1. hobbies and interests
2. reading

Cognition,
51 (1994)
OOlO-0277/94/$07.00
91
91-103
@ 1994 - Elsevier
Science
B.V. All
rights
reserved.
The influence of orthographic consistency
on reading development: word recognition
in English and German children
Heinz Wimmer*‘“, Usha Goswamib
“Institut fiir Psychologie,
Hellbrunnerstrasse
34, A 5020 Salzburg, Austria
bDepartment of Experimental Psychology,
University of Cambridge, Cambridge
Received
August
27, 1992, final version
accepted
August
CB2 3EB,
UK
30, 1993
Abstract
Groups of 7, 8, and 9-year-old children who were learning to read in English and
German were given three different continuous reading tasks: a numeral reading
task, a number word reading task, and a nonsense word reading task. The
nonsense words could be read by analogy to the number words. Whereas reading
time and error rates in numeral and number word reading were very similar across
the two orthographies, the German children showed a big advantage in reading the
nonsense words. This pattern of results is interpreted as evidence for the initial
adoption of different strategies for word recognition in the two orthographies.
German children appear to rely on assembling pronunciations via graphemephoneme conversion, and English children appear to rely more on some kind of
direct recognition strategy. A model of reading development that takes account of
orthographic
consistency is proposed.
Introduction
Models of reading development have made the implicit assumption that stages
of reading development are more or less uniform across different alphabetic
We would like to thank the teachers and children of St. Laurence’s
Primary School, Cambridge,
England,
and the Volkschule Bischofshofen,
Austria. We also thank Michaela Juritsch, Sarah Pressley,
and Barbara
Klampfer
for help in collecting
and analysing
the data. This research
was partially
supported
by grants from the University
of Salzburg, Austria,
and St. John’s College, Cambridge.
*Corresponding
author.
SSDZ
0010-0277(93)00581-W
92
H. Wimmer.
orthographies.
U. Goswami
A plausible
I Cognition 51 (lY94) 91-103
alternative,
however,
is that the transparency
orthography
will affect the development
of reading. This
direct and indirect. If an orthography
is highly transparent,
of the
effect could be both
with very consistent
mappings
from spelling to sound,
then grapheme-phoneme
should
be easier to detect and use: a direct effect. In
correspondences
a less transparent
orthography,
the underlying
rules will be less consistent,
and may be more
complex in terms of being context-sensitive
and operating at different phonological levels.
With
such orthographies,
it may be more
adaptive
initially
to learn
spelling patterns for individual
words, and then to use various strategies such as
analogy to try and read new words.
One can also speculate that the consistency
of the orthography
could have an
indirect effect on reading development
via teaching methods. For a highly regular
orthography
grapheme-phoneme
correspondences
are obviously
easy to teach,
and the regularity
of the orthography
guarantees
that grapheme-phoneme
conversion
and blending quite reliably results in correct word recognition.
So a
“phonics”
approach
is a quite convenient
method
to introduce
children
to
reading. These advantages
of a phonics approach are obviously reduced for a less
regular orthography.
If orthographic
transparency
affects reading development
in the indirect and
direct ways that we have described,
then the assumption
that stages of reading
development
are uniform
across different
alphabetic
orthographies
can be
questioned.
For example, two recent influential
models of reading development
have assumed that beginning readers will adopt a strategy of logographic
reading,
in which access to the lexicon is direct and visually based (Frith, 1985; Marsh,
Friedman,
Welch, & D&berg,
1981). In both models, this foundational
logographic stage is followed by an alphabetic stage of decoding, in which graphemephoneme
correspondences
are used to assemble pronunciations.
Frith has further
proposed
that the final stage in development
is a return to direct access, with
orthographic
word recognition
replacing
or supplementing
alphabetic
reading.
Orthographic
reading is characterized
by the use of spelling units that represent
morphemes.
However,
while it is possible to argue that this sequence
of stages may be
characteristic
of learning to read English, which has a relatively
opaque orthography, this sequence
may not apply to learning
to read a highly transparent
orthography
such as German.
In German,
the mapping
from graphemes
to
phonemes
is largely consistent,
especially
for vowel graphemes.
An adaptive
strategy for young German
readers would thus be to use grapheme-phoneme
translation
from the beginning
of reading, omitting an initial stage of logographic
access. In English, however, mappings at the grapheme-phoneme
level are quite
inconsistent,
with vowel graphemes
being particularly
ambiguous.
