vocabulary
The British Picture Vocabulary Scale (BPVS; Dunn, Dunn, Whetton & Pintilie, 1982), a British
version of the Peabody Picture Vocabulary Test, was used at each assessment to measure receptive
vocabulary. The child selects from an array the picture that matches a word spoken by the tester. Scores
were converted to standard scores with a mean of 100 and S.D. of 15 (based on standardization data).

Understanding of grammatical contracts


The Test for Reception of Grammar (TROG) (Bishop, 1989) was administered at each assessment.
This is a multiple-choice test in which the child is required to select from an array the picture that
matches a phrase or sentence spoken by the tester. Vocabulary is kept simple, but grammatical complexity
increases as the test proceeds. The test is scored in terms of number of blocks correct out of 20, with
each block of four items testing comprehension of a particular type of grammatical contrast. Logarithmic
error scores were converted to standard scores with a mean of 100 and S.D. of 15 (based on standardization
data).

General comprehension
In the first three assessment sessions, the BAS verbal comprehension subtest (Elliott, Murray &
Pearson, 1978) was administered as a general measure of ability to carry out instructions. As norms
for this subtest do not extend beyond 4 years, 11 months, scores were transformed to scaled scores
for the first two test sessions only.
At the follow-up assessment at 8V2 years, the WISC-R verbal comprehension subtest was included.
Unlike the other two comprehension tests (BPVS and TROG), which simply involve matching the literal
meaning of a word or sentence to an item from an array, this subtest requires the child to use general
and social knowledge to work out appropriate answers to questions. Scores were transformed to standard
scores with a mean of 100 and S.D. of 15 (based on standardization data).

Expressive phonology
In the first three assessments, a set of photographs (Newcastle speech assessment) was given to the
child to name in order to elicit a wide range of consonants and consonant clusters. Responses were
transcribed by the tester, who was trained in phonetic transcription.
For the 8 5^-year old follow-up, responses to the Word-Finding Vocabulary Scale (see below) were
used to identify phonological errors in children's speech. Phonological errors when naming were
transcribed, and the child's total responses to the test were analysed to compute the percentage of
consonants correct (Shriberg & Kwiatkowski, 1982).
1030 D. V. M. Bishop and C. Adams

Expressive vocabulary
In the first three assessments, the BAS naming vocabulary subtest (Elliott et al., 1978) was administered,
scores being converted to standard scores.
For the 8 Vz -year old assessment we used Renfrew's Word-Finding Vocabulary Scale (Renfrew, 1980),
a confrontation picture-naming test suitable for children of this age. Our own control data were used
to convert scores into standard scores based on a mean of 100 and S.D. of 15.

Mean length of utterance (ML U)


For the first three assessments, MLU (in morphemes) was computed from all utterances given as
responses to the Action Picture Test (Renfrew, 1966) and the Bus Story Test (Renfrew, 1969). Control
data from children of the same age were used to convert scores to standard scores.
Lack of an adequately standardized procedure suitable for older children led us to formulate our
own test materials for the 8J4-year old assessment. A commercially available picture-book {Not Now,
Bernard, David McKee, 1980) was modified to give a story which could be told to the child with pictures
before asking him or her to go through the pictures again retelling the story. The child's story was
tape-recorded and transcribed by one of the authors, and used to compute MLU in morphemes. Control
data gathered in this study were used to transform scores into standard scores with a mean of 100 and
S.D. of 15.

Expression of semantic relations


T,he Bus Story information score was used as an index of the ability to express semantic relationships
at 4, 4J4 and bYz years. Control data were used to transform scores to standard scores.
The Not Now Bernard story, used at 8 /4 years, was also analysed for semantic content. Each phrase
in the story was coded according to whether it described a central idea of the story or an incidental
detail, and the child's account was credited 2 points for each central idea and 1 point for each incidental
detail mentioned. Alterations in wording were credited provided the basic meaning was preserved. Our
own control data were used to transform scores into standard scores with a mean of 100 and S.D. of 15.

Reading ability
The Neale Analysis of Reading Ability, form C (Neale, 1966), was administered at the follow-up
assessment at 8 ^ years. The child reads short stories of increasing difficulty until 12 or more errors
are made on a passage. If a child does not produce a response to a word within 4 seconds, the tester
provides it and an error is scored. A set of comprehension questions is provided after each passage
to assess the child's understanding of the story. Separate reading ages can be computed for accuracy
and comprehension. Data from an unselected population of 168 British 8J4-year olds (mean age 101
months, S.D. 6.9 months) (Bishop & Butterworth, 1979, 1980) were used as a basis for converting
scores to standard scores with a mean of 100 and S.D. of 15, after applying a logarithmic transform
to raw reading scores to reduce skew. Mean scores from the Bishop and Butterworth sample were close
to expected population values: the mean difference between reading accuracy and chronological age
(in months) was - 1.25. Reading comprehension data from this population have not previously been
published: the mean difference between reading comprehension age and chronological age in this sample
was -0.53 months.

Spelling
The Graded Word Spelling Test (Vernon, 1977) was administered at the follow-up assessment. The
child is required to spell words of increasing difficulty, each of which is dictated both in the context
of a sentence and in isolation. Testing is discontinued when 10 errors have been made. Published test
norms (for English children) were used to transform scores to standard scores with a mean of 100 and
S.D. of 15.

Test of non-word spelling and reading


Sixteen non-words were prepared, eight for spelling (mim, zab, pog, tep, plom, frool, samkin, hingo)
and eight for reading (bab, wob, zok, pim, stig, drune, binkol, shavim). Four items in each set were
Language impairment and reading retardation 1031

consonant-vowel-consonant monosyllables, two were monosyllables containing a consonant cluster,


and two were bisyllabic. Easy early items involved one-to-one correspondences between graphemes and
phonemes, but later items required knowledge of more complicated relationships between letters and
sounds (e.g. long vowels, or consonants spelt by a digraph). The first two items in each set were used
as practice items, which were not scored. For the spelling task, the child was presented with a card
showing a line-drawing of a space monster, whose name was written on the back of the card. The tester
spoke the name of the space monster and asked the child to repeat it and then to write it. For the practice
items, the child was helped to work out the correct answer if errors were made. The tester recorded
whether or not the child had repeated the non-word correctly. Feedback was given for all the spelling
items, with the child checking the back of the card after attempting to write the name. For the reading
task, the child was given cards showing a space monster and a non-word written in lower case, and
asked to read adoud the name of the space monster. Feedback was given for the two practice items.
Scoring of non-word reading and spelling was in terms of an error measure, phonemic distance from
target, with a point being added for each phoneme represented incorrectly, omitted or added (see Bishop,
1985, for further details). The total phonemic distance was added across all non-words, giving a maximum
error score of 25 for reading and 25 for spelling.