For young
readers of English, it would be initially more adaptive to use varieties of direct
access strategies,
memorizing
spelling patterns in order to build up a lexicon of
H. Wimmer, U. Goswami
orthographic
recognition
units from which analogies
I Cognition 51 (1994) 91-103
93
to new words could be made,
and context-sensitive
grapheme-phoneme
mappings could be derived.
There is already indirect evidence that German children may not pass through
an initial stage of logographic
reading. Wimmer and Hummer
(1990) found that
the majority of errors made by children learning to read German were nonsense
words, in marked contrast to children learning to read English, whose errors were
largely
(the wrong)
real words (e.g.,
Seymour
& Elder,
1986; Stuart
& Coltheart,
1988). The preponderance
of nonsense
word errors in German implies that the
children
were reading
words indirectly,
by assembling
pronunciations
from
The English
children
were apparently
grapheme-phoneme
correspondences.
attempting
to use direct access strategies to read words, resulting in real word
rather than nonsense
word errors.
The present study attempts to provide a more direct test of our assumptions
about the effects of orthographic
consistency
on reading development.
In order to
see whether English and German children in the early phases of learning to read
differ in their reliance on direct access and assembled pronunciation,
we chose to
compare reading of simple number words with reading of nonwords derived from
the number
words. The number
words chosen are very similar in the two
orthographies
(e.g., three, drei; seven, sieben). They are also familiar to young
children,
and may well be over-learned,
making it likely that direct recognition
units for these words are established
quite early. The children were also asked to
read aloud the corresponding
numerals,
for which pronunciation
must be directly
accessed. Finally, the children were asked to read nonsense
words, which were
derived from the number words by exchanging
the initial consonantal
graphemes,
so that the spelling patterns corresponding
to the sub-syllabic
units of onsets and
rimes remained
undisturbed.
For example, se12 was created from seven and ten,
and zwehn was created
from zwei and zehn. This method
necessitated
the
omission of the numbers “one”, “eight” and “eleven”;
otherwise all the numbers
from two to twelve were used. Such nonsense words must be read indirectly,
by
assembling
pronunciations.
A full list of the nonsense words used is given in the
Appendix.
If our assumption
about the initial difference in reading strategy between the
two orthographies
is correct, then English children, who may be more dependent
on direct access strategies,
should have difficulty in reading the nonsense
words.
The number words should present few problems.
The German children,
on the
other hand, should have little difficulty with the nonsense
word reading task, as
they are used to assembling pronunciations.
Number words should also be read by
assembling
pronunciations,
although German children might already have direct
recognition
units for these typically over-learned
words. It follows, therefore,
that
there should be an advantage
for the German children compared
to the English
children in reading the nonsense words. No such difference would be expected for
the number
words. It is even possible that the English children might show an
94
H. Wimmer,
early advantage
pronunciations
children
U. Goswami
I Cognition
51 (1994) 91-103
in number word reading if the German children
for these words. For the numerals,
the English
were expected
to perform
need to assemble
and the German
comparably.
Method
Subjects
Seventy-two
English children (28 boys, 44 girls) and 81 Austrian
children (41
boys, 40 girls) took part in the study (as the Austrian
children were learning to
read in German, they are referred to as “German”
children throughout
this paper
for ease of exposition).
There were three different age groups: 7-year-olds
(23
English, 30 German),
&year-olds
(24 English, 30 German),
and 9-year-olds
(25
English, 21 German).
The children of each country were chosen to be at about
the same point in their school career. This meant in practice that the German
children
were somewhat
older than the English children.
The youngest
group
were in the middle of their second year of schooling (English children: mean age
7 years 0 months, range 6 years 3 months to 7 years 6 months; German children:
mean age 7 years 10 months, range 7 years 4 months to 8 years 5 months), the
middle group were in their third year of schooling (English children: mean age 8
years 0 months, range 7 years 7 months to 8 years 6 months; German children: 8
years 10 months, range 8 years 4 months to 9 years 10 months), and the oldest
group were in their fourth year of schooling (English children: 9 years 0 months,
range 8 years 7 months to 9 years 5 months; German children: 9 years 11 months,
range 9 years 2 months to 10 years 10 months).
The method of teaching reading was somewhat different for the German and
the English children. The school attended by the German children used a rather
systematic
phonics approach.
The main characteristics
are that all graphemes
including the multi-letter
graphemes are slowly introduced
and immediately
used
for word recognition
via grapheme-phoneme
conversion
and blending.
The first
graphemes
are for vowels and continuants
to make phoneme blending easier. In
the beginning
children
are induced
to utter the assembled
pronunciations
as
“word preforms”,
which are then followed by the correct pronunciations.
There
are also attempts to make children aware of correspondences
between larger parts
of word spellings and word pronunciations
by presenting
lists of rhyming words
with identical spelling patterns.
The school attended
by the English children introduced
reading by using a
combination
of a phonics and a whole word reading scheme. The whole word
scheme
used “look-and-say”,
and the phonics
scheme placed emphasis
on
individual
grapheme-phoneme
correspondences
for single letters and blends.
“Word patterns”
were also taught to the children in terms of “families”
with the
H. Wimmer, II. Goswami I Cognition 51 (1994) 91-103
same
rhyming
or beginning
to be used throughout
sounds.
This mixture
of teaching
methods
95
continued
the first few years.
Procedure
Each
continuous
child was seen for a single experimental
reading
tasks were
administered.
session,
These
in which three
were respectively
different
a numeral
reading task, a number word reading task, and a nonsense
word reading task.
Each task consisted of two lists of 18 items, created by including each of the nine
items selected for the study twice in each list. The presentation
of the lists was
intended
to mimic “real” reading, and so the items appeared in sequence printed
left-to-right
on a single page in separate
lines of text. The numerals
were
positioned
to correspond
with the positions of the first letter in each number word
or nonsense
word.
The two lists for each task were given in immediate
succession.
However,
although the order of the items in each pair of lists was varied, the order of the
items in the lists for each task was the same. Thus if one numeral list began “2,
10, 7 . .“) the corresponding
number
word list would
begin
“two,
ten,
seven. . .“, and the nonsense
word list would begin “thro, sen, feven . .“. An
additional
ordering
constraint
was that the numbers
never appeared
in ordinal
sequence in any list. There were six different orders of the pairs of lists. The child
was asked to read each list as quickly and as accurately
as possible.
If a child
paused for too long on a particular
item, he or she was encouraged
to miss that
item out and continue reading. The children were timed on their reading of each
of the six lists, and any errors were noted.
Prior to receiving the experimental
tasks, the children were given a practice
session in which examples
of all three tasks were presented.
In the practice
session, lists of six items were used, and the child was encouraged
to read through
the items as quickly as possible. The aim was to familiarize the children both with
the need to read fast, and with the characteristics
of the different tasks.
Results
Performance
was analysed in two ways: in terms of reading speed per item, and
the overall number of errors per task. Reading speed was averaged across the two
lists used for each task.’ Very few children made errors in reading the numerals or
‘These measures
were highly reliable across lists, although there were so few errors on numeral
and number word reading that reliabilities
could only be computed
for reading time. The partial
correlations
(with age partialled out) between the reading time for the two lists of numerals were 0.79
(English)
and 0.83 (German).
The corresponding
correlations
for the number words were 0.86 and
0.93, respectively.
For the nonsense words, the correlations
were 0.87 (English) and 0.85 (German).
96
H. Wimmer,
the number
U. Goswami
words
aloud.
I Cognition
Four
51 (1994) 91-103
German
children
made
a single
error
in reading
numerals,
and one child made two errors. Six German
children made a single
error in reading the number words, and one made two errors. No English children
made errors in reading the numerals,
but seven children made errors in reading
the number
words,
the number
of errors
ranging
from
2 to 16. Although
the
number
of children
erring with the number
words was the same in the two
orthographies,
some of the English children who erred made many more errors.
The most striking difference
in the number of errors across the two orthographies occurred for the nonsense
words, despite the fact that a very lenient
scoring criterion was used. Nonsense word pronunciations
were scored as correct
whenever a real word analogue for the chosen pronunciation
existed. Thus for the
nonsense
word nour, pronunciations
that rhymed with our and tour as well as
with four were counted as correct. The number of errors made by the different
age groups is shown in Table 1. For the German children, the number of errors
was low even for the youngest
children,
whereas for the English children the
number of errors was much higher, and remained much higher. It is remarkable
that the oldest English children were making more errors in reading the nonsense
despite their greater reading
exwords than the youngest
German
children,
perience.