Procedure
The procedures for the first three assessments are described fully by Bishop and Edmundson (1987a).
Briefiy, each child was individually assessed in a single test session and given the whole battery of tests,
except where lack of cooperation or inattention developed to such a point that testing had to be
discontinued: in effect, this meant that some children were not given the BPVS and/or TROG at the
age of 4 or 4 ^ years.
For the follow-up assessment, each child in the experimental group was seen by one of the authors
in a quiet room at home, school or clinic for a single session lasting from 60 to 90 minutes, in which
all of the aforementioned assessments were administered.
Control 8 5^2 -year olds were seen individually at their school and were given the following tests: Not
Now Bernard story, WISC-R picture completion, Neale Analysis of Reading Ability, Vernon spelling
test, non-word reading and spelling and Word-Finding Vocabulary Scale.

Results
Mean scores on language and non-verbal tests are shown in Table 1 for control
children and for language-impaired 8-year olds divided according to status at 5 ^/4 years.
Scores on standardized tests for each subgroup in Table 1 were compared to mean
values from normative data using one-way analyses of variance. For non-standardized
tests, scores of experimental groups were compared with the control group mean using
one way analysis of variance. All F-ratios were significant at the 0.01 level. The general
delay group differed significantly from control values on all tests, verbal and non-
verbal. The children with poor outcome at 5 years scored significantly below control
values on all verbal assessments except for Vernon spelling, where the difference fell
just short of statistical significance. The mean score of this group on the non-verbal
block design test was significantly below the norm, but the picture completion score
was not. Children with good outcome at 5 years did not differ from normal controls
except on two comprehension tests, TROG and WISC-R verbal comprehension, where
mean scores were significantly below those of the standardization samples.

Reading ability relative to intellectual level


It can be seen from Table 1 that children with poor outcome at 5 years and those
1032 D. V. M. Bishop and C. Adams

Table 1. Mean (S.D.) scaled scores on language tests at 8 ^ years


Control Good outcome Poor outcome General
at 5 years at 5 years delay
30) (^ = 29) (A^=37) (A^= 16)

Age (months) 100.5 (2.94) 99.9 (0.86) 100.1 (1.35) 99.8 (1.24)
BPVS 96.5 (11.8) 83.6"(13.6) 76.1"(15.2)
TROG 92.0'' (13.0) 81.1" (9.4) 74.8" (7.8)
Verbad comprehension 91.7 " (16.3) 78.1" (12.2) 72.5" (12.2)
Word-finding 100.0*^ (14.9) 105.1 (11.2) 83.5" (16.3) 73.0" (19.0)
MLU 99.9^ (14.9) 98.5 (17.4) 85.0" (18.4) 78.6" (19.0)
Expressive semantics 97 9'* (16.6) 100.4 (17.2) 77.8" (17.9) 64.3" (12.7)
Reading accuracy 100.5 (15.1) 105.2 (12.3) 89.0" (14.6) 80.2" (20.2)
Reading comprehension 97.9 (18.7) 102.0 (12.6) 84.0" (14.9) 72.6" (18.6)
Spelling 94.1 (16.9) 96.4 (21.3) 83.0 (12.8) 77.5" (12.4)
Block design 105.0 (12.4) 91.5" (17.2) 87.5"(25.7)
Picture completion 98.7 (13.5) 101.7 (12.4) 94.1 (11.6) 85.9" (12.6)

"Significantly below control/normative value, p<0.05.


'^Used as basis for computing scaled scores.

with general delay had lower scores than the control group on non-verbal as well
as language tests. The first question to ask is whether their poor reading attainments
can be explained in terms of relatively low non-verbal ability.
The reading scores of the language-delayed group (all three subgroups combined)
were compared with control data from the Bishop and Butterworth (1980) sample
in an analysis of covariance, adjusting for non-verbal ability (i.e. summed scaled scores
on the two WISC-R performance subtests, block design and picture completion).
(Seventeen control children aged less than 96 months or more than 104 months were
excluded from this comparison.) Figure 1 shows the regression lines for predicting
140 1

120-
>
u
re
u 100-
u
re

80-
re

60-

40
40 60 80 100 120 140 160

block design •«• picture completion

Fig. 1. Scatterplot showing relationship between non-verbal ability and reading


accuracy at 8}-^ years for the language-delayed group. The regression hnes
for control and language-delayed groups are also shown: the control line is
the higher one. Both non-verbal ability and reading ability are shown as scaled
scores with a mean of 100 and S.D. of 15. (D) Good outcome at 5 years;
( B ) poor outcome at 5 years; (A) general delay.
Language impairment and reading retardation 1033

reading accuracy from non-verbal ability for these two groups, with the language-
delayed group coded according to 5}/2-year old status. Neither the slopes nor the
intercepts of these lines differed significantly. For reading accuracy, the adjusted scaled
scores were 99.45 for the controls and 97.05 for the language-delayed sample.
A different picture was obtained when reading comprehension scores were considered.
Regression lines for control and language-delayed children are shown in Fig. 2, together
with data points for individual children in the language-delayed sample. Slopes of
the two lines did not differ significantly, but the intercepts were significantly different
[F(l, 230) = 6.6; /)< 0.05]. The adjusted mean reading comprehension score for the
control group was 99.39, and for the language-delayed group, 92.54.

I4U -

o 120- D • ° ^ ^
"w •
a>
a> 100- D
• ••
jJ2P;jcf^°D
uo -X"^ • A •
80-
c 1 BA ••
• • • AAD •
CO
60- • • •
A AA A •

A
40- 1 1 1 1 1

40 60 80 100 120 140 160

block design + picture completion

Fig. 2. Scatterplot showing relationship between non-verbal ability and reading


comprehension at 8'/2 years for language-delayed group. The regression lines
for control and language-delayed groups are also shown: the control line is
the higher one. Both non-verbal ability and reading ability are shown as scaled
scores with a mean of 100 and S.D. of 15. (CH) Good outcome at 5 years;
( H ) poor outcome at 5 years; (A) general delay.

These results suggest that children in the language-delayed group obtain reading
comprehension scores that are disproportionately poor relative to their reading accuracy.
A further analysis of covariance confirmed this impression: reading comprehension
scores of language-delayed and control groups were compared after adjusting for
reading accuracy. The effect of group was highly significant \F (1, 230) = 26.31;

Thus when non-verbal ability is taken into account, the language-delayed group
does not differ from the control group on reading accuracy, but is impaired on reading
comprehension. The next question to ask is whether this deficit in comprehension
is specific to written language, or whether it is one aspect of more general verbal
impairment. It was noted above that the language-delayed children tended to do poorly
on WISC-R verbal comprehension. To see how far poor reading scores could be
explained in terms of this, a further analysis of covariance was carried out, this time
adjusting reading scores relative to verbal as well as non-verbal ability, by using the
summed scaled scores of all three WISC-R subtests (block design, picture completion
1034 D. V. M. Bishop and C. Adams

aind verbal comprehension) as covariate. There was no significant effect of group for
either reading accuracy [F(l, 230) < l] or reading comprehension [F(l, 230) = 1.57].
We may conclude that the poor reading comprehension scores obtained by the
language-delayed group can be explained in terms of their generally poor level of
language understanding.

Ability to read and spell non-words


Although the reading accuracy scores were very similar for those with good outcome
at 5 years and the control group, it could be that the two groups were using different
strategies to read. In particular, we were interested in the possibility that the children
with a history of SLI might have difficulty learning phonics and so might rely more
heavily on learning whole orthographic patterns, in which case they should be especially
poor at non-word reading. Non-word error data are shown in Table 2. A logarithmic
transformation was applied to non-word error data to stabilize variances, and a two-
way analysis of variance used to compare the control group with the three subgroups
of language-delayed children, treating non-word reading/spelling as a repeated
measure. There was a highly significant main effect of group {F = 10.96; d.f. = 3,
108; p < 0.01) and of read/speU (F = 22.31; d.f = 1, 108; p < 0.01), but the interaction
was not significant. The overall score of children with poor outcome at 5 years and
those with general delay was significantly below that of control children and the group
with good outcome at 5 years, who did not differ significantly from one another.

Table 2. Mean (S.D.) phoneme errors (out of 25) in reading and spelling non-words
Control Good outcome Poor outcome General
at 5 years at 5 years delay
(A^=30) (^ = 29) (A^=37) (N= 16)
Non-word
reading 7.7 (7.79) 5.3 (5.87) 12.3 (7.0) 16.1 (9.45)
Non-word
spelling 5.7 (6.38) 3.4 (3.84) 8.1 (6.14) 16.3 (10.27)

The scatterplot relating Neale reading accuracy to non-word reading errors is shown
in Fig. 3. If the experimental group had disproportionate difficulty with non-word
processing, their scores should cluster below the control regression line. It is clear
from inspection that this is not the case. An analysis of covariance confirmed that
neither the slopes nor the intercepts of the regression lines for predicting non-word
errors from reading accuracy differed between control and language-delayed groups
(all F-ratios< 1).

How many children are reading retarded?


The above analyses make it clear that the numbers of children classified as reading
retarded will depend crucially on the definition adopted.
Although not all authors accept this procedure (e.g. Siegel, 1988), it is common
for a distinction to be drawn between those children whose low reading attainments
Language impairment and reading retardation 1035

>u
a

(0

'•3

-2
o
o

40 60 80 100 120 140

Neale reading accuracy (scaled)

Fig. 3. Scatterplot showing relationship between Neale reading accuracy and


non-word reading accuracy. The plotted line is the regression of non-word
reading on reading accuracy for controls. ( x ) Control; (D) good outcome
at 5 years; ( I ) poor outcome at 5 years; (A) general delay.

are compatible with their limited intelligence, and those poor readers of normal
intelligence. Yule and Rutter (1976) have recommended defining specific reading
retardation (SRR) in terms of the regression equation relating reading ability to I.Q.
A child whose reading score is significantly below the level predicted from I.Q. is
regarded as specific2dly reading retarded, whereas one whose reading is poor but
consistent with I.Q. is termed a 'backward reader'. Using the control regression lines
relating non-verbal ability to reading accuracy and comprehension (see Figs 1 and 2),
children in the language-delayed sample were categorized as follows.
SRR: A—specific reading retarded on accuracy: this group consisted of children
whose obtained reading accuracy was more than 1.96 S.D.s below the value predicted
from their summed block design and picture completion scores.
SRR:C—specific reading retarded on comprehension: these were children whose
obtained reading comprehension was more than 1.96 S.D.s below the value predicted
from their summed block design and picture completion scores.
BR—backward readers: these were children whose reading scaled score was below
71 on either scale, but who did not fall into either SRR group. (In Bishop and
Butterworth's normal sample, 3% of children fell into this category.)
All those who did not fit the criteria for one of these categories were classed together
as normal readers.
Note that the two categories of SRR are not mutually exclusive. Five children met
the criteria for both SRR: A and SRR:C. Two of these had poor outcome at 5 years
and three were from the general delay group. Two children (both with poor outcome
at 5 years) were included in the category SRR:A but not SRR:C, and five children
(one good outcome at 5 years, one poor outcome at 5 years, and three general delay)
met criteria for SRR:C but not SRR:A. Overall, seven out of 82 children (8%) met
1036 D. V. M. Bishop and C. Adams

criteria for SRR:A, and 10 (12%) met criteria for SRR:C. In the general population
we would predict that 5% of children should fall in each category. The frequency
of SRR:A is not significantly different from expectation (chi-square with Yates'
correction = 1.48, d.f = 1), but the frequency of SRR:C is significantly above
expectation (chi-square = 7.49, d.f. = 1, p<O.Oi).
Five children (four with poor outcome at 5 years and one with general delay) were
backward readers: in all but one case they were backward on the comprehension score
only.
Thus 28 out of the 29 (96%) children with good outcome at 5 years were normal
readers, compared with 28 out of 37 (75%) cases with poor outcome at 5 years, and
nine out of 16 (56%) cases of general delay.
We have concentrated so far on defining reading retardation relative to non-verbal
ability. However, in this sample, most children had weaker verbal than non-verbal
skills. We might expect, then, that those children identified as specifically reading
retarded on this criterion would have difficulties with oral as well as written language.
Table 3 shows the language profiles of those seven children who fitted the profile
of SRR: A at 8 ^ years. It is evident that all had poor scores on tests of oral as well
a3 written language.

Table 3. Test profiles at 8 ^ years of seven children who fit criteria for specific reading retardation
A B C D E F G
'P.I.Q..' 110 73 104 96 71 76 103
Reading SS 75 55 55 55 55 55 66
Reading age 6:9 non-reader 5:11 5:10 5:11 5:10 6:7
BPVS SS 81 90 83 66 79 90 79
TROG SS 80 74 77 77 66 80 74
Verbal 85 60 70 65 65 75 60
comprehension SS
Word-fmding SS 79 94 79 70 55 88 67
MLU SS 82 57 92 55 68 73 69
Semantic SS 82 59 93 55 56 75 63

It follows that if we were to adopt a stricter definition, in which predicted reading


level was computed from both verbal and non-verbal ability, then f"ar fewer children
would be identified as cases of specific reading retardation. If verbal comprehension,
as well as block design and picture completion, is used to predict reading score, then
we do indeed find that the prevalence of specific reading retardation is no higher
than expectation (two cases of SRR:A only, three of SRR:C only and two fitting
both categories).

Discussion
This study confirmed that there is a link between early language impairment and
later literacy problems, but the nature of the link was rather different from that
anticipated.
Language impairment and reading retardation 1037

'Delayed language' as a precursor of specific reading impairment


The first unexpected finding was that children whose SLI had apparently resolved
by 5 ^ years continued to make good progress, with no significant impairments in
language or literacy measures. Furthermore, these children were just as proficient
as other 8}4-year olds in using phonics to read and spell non-words. Those children
who still had evident language difficulties at 5 ^ years were poor at reading at 8 V2
years. However, this was not an isolated impairment, but occurred in the context
of widespread verbal deficits, and rather poor non-verbal skills. Many children in
this subgroup were able to read accurately, but with poor comprehension of what
they read.
At first glance, these findings appear somewhat discrepant with the literature
reviewed in the Introduction. However, on closer inspection the inconsistencies are
not so great.