The raw error scores were subjected
to a 2 X 3 (orthography,
English,
German) x Age (7, 8 and 9 years) analysis of variance. The analysis showed only
a significant
main effect of orthography,
F(1,147) = 42.9, p < ,001, confirming
that the German children made significantly
fewer errors at every age level. This
implies that the German
children
found it easy to assemble
pronunciations,
whereas the English children did not. The high number of errors made by the
English children in nonsense word reading cannot be explained by a tendency to
trade accuracy for speed. In fact, there was a positive correlation
between the
number
of errors
and reading
time in both orthography
groups,
but this
correlation
was much higher for the English children
(r = .74) than for the
German
children (r = .22).
As well as the difference
in the central tendency
of the error scores across
orthography,
there was a large difference in the dispersion of the errors as well.
Table 1 shows that the standard deviations of the English children were between
Table
1.
Means (medians)
words
and standard deviations of errors for the nonsense
M W4
SD
Y-year-olds
8-year-olds
7-year-olds
English
German
English
German
English
German
12.3 (12)
9.9
4.8 (3.5)
4.4
13.4 (12)
11.6
2.6 (2)
2.0
8.8 (4)
9.7
3.4 (3)
2.5
H. Wimmer,
U. Goswami
two and five times larger than those of the German
I Cognition
51 (1994) 91-103
children.
This reflects
97
the fact
that the range of errors was much greater for the former group. Inspection
of
individual
error scores showed that 22 English children read the majority (19 or
more) of the nonsense words incorrectly,
whereas 16 English children made either
no errors,
or only one error.
age of these two groups.
words incorrectly.
Interestingly,
No German
child
there
read
was no difference
the majority
in the mean
of the nonsense
The remarkable
differences
in the nonsense word reading skills of the English
children
could be taken as evidence for two distinct reading styles: an indirect
style of word recognition
relying on letter-sound
conversion,
and a direct style
based on whole-word
recognition.
If this were the case then one might expect that
the good nonsense
word readers (who rely on indirect strategies) ,should be slow
in reading the number words, whereas the poor nonsense
word readers (direct
strategies)
should exhibit fast and accurate reading of these frequent
number
words. Closer inspection
of the data, however, revealed that this was not the case.
The poor nonsense
word readers were also much slower at reading the number
words, and made errors on them. They took on average 1.26 s per number word
versus 0.5 s for the good nonsense word readers, and some of the poor nonsense
word readers made a large number of errors (four of the children made 10 or
more errors, while 14 made no errors). Therefore,
the poor non-word readers are
better characterized
as being generally poorer readers than as readers who rely on
wholistic word recognition.
The kind of errors made by readers of the two orthographies
provided further
evidence for a difference in reading strategy. The German children never refused
to attempt to read a nonsense
word, whereas there were 37 refusals from the
English children, 28 of these from the youngest group. For the youngest German
group, the majority of errors were other nonsense words (e.g., “zweir” for zwier,
“sum” for stint, “breun”
for dreun). This supports the view that these children
were reading the nonsense words by assembling pronunciations.
Furthermore,
an
extremely
high correlation
between
number
word reading speed and nonsense
word reading speed (r = .93) was found for this group, suggesting that the same
strategy was being adopted in reading the number words. For the young English
children, the majority of the errors were real words. Many of these were plausible
real word guesses based on orthographic
similarity.
The most frequent error was
to read sen as “seen”.
Other examples of errors include “thing” for thrine, and
“twix” (the name of a popular chocolate snack) for tix. The two older German
groups made very few errors overall. The errors that were made were largely real
word errors, many of these being number words (e.g., “sieben”
for zieben). The
older English groups continued
to produce real word errors, and very few of these
errors were number
words. Finally,
for the English children,
the correlations
between
number word reading time and nonsense
word reading time increased
with age, r = .58, r = .57 and r = .71, respectively.
For the German
children,
98
H. Wimmer, ZJ. Goswami
Table
2.