Studies of 'language delay' in children with 'developmental dyslexia'


In the field of developmental reading disorders, there has been much debate not
only over terminology (e.g. whether one should talk of 'developmental dyslexia' or
'specific reading retardation'), but also over defining criteria. We found that whether
or not one finds continuity between specific language impairment and 'developmental
dyslexia' depends crucially on how the latter is defined. Our study, like that of
Kinsbourne and Warrington (1963), suggests that only those reading-impaired children
who still have measurable oral language deficits are characterized by a history of
language delay. It is not the case that oral language problems disappear to be
superseded by reading problems. Rather, some children have oral language problems
that persist, but the focus of concern tends to shift to reading and spelling as they
grow older, perhaps because this is the most obvious sign of educational difficulty.
It follows that if one adopts a definition of developmental dyslexia that requires
that the child has normal verbal and non-verbal intelligence in the face of a severe reading
problem, then there will be little or no overlap between early SLI and later dyslexia.
If, however, one requires only that non-verbal skills of dyslexic children be in the normal
range, then some overlap between SLI and dyslexia will be found (see also Bishop
& Butterworth, 1980, for further discussion of these issues).
Another aspect of the definition of developmental dyslexia concerns which reading
measure is used. If, like Rutter et al. (1970), we diagnosed 'specific reading retardation'
in any child who did poorly relative to I.Q. on either reading accuracy or reading
comprehension, we found a significant excess of such cases in children with a history
of language delay (provided only non-verbal I.Q. was considered). If, however, we
used a definition concerned only with reading accuracy, the weak association with
language delay was not significant.

Other follow-up studies of language-impaired children


The impression that early language delay leads on to later literacy problems is
reinforced by the longitudinal studies of Fundudis et al. (1979), Richman et al. (1982)
and Silva et al. (1987), all of whom found poor literacy scores in children who had
been identified as having language delay in the preschool period.
1038 D. V. M. Bishop and C. Adams

However, these differences can easOy be accounted for in terms of sampling


procedures. In all these studies, the criterion for inclusion in the 'language-delayed'
sample was poor performance on some index of language in the preschool years,
without regard to non-verbal ability. Consequently, many children included in these
studies had general developmental delays, with language often appearing as the most
obvious manifestation. Their problems with reading and spelling at school-age were
not specific difficulties, but rather part of more global mild retardation that affected
scores on all intellectual tests, both verbal and non-verbal. Indeed, Richman et al.
went on to show that specific language delays were not associated with later reading
retardation, and that those children whose early language delay was associated with
reading and spelling problems tended to have generally low ability on a range of
measures. Similarly, Silva, McGee and Williams (1985) identified from their cohort
those children whose reading ability was significantly below the level predicted from
P.LQ. at 9 years, and showed that they did not differ from control children in scores
obtained with the Reynell Developmental Language Scales at the ages of 3 or 5 years,
nor on two subtests of the ITPA administered at 7 and 9 years. They did, however,
obtain significantly lower scores on an articulation screen at 7 and 9 years of age.
, In contrast to these other longitudinal studies, our sampling procedures were
designed to exclude children with general developmental delays (although we were
not entirely successful in this aim). Furthermore, because we recruited children from
speech therapists and paediatricians, we included some cases who would probably
not have met criteria for language delay in other studies, namely those children with
selective phonological disorders. As will be discussed below, these children had a
particularly good prognosis.

Part II: Relationship between Early Language Test Scores and Later Reading
Ability
The language-delayed children in this study included a wide range of different types
of problem when first seen. Those children with milder problems and good outcome
tended to be initially impaired only in the areas of expressive phonology and grammar,
with normal language comprehension and expressive semantics (Bishop & Edmundson,
1987a). The wider the range of language functions affected, the poorer the outcome
at 5 34 years. In general, there were strong intercorrelations between different language
tests scores. However, severity of initial phonological impairment, as assessed by the
percentage of consonants produced correctly, was not so strongly related to other
scores and did not predict outcome: many children who had good outcome at 5 years
had had severe phonological problems at the age of 4 years.
Given this diversity of language problems, we may ask which components of
language best predict reading and spelling competence. To date, there have been
few data on this question from disordered populations, although there have been several
longitudinal studies looking at predictors of reading progress within normal
populations.
In one of the best-known studies of this kind, Bradley and Bryant (1983, 1985)
showed that proficiency in phonological segmentation at 4-5 years, as evidenced by
ability to judge rhyme or alliteration of spoken words, was a significant predictor
Language impairment and reading retardation 1039

of later reading achievement, even after differences due to general intelligence and
vocabulary level had been allowed. This finding has been replicated by several
independent studies (see Wagner & Torgesen, 1987, for a review). Jorm (1979) has
argued that phonological segmentation skills are particularly important for learning
correspondences between letters and sounds, and that dyslexic children are deficient
in these skills. Jorm and Share (1983) further suggested that phonological recoding
of text may facilitate comprehension.
Other studies have concentrated on the role of metalinguistic skills, i.e. the ability
to reflect upon and manipulate the structural features of language. Tunmer, Herriman
and Nesdale (1988) found significant correlations between reading development and
performance on a range of metalinguistic tasks, including, among others, tests involving
phonological processing. Such findings suggest an alternative explanation of the results
obtained by other workers using tests of phonological awareness. It could be that
the crucial component of such tasks is not their phonological content so much as their
metalinguistic nature. If so, we would predict that phonological skills should be
unrelated to reading competence if they were assessed in a way that did not implicate
metalinguistic awareness.
Yet other workers have argued that basic competence in syntax and semantics are
important in reading acquisition. An extreme position had been taken by Smith (1973),
who argued that it is neither helpful nor natural for a child to decode written language
into spoken language by the use of letter-sound correspondences. Rather, the child
learns to understand written language by extracting meaning directly from written
words, without first accessing their phonology, and, when an unfamiliar word is
encountered, the reader engages in a form of top-down processing, making use of
syntactic and semantic redundancies to deduce the meaning. Bowey and Patel (1988)
obtained data consistent with this viewpoint. They found that a sentence repetition
test and a receptive vocabulary test accounted for significant amounts of variance
in reading ability in a normal sample, and that tests of phonological awareness and
other metalinguistic skills did not account for additional variance, when these factors
had been allowed. However, this was not a longitudinal study and it could be argued
that these language skills were enhanced by reading proficiency as much as the other
way around.
Consideration of these views of reading development in normal children suggests
several hypotheses about relationships between language disorders and later literacy
outcome.
The first prediction is that children with phonological disorders will be at particular
risk for reading and spelling difficulties, and their problems will be characterized by
lack of awareness of letter-sound correspondences, i.e. they should have particular
difficulty with non-words. There is indeed some evidence that supports this position.
Snowling and Stackhouse (1983) found that children diagnosed as suffering from
'developmental verbal dyspraxia' were quite unable to spell non-words, and their
spelling of real words indicated little understanding of letter-sound correspondences.
However, the children they studied had exceptionally severe and unusual phonological
problems, with inconsistent articulatory errors persisting to 8 years or over. Other
studies, however, have reported that children with isolated phonological problems
are less likely to have reading difficulties than those with other types of language
1040 D. V. M. Bishop and C. Adams

impairment (Debray-Ritzen, Mattlinger & Chapuis, 1976; Levi, Capozzi, Fabrizi


& Sechi, 1982).
Of course, the phonological disorders seen in our language-delayed children were
very different from the deficits in phonological processing studied by Bradley and
Bryant. If it is the metalinguistic component of phonological awareness tasks that
is crucial, then a phonological index such as ours, which simply reflects ability to
produce speech sounds accurately, might well prove to be unrelated to reading
development.
A further prediction, derived from the views of Smith (1973), is that competence
in syntax and semantics will be much more important in reading acquisition than
phonological skills. If the vocabulary and syntactic complexity of the text are beyond
the child's competence, then he or she will experience difficulty in using context to
deduce meaning of unfamiliar words. There is, indeed, some evidence that dyslexic
children are impaired on tasks of syntactic processing (Vogel, 1975), although this
has sometimes been explained as a secondary consequence of phonological limitations
(Mann, Shankweiler & Smith, 1984).
The positions adopted by such workers as Jorm and Smith are strongly polarized,
with the former regarding letter-sound decoding as all-important, and the latter
regarding it as irrelevEint to reading progress. Many researchers adopt a more balanced
position. Vellutino (1979), for instance, in a review of research on developmental
dyslexia, concluded: "It seems quite likely that given normal intelligence, intact visual
and auditory acuity, and adequate exposure to and investment in reading as a process,
success in learning to read depends, first, upon linguistic ability in general and, second,
upon the ability to make one's knowledge of language explicit. By extension, deficiencies
in any aspect of linguistic functioning will presumably result in difficulty in reading."
(p. 343: author's italics). In this view, all verbal tests should play some part in predicting
literacy skills.
Analyses designed to investigate which aspects of language were most important
in predicting reading difficulties were conducted using the same data as described
in Part I.

Results
One simple way of investigating the predictive significance of phonological disorders
is to consider the outcome of those cases whose early difficulties were restricted to
phonology. At 4 years of age, 12 of the children in this sample had isolated phonological
problems, with scaled scores of 80 or above on all other language measures. According
to the criteria of Shriberg and Kwiatkowski (1982), the 4-year phonological problems
were mild in one case, mild-moderate in seven, and moderate-severe in four. This
subgroup of children did particularly well in reading and spelling at 8 ^ years, with
a mean reading accuracy scaled score of 111 (S.D. = 11.02) and spelling scaled score
of 105.7 (S.D. = 18.3). A similar analysis was conducted on data obtained at the
5 ^ -year old assessment. Most of the children who had had pure phonological problems
at 4 years of age had normal speech when seen at 5 34 years, but other children who
previously had had more widespread problems had improved to the point where only
phonology remained impaired. Table 4 shows data for all those children who scored
Language impairment and reading retardation 1041
Table 4. Test profiles of six children with isolated phonological impairment at 5J4 years
Child identifier H I J K L M
% Consonants correct at 5'/2 yr 89 85 89 80 11 56
Phonenie problems at 8 54 yr? no no no no no mild
'P.I.Q..' at 8J4 years (2 subtests)^ 76 103 121 100 114 71
Reading accuracy^ 101 115 100 83 114 100
Reading comprehension^ 88 111 96 75 111 93
Spelling^ 100 120 125 83 115 93
Non-v^ord reading error 3 1 9 16 1 6
Non-word spelling error 3 0 3 3 4 3
^Scaled scores.
less than 90 % on the measure of percentage consonants correct at 5 J4 years but who
scored above the 10th centile on all other measures at this age. It is evident from
inspection that, for all but one of them, literacy skills are as good as those of 8-year
old controls. The number of children whose reading is below the level predicted from
non-verbal ability is no greater than the number above prediction. Child K is the
only one for whom reading and spelling skills are rather poor relative to non-verbal
ability.
These analyses indicate, then, that isolated phonological problems in pre-literate
children are not predictive of reading difficulties.
Are other language measures better predictors of literacy outcome? Correlations
between language tests at 4, 4 ^ and 5 ^ years and reading and spelling scores at
8^/4 years are shown in the Appendix. All correlations are based on standard scores
except for percentage consonants correct, where normal children's performance
approaches ceiling. It is noteworthy that phonological status at 4 and 4 V^ years shows
weaker correlations with later reading and spelling ability than other language tests,
although the correlation becomes stronger at 5^2 years.
The measures of receptive and expressive semantic and syntactic skills are strongly
intercorrelated, so it becomes interesting to ask whether these make independent
contributions to outcome, or whether they reflect the effect of a common 'language
competence' factor. To investigate this question, multiple regression was used. In
all analyses, non-verbal ability (summed scaled score on picture completion and block
design) was forced as the first variable to be entered, with language variables from
the earlier assessments then entered using stepwise selection. Results for analyses
using data from 4 V2 -year old and 5 Vi -year old assessments as predictors are shown
in Table 5. The strongest and most consistent 4 V2 -year old predictor of subsequent
reading ability is MLU. Once scores have been adjusted for this variable, other
language variables do not account for additional variance in reading accuracy or non-
word reading, although the TROG score accounts for additional variance in the two
spelling tests. For reading comprehension, the expressive semantics score accounts
for a substantial amount of variance, but other language variables do not enter the
equation. The pattern is rather similar when predicting outcome from 5^-year old
language measures, with MLU and TROG again accounting for most of the variance
in outcome. However, the expressive phonological measure, percentage consonants
correct, does account for a small but significant proportion of the variance in reading
accuracy, after allowing for the effect of MLU.
1042 D. V. M. Bishop and C. Adams

Table 5. Stepwise regression analyses predicting reading and spelling at 8 years, with literacy
measures adjusted for non-verbal ability at 8 years
(a) Prediction from 454 year assessments (A'^= 75)
Dependent variable Predictors Total % variance Varizmce increase
Reading accuracy 854 year 'P.LQ.'^'' 39.1
4/2 year MLU 48.0 8.9

Reading comprehension 8 54 year 'P.LQ.'* 41.6


454 year semantics 56.6 15.0

Spelling 854 year 'P.LQ.'^ 31.7


4/2 year MLU 44.9 13.2
454 year TROG 48.6 3.7

Non-word reading 854 year 'P.LQ.'^ 37.2


4/2 year MLU 43.3 9.7

Non-word spelling 8K2 year 'P.LQ.'^ 33.6


454 year MLU 43.3 9.7
454 year TROG 47.4 4.1

(b) Prediction from 5J4 year assessments (A'^=81)


Reading accuracy 8 54 year 'P.LQ.'* 29.8
51/2 year MLU 42.7 12.9
5 5/2 year TROG 46.3 3.6
5 54 year % consonants correct 50.3 4.0

Reading comprehension 8 54 year 'P.I.Q.'* 35.0


51/2 year BPVS 56.8 21.8
5 54 year % consonants correct 60.5 3.6
5 54 year TROG 62.5 2.1

Spelling 8 54 year 'P.I.Q.'^ 26.1


5 54 year MLU 38.5 12.4
5 54 year naming 41.7 3.2

Non-word reading 8 54 year 'P.I.Q.'* 32.3


5 K2 year MLU 46.5 14.2
5 54 year TROG 49.7 3.2

Non-word spelling 85/2 year 'P.LQ.'^ 27.6


5 54 year MLU 43.5 15.9
5 K2 year TROG 47.0 3.5
^Variable forced into equation.
''Summed scailed scores of picture completion and block design.