I Cognition 51 (1994) 91-103
Means (medians) and standard deviations of the reading time per item
for numerals,
number words and nonsense words
7-year-olds
8-year-olds
9-year-olds
Tasks
English
German
English
German
English
German
Numerals
M (Md)
SD
0.56 (0.53)
0.12
0.68 (0.68)
0.19
0.59 (0.58)
0.13
0.59 (0.58)
0.13
0.52 (0.50)
0.13
0.53 (0.50)
0.12
Number words
M (Md)
0.82 (0.61)
SD
0.57
1.39 (1.07)
0.89
0.83 (0.64)
0.66
0.71 (0.64)
0.25
0.62 (0.56)
0.26
0.60 (0.50)
0.22
Nonwords
M (Md)
SD
2.14 (1.88)
1.14
2.90 (2.50)
1.95
1.53 (1.42)
0.61
2.03 (1.39)
1.35
1.30 (1.22)
0.48
3.30 (2.31)
2.83
these correlations
decreased with age (r = .93, r = .72 and r = .65, respectively).
Differences
in reading time largely reflected the pattern of differences
found
for the errors. Mean reading times per item for the three tasks are shown in Table
the largest differences
2, separated
by orthography
and age. As expected,
occurred for the nonsense
words, which the German children read consistently
more quickly than the English children. For the number words and the numerals,
there was an apparent difference in reading time for the the youngest groups only,
the German children being slightly slower.
A 2 x 3 x 3 (orthography,
German,
English) x Age (7, 8 and 9 years) X Task
(numerals,
number
words, nonsense
words) analysis of variance
taking mean
reading time per item as the dependent
variable showed the predicted interaction
between
orthography
and task, F(2,294) = 24.3, p < .OOl. Exactly the same
interaction
was found when the mean reading
time per item measure
was
log-transformed
to reduce the effect of outliers, F(2,294) = 27.5, p < .OOl. Post
hoc inspection
(Newman-Keuls)
of the former interaction
confirmed that it arose
from a difference
in reading time for the nonsense
words only. The German
children read these words significantly
faster than the English children (p < .Ol).
No other interactions
with orthography
were significant.
As was the case for
errors, the variance in speed was much greater for the English children than for
the German children. The effects of orthography
are thus highly consistent across
both measures.
Discussion
The results of this study are very straightforward.
The only difference,
was a remarkable
one, between
the English and the German
children
and it
was in
H. Wimmer,
reading
the nonsense
words.
A substantial
U. Goswami
number
I Cognition
of English
51 (1994) 91-103
children
99
at each age
group had enormous
difficulty in deriving acceptable
pronunciations
for these
words, while for German children - even for the youngest ones - nonsense
word
reading posed little difficulty. In number word and numeral
reading,
however,
children
learning
to read in the two orthographies
performed
at very similar
levels. The youngest
English children were even slightly faster at these tasks,
ruling out the possibility
that differences
in age or in amount
of schooling
accounted
for the nonsense
word differences
that were found.
The difficulties that the English children experienced
in reading the nonsense
words are even more surprising when the nature of these words and the set-up of
our task are taken into account. The nonsense words shared the same onset and
rime grapheme
clusters as the number words, and were repeated four times over
the two lists. Furthermore,
the children received a practice trial with number
words before beginning
the experiment,
which may have encouraged
them to
think of using number word analogies when they saw the nonsense words. So the
nonsense
word reading task should have been an easy one, especially given the
lenient scoring system that was used, in which every plausible pronunciation
for
the nonsense
words was counted as correct.
The fact that even the youngest
German
children
experienced
very little
difficulty
with nonsense
word reading
suggests that their preferred
reading
strategy is an indirect one. This interpretation
is also supported
by the close-toperfect
correlation
between
number
word reading
time and nonsense
word
reading time, the tendency
to produce nonsense
word errors, and the lack of
reading refusals shown by these children.
In contrast,
the difficulty in reading
nonsense
words, the lower correlation
between number word reading time and
nonsense
word reading time, the preponderance
of word errors, and the reading
refusals shown by the young English children suggest that a substantial
number of
them may have been relying on some kind of direct recognition
strategy
in
reading with little ability to assemble pronunciations
for nonsense
words. This
difficulty in nonsense
word reading was still observed
for some of the oldest
English
children.
On the other hand the older German
children
might be
expected
to show a corresponding
impairment
in the acquisition
of fast direct
word recognition.