How do these results compare with those obtained by Bradley and Bryant? It is
important to note that in their multiple regression analyses, they first adjusted reading
scores not only for non-verbal ability, but also for verbal ability measured at follow-
up, by entering into the regression equation receptive vocabulary (English Picture
Vocabulary Test) and full-scale LQ. (rather than just performance LQ.). It was
Language impaimtient and reading retardation 1043

possible for us to follow a similar procedure, by forcing both 8 ^/^ -year old BPVS scaled
score and the summed scaled scores on the three WISC-R subtests into the equation,
before entering the language variables from the earlier assessments, again using
stepwise selection procedures. Results from this procedure are summarized in Table 6.
The measure of expressive phonology, percentage consonants correct, emerged as
the strongest predictor of adjusted reading accuracy at 4 ^ years (accounting for 5.4%
of variance in outcome). An analogous analysis conducted with 5 V2 -year old language
variables gave a similar result, with percentage consonants correct accounting for
9.9% of variance. These results were specific to reading accuracy. Early phonological
status did not emerge as a significant predictor when multiple regression was used
to predict adjusted reading comprehension scores. Percentage consonants correct at
5 54 years was also the best predictor of LQ.- and vocabulary-adjusted spelling score
at 8 ^ years (9.9% of variance). However, it was not a significant predictor of non-
word reading and spelling at any age.
The results from the analyses of expressive phonology at 4 5/2 and 5 5/2 years as a
predictor of reading accuracy scores thus agree with those reported by Bradley and
Bryant using a test concerned with a different aspect of phonological processing.
Figure 4 shows the scatterplot relating phonological status at 5 V2 years to reading
accuracy after adjusting for verbal and non-verbal status at 8 ^2 years (expressed as
a standardized residual). One can see that while many children with good phonology
scores do have poor reading scores, children scoring below 70% on the measure of
percentage consonants correct are seldom good readers. Note that most of these latter
children were not included in Table 4 because they obtained low scores on measures
other than phonology.

Discussion
A significant proportion of variance in reading attainment can be explained by
early language scores. All the preschool tests of expressive and receptive syntax and
semantics showed healthy correlations with later reading and spelling ability. These
measures were strongly intercorrelated, and MLU emerged as the best single predictor.
Expressive phonology did not begin to show substantial correlations with literacy
test scores until 5 34 years of age, when the child was first learning to read. Phonological
competence in 4-5-year olds was relatively independent of other language test scores.
However, it did account for a significant proportion of variance in reading accuracy
once effects of general language competence had been allowed, either by first adjusting
scores for early MLU, or by using a procedure similar to that adopted by Bradley
and Bryant, where literacy measures are first adjusted for the effects of verbal and
non-verbal ability at outcome.
The finding that non-phonological aspects of language are important predictors
of reading development is not, in fact, inconsistent with the results obtained by Bradley
and Bryant. They obtained substantial correlations between scores on the EPVT (a
test of receptive vocabulary) and reading ability. They also drew attention to the fact
that while good phonological competence in preschool was a reasonable predictor of
which children would become unusually good readers, the converse did not apply,
and phonological categorization tasks at this age were poor at identifying children
1044 D. V. M. Bishop and C. Adams

Table 6. Stepwise regression analyses predicting reading and spelling at 8 ^ years, with literacy
measures first adjusted for verbal and non-verbal ability at 8 54 years
(a) Prediction from 454 year assessments (N= 75)
Reading accuracy 854 year BPVS" 29.7
8»/2 year 'LQ.'^'' 44.0 14.3
454 year % consonants correct 49.4 5.4
4^2 year TROG 52.1 2.7

Reading comprehension 854 year 43.0


854 year 'LQ. 57.1 14.1
41/2 year MLU 61.4 4.3

Spelling 854 year 22.8


8 54 year 'LQ.'=* 37.1 14.3
454 year MLU 44.5 7.4
454 year TROG 47.6 3.1

Non-word reading 81/2 year 30.5


854 year l.Q. 43.9 13.4
41/2 year MLU 49.2 5.3

Non-word spelling 8 54 year 24.6


8K2 year LQ. 37.6 13.2
454 year TROG 42.8 5.2
4^2 year MLU 46.0 3.2

(b) Prediction from 5 54 year assessments (N=81)


Reading accuracy 8 5/2 year 25.3
854 year 'LQ.'^ 37.7 14.4
5 54 year % consonants correct 47.6 9.9
5 54 year TROG 51.2 3.6

Reading comprehension 8 ^2 year 41.2


8 54 year 'I.Q.' 53.7 12.5
5 54 year MLU 60.1 6.4
51/2 year ' L Q . ' 62.4 2.3

Spelling 854 year 20.0


8 54 year 'LQ.'^ 33.3 13.3
5 54 year % consonants correct 43.2 9.9

Non-word reading 8 54 year BPVS^ 27.6


8 54 year 'LQ.'^ 40.9 13.3
51/2 year MLU 48.4 7.5

Non-word spelling 8/2 year BPVS^ 22.4


8K2 year 'LQ.'^ 33.7 11.3
51/2 year MLU 43.8 10.1
^Variable forced into equation.
'^Summed scale scores of picture completion, block design and verbal comprehension.
Language impairment and reading retardation 1045

3'

2
re
3

I 1

1 0
El
re
-1
c Q •
3
(A
-2

-3
20 30 40 50 60 70 80 90 100
% consonants correct

Fig. 4. Percentage consonants correct at 5 54 years versus reading accuracy


adjusted for BPVS and LQ. (verbal comprehension, block design and picture
completion) at 8 54 years (expressed as a standardized residual).

who would go on to have reading problems. They concluded: "Results like these
support our belief that learning to read and spell is a complicated business which
involves a number of different skills, of which we are tapping only one" (Bradley
& Bryant, 1985, p. 105).
We may conclude that phonological proficiency is not the main determinant of
reading acquisition: syntactic and semantic ability are responsible for the major part
of variation in reading ability. Nevertheless, it is noteworthy that, despite the differences
in sampling and assessment procedures, we confirmed Bradley and Bryant's finding
that phonological processing accounted for measurable variance in reading outcome
after allowing for the powerful effects of other language measures. This was so despite
the fact that our measure of phonological competence had no metalinguistic component.
Although phonological competence at 5J4 years accounted for a significant
proportion of^ variance in later reading accuracy, the effect was small and, in this
sample of language-impaired children, provided that a child's difficulties were restricted
to phonology and did not occur in the context of other verbal difficulties, normal
progress in reading and spelling could be anticipated. Having said that, it should
be noted that isolated phonological problems were rare in this sample and seldom
persisted beyond 5 years of age. Those few children who produced less than 70%
of consonants correctly at 5 V2 years of age mostly had wide-ranging deficits involving
syntax and semantics. The finding that such children tended to be poor at reading
relative to their intelligence and other language skills is open to two possible
interpretations. The first explanation, which may be termed the Critical Age
hypothesis, is that an expressive phonological problem which is still present when
the child is first learning to read will be detrimental to reading development, provided
it is severe enough. Milder problems and those that resolve by the time the child
starts formal schooling will not have an effect. An alternative explanation, the Synergy
hypothesis, maintains that expressive phonological impairments only assume
importance for reading acquisition when they occur in the context of other language
1046 D. V. M. Bishop and C. Adams

deficits. In this view, the combination of phonological disorder and other language
difficulties has a worse impact on learning to read than would be predicted from either
effect considered in isolation. To choose between these hypotheses we need
longitudinal studies of children who have severe and specific phonological disorders
extending beyond 5 years, so that the effects of age and presence of other language
difficulties can be unconfounded.

Clinical Implications
A high proportion of children in our sample did have reading problems, but not
of the kind anticipated. Relatively few children had reading accuracy scores below
the level predicted from their non-verbal ability, and those that did all had widespread
impairments of spoken as well as written language, with poor scores on comprehension
as well as expressive tests. We did not find any children who had significant reading
or spelling difficulties in the context of otherwise normal verbal functions.
We had expected that language impairment, and especially phonological problems,
might lead to difficulty in learning letter-sound correspondences, so that children
with normal verbal comprehension would be frustrated in their attempts to decode
written material. However, the pattern of deficit we observed suggested just the
opposite. The commonest type of reading problem was where the child had normal
reading accuracy but poor comprehension for what had been read. These children
typically also had poor understanding for spoken material. The term 'hyperlexic'
is sometimes used to describe children whose reading accuracy is well in advance
of reading comprehension, but we feel this term should be reserved for those unusual
children who learn to read precociously and whose accuracy in reading is well in
advance of their chronological age. This was not the case for the children in this sample,
none of whom learned to read before starting school.
If, as Bishop and Butterworth (1980) recommended, we define specific reading
retardation as a difficulty in reading that is discrepant with both non-verbal and verbal
abilities, then not a single child in this sample fitted this picture. It would, of course,
be quite wrong to conclude that such children do not exist: the literature contains
many well-documented cases. What we can question, however, is the view that a
specific language impairment in the preschool period manifests itself as a specific
reading disorder later on. Contrary to what we expected, children who grow out of
their early language difficulties (our group with good outcome at 5 years) were not
at risk for literacy problems. Children who still have evident language impairment
at the age of 5 }^ years are likely to have reading and spelling difficulties, but these
will not be isolated problems, but will occur in the context of persisting deficits in
comprehension and expression of spoken language. In many children with persisting
language impairment, comprehension of written material is a much greater problem
than learning to decode from printed text to speech. It seems probable that their
problems are especially likely to be overlooked because of their apparent facility in
reading accuracy and spelling.
Our study indicates that expressive phonological competence does play a part in
learning to read, but it suggests that the importance of phonological processing may
have been overstated. Phonological factors are of particular theoretical interest because
Language impairment and reading retardation 1047

they seem able to explain variation in reading acquisition that is not accounted for
in terms of other, more general, verbal abilities. However, it should be emphasized
that other language skills exert the major influence on reading progress.

Acknowledgements—We would like to thank all the parents and children involved in this study for their
continuing support and cooperation. In addition, thanks are due to the many speech therapists and
paediatricians in the north east and north west of England who first referred children to this study,
and who facilitated this research in all its stages. Many school staff helped to make this study possible,
with especial thanks being due to those at the schools where we saw control children: Reedley Gounty
Primary School, Lancashire and St Edward's Primary School, Manchester. Janice Kay kindly assisted
with the creative design of space monsters. This study was supported by a project grant from the Medical
Research Council.

References
Bishop, D. V. M. (1985). Spelling ability in congenital dysarthria: evidence against articulatory coding
in translating between phonemes and graphemes. Cognitive Neuropsychology, 2, 229-251.
Bishop, D. V. M. (1989). Test for reception of grammar (2nd edn). Published by the author at Department
of Psychology, University of Manchester, Manchester.
Bishop, D. V. M. & Butterworth, G. E. (1979). A longitudinal study using the WPPSI and WISC-R
with an English sample. British fournal of Educational Psychology, 49, 156-168.
Bishop, D. V. M. & Butterworth, G. E. (1980). Verbal-performance discrepancies: relationship to birth
risk and specific reading retardation. Cortex, 16, 375-389.
Bishop, D. V. M. & Edmundson, A. (1987a). Language-impaired 4-year-olds: distinguishing transient
from persistent impairment. Journal of Speech and Hearing Disorders, 52, 156-173.
Bishop, D. V. M. & Edmundson, A. (1987b). Specific language impairment as a maturational lag:
evidence from longitudinal data on language and motor development. Developmental Medicine and
Child Neurology, 29, 442-459.
Bowey, J. A. & Patel, R. K. (1988). Metalinguistic ability and early reading achievement. Applied
Psycholinguistics, 9, 367-383.
Bradley, L. & Bryant, P. E. (1983). Categorizing sounds and learning to read—a causal connection.
Nature, 301, 419-421.
Bradley, L. & Bryant, P. E. (1985). Rhyme and reason in reading and spelling. IARLD Monographs, No. 1.
Ann Arbor: University of Michigan Press.
Debray-Ritzen, P., Mattlinger, M. & Chapuis, C. (1976). Etude statistique du devinir lexique des
enfants presentant un retard de langage. Revue Neurologique, 132, 651-659.
Dunn, L. M., Dunn, L. M., Whetton, C. &Pintilie, D. (1982). British Picture Vocabulary Scale. Windsor:
NFER-Nelson.
Elliott, C. D., Murray, D. J. & Pearson, L. (1978). British Ability Scales. Windsor: NFER-Nelson.
Fundudis, T., Kolvin, I. & Garside, R. (1979). Speech retarded and deaf children: their psychological development.
London: Academic Press.
Harris, D. B. (1983). Goodenough-Harris Drawing Test. New York: Harcourt Brace Jovanovitch.
Jorm, A. F. (1979). The cognitive and neurological basis of developmental dyslexia: a theoretical
framework and review. Cognition, 7, 19-33.
Jorm, A. F. & Share, D. L. (1983). Phonological recoding and reading acquisition. Applied Psycholinguistics,
4, 103-147.
Kinsbourne, M. & Warrington, E. (1963). Developmental factors in reading and writing backwardness.
British Journal of Psychology, 54, 145-156.
Leiter, R. G. (1948). Leiter International Performance Scale. Windsor: NFER-Nelson.
Levi, G., Capozzi, F., Fabrizi, A. & Sechi, E. (1982). Language disorders and prognosis for reading
disabilities in developmental age. Perceptual and Motor Skills, 54, 1119-1122.
1048 D. V. M. Bishop and C. Adams