However,
there was little evidence
for this. The 9-year-old
German
children tended to read the number words as fast as the corresponding
digits. This suggests that in the case of frequent words they relied on direct access
to pronunciation.
In summary,
the German children appeared
to move into reading by relying
heavily on word recognition
via assembled
pronunciation,
and from there move
on to direct word recognition
for frequently
encountered
words. The English
children,
in contrast,
tended to move into reading by relying on direct word
recognition.
These differing
approaches
may plausibly
be interpreted
as a
combined
effect of the difference
in orthographic
consistency
between German
and English and the instructional
regimes in the two countries.
As mentioned,
the
100
H. Wimmer,
CJ. Goswami
I Cognition
51 (1994) 91-103
German children had been exposed to a systematic phonics based approach, while
the English children
had been exposed to a combination
of whole-word
and
phonics
methods.
However,
it is unlikely
that the present
findings
can be
explained
difference
certainly
solely in terms of the differences in instructional
in teaching could itself arise from orthographic
easier to teach grapheme-phoneme
correspondences
English.
Secondly,
the finding that the oldest
more errors in reading nonsense
words than
approach. Firstly, this
consistency,
as it is
in German
than in
group of English children made
the youngest
group of German
children
goes directly against such an explanation.
The oldest English children
would have received more instruction
in phonics than their younger
German
counterparts
even though this instruction
was combined
with other methods.
The pattern
shown by the young English children - successful performance
with the number
words, and difficulty with the nonsense
words-can
be interpreted in two ways. The traditional
dual-route
assumption
has been that nonsense
word reading
depends
on assembled
phonology
(grapheme-phoneme
conversion), which in some well-known
developmental
models of reading is represented
by a stage of alphabetic
reading that supplements
the initial direct (logographic)
stage (Frith, 1985; Marsh et al., 1981). According
to these models, the English
difficulty in nonsense word reading would be explained by some children relying
largely on a logographic
reading strategy, and having a rather under-developed
alphabetic
strategy.
More recently,
an alternative
conceptualization
of the logographic
stage has
been proposed.
According
to this reformulation,
even the earliest recognition
units are not purely visual, but are linked to phonology (e.g., Ehri, 1992; Perfetti,
1992; Stuart & Coltheart,
1988). The proposal is that the recognition
units are
composed
of some letters which are linked. to phonemic
segments
of the
corresponding
entries in the mental lexicon. For example, the recognition
unit for
a word like “bat” might be b. t. As this recognition
unit is only partial, it would
be instantiated
not only by bat, but also by words like bet and boat.
While we agree that successful young readers of English use a direct access
strategy
that has phonological
underpinning,
in our view this phonological
underpinning
would initially be predominantly
at the onset-rime level. There are
several pieces of evidence
for this idea. Firstly, it has been shown that an
important
predictor of early reading in English is a pre-school sensitivity to onset
and rime (e.g., Bradley & Bryant,
1985; Maclean,
Bryant,
& Bradley,
1987).
Secondly,
it has been shown that children can read new words by analogy from
the beginning
of learning to read, and that these analogies are based on onsetrime units. If children are taught to read a word like beak, then they can use this
word as a basis for reading a new word like peak (rime analogy), and if they are
taught to read a word like trim, then they can use it as a basis for reading trot
(onset analogy; Goswami, 1986, 1988, 1991). Finally, phonologically
able children
make more such analogies
(Goswami,
1990; Goswami
& Mead, 1992). These
H. Wimmer,
that
children
with
U. Goswami
good
I Cognition
phonological
51 (1994) 91-103
awareness
skills
101
findings
suggest
establish
in terms
recognition
units for beak and trim that are phonologically
underpinned
of the grapheme clusters that correspond
to the onset and the rime (see
will
also Goswami,
1993). Children
with poorer phonological
awareness
skills may
develop recognition
units with erratic or weak underpinning
at the onset-rime
level.
The idea that some children establish phonologically
underpinned
recognition
units from the beginning
of learning to read while others do not concurs with the
tremendous
variation
in English nonsense
word reading found in the present
study. Some English children,
even in the youngest group, had no difficulty in
reading the nonsense words correctly, suggesting that they were making analogies
to known words, or had already developed
some context-sensitive
graphemephoneme
rules. Others, however, made many mistakes in reading the nonsense
words, despite reading