Mann, V. A., Shankweiler, D. & Smith, S. T. (1984). The association between comprehension of spoken
sentences and early reading ability: the role of phonetic represcntaition. Journal of Child Language
11, 627-643.
McKee, D. (1980). Not now, Bernard. London: Arrow Books.
Naidoo, S. (1972). Specific dyslexia. London: Pitman.
Neale, M. D. (1966). Neale analysis of reading ability. London: Macmillan.
Renfrew, C. E. (1966). Action Picture Test. Published by the author at North Place, Oxford, U.K.
Renfrew, C. E. (1969). The bus story: a test of continuous speech. Published by the author at North Place
Oxford, U.K.
Renfrew, C. E. (1980). Word-Finding Vocabulary Scale. Published by the author at North Place, Oxford
U.K.
Richman, N., Stevenson, J. & Graham, P. (1982). Preschool to school: a behavioural study. London: Acade
Press.
Rutter, M., Tizard, J. & Whitmore, K. (Eds). (1970). Education, health and behaviour. London: Longman
Shriberg, L. & Kwiatkowski, J. (1982). Phonological disorders. III. A procedure for assessing severity
of involvement. Journal of Speech and Hearing Disorders, 47, 98-112.
Siegel, L. S. (1988). Evidence that I Q scores are irrelevant to the defmition of reading disability. Canadian
Journal of Psychology, 42, 201-215.
Silva, P. A., McGee, R. & Williams, S. M. (1985). Some characteristics of 9-year old boys with general
reading backwardness or specific reading retardation. yo«r«a/ of Child Psychology and Psychiatry, 26,
407-421.
Silva, P. A., Williams, S. M. & McGee, R. (1987). A longitudinal study of children with developmental
language delay at age three: later intelligence, reading and behaviour problems. Developmental Medicine
and Child Neurology, 29, 630-640.
Smith, F. (19737. Psycholinguistics and reading. New York: Holt, Rinehart & Winston.
Snowling, M. (1987). Dyslexia: a cognitive developmental perspective. Oxford: Blackwell.
Snowling, M. & Stackhouse, J. (1983). Spelling performance of children with developmental verbal
dyspraxia. Developmental Medicine and Child Neurology, 25, 430-437.
Tunmer, W. E., Herriman, M. L. & Nesdale, A. R. (1988). Metalinguistic abilities and beginning
reading. Reading Research Quarterly, 23, 134-158.
Vellutino, F. R. (1979). Dyslexia: theory and research. Cambridge, MA: MIT Press.
Vernon, P. E. (1977). Graded Word Spelling Test. London: Hodder & Stoughton.
Vogel, S. A. (1975). Syntactic abilities in normal and dyslexic children. Baltimore: University Park Press
Wagner, R. K. & Torgesen, J. K. (1987). The nature of phonological processing and its causal role
in the acquisition of reading skills. Psychological Bulletin, 101, 192-212.
Wechsler, D. (1974). Manual of the Wechsler Intelligence Scale for Children—Revised. New York: Psycholo
Corporation.
Yule, W. & Rutter, M. (1976). Epidemiology and social implications of specific reading retardation.
In R. M. Knights & D. J. Bakker (Eds), The neuropsychology of learning disorders (pp. 25-39). Baltim
University Park Press.
Language impairment and reading retardation 1049
Appendix
bo
C
csr
CO
u 1
' V
-< o
Z
z
o CO
u:> oi
d. CO r>«
^ d d oi

u =5 ^
o in (23 O
u
09 C O
o -s
i
u
u
O O^ LO t^
-a cy> lO t o o
2
'o
CO CO 00 CO 'S 13 ««

u c
C o
O CO
c
00 O o o
d> d> d d
a CO

CO —1 —I O
in in - ^ CM en -^
O o •'t' in o O^ "^ CO
u ^ o o o d d
CM •*' U

E
u
be en c
d d
c
^
^
11^
« CO U

bo O CTi i n i n CTi t ^ • ^ m -^ (U
Cl r^ i n Th CTi O CM i n > >
en •^ C M C O CT! -^ C<^
§ 2
h ^ddddddd in in
o ,
d d
O J3 X
O
CL,
a •^dddddddd t-H w «
O IT
u
be
"^ J J pa U CO

u C
oco^ooocn •^ bD b/j

G
o
I Tti CO <x> o
ddddddddd
•* in • * • * '^^ O S

a,
' 5 CM en rt S CO CO
c
u
u CO® °
O g o o
•^dddddddddd J5 O O
bb^ o w
o w u
rj S.H t»
t- O O
^ ^ y

ddddddddddd a. .- ..
c i^ en E'E &•
u r^ en o
u II II

bo
ES 13 J3 -C
•(j -w •• r,
en
'4.' oj
W
n
O

3 E''
^ GO
1050 D. V. M. Bishop and C. Adams

(b) Correlations between 4 ^ -year old language measures and 8 ^ -year old literacy scores
MLU Sem Naming BPVS TROG V. comp % Cons
MLU 1
Sem 0.737 1
Naming 0.509 0.678 1
BPVS 0.402 0.573 0.615 1
TROG 0.470 0.620 0.573 0.589 1
V. comp 0.398 0.555 0.541 0.638 0.665 1
% Cons 0.475 0.209 0.119 0.116 0.022 -0.063 1
Read a 0.526 0.542 0.562 0.418 0.510 0.412 0 .256
Read c 0.592 0.656 0.622 0.537 0.538 0.549 0 .138
Spell 0.567 0.599 0.573 0.412 0.538 0.405 0 .230
Nw read 0.559 0.560 0.571 0.483 0.509 0.401 0 .206
Nw spell 0.521 0.517 0.541 0.350 0.533 0.354 0 .153

N= 74. A correlation of 0.229 is significant at the 0.05 level, and 0.298 at the 0.01 level.
Key: as in part (a).

(c) Correlations between 5 H -year old language measures and 8 i4 -year old literacy scores
MLU Sem Naming BPVS TROG % Cons

MLU 1
Sem 0.733 1
Naming 0.641 0. 611 1
BPVS 0.642 0. 760 0. 619 1
TROG 0.502 0. 702 0. 381 0.550 1
% Cons 0.522 0. 285 0. 417 0.190 0 .142 1
Read a 0.551 0. 496 0. 561 0.516 0 .509 0.433
Read c 0.623 0. 632 0. 628 0.673 0 .554 0.368
Spell 0.539 0. 520 0. 581 0.497 0 .469 0.439
Nw read 0.585 0. 534 0. 526 0.550 0 .519 0.391
Nw spell 0.582 0.467 0.493 0.459 0.514 0.409
A^=80. A correlation of 0.217 is significant at the 0.05 level, and 0.283 at the 0.01 level.
Key: as in part (a).

You might also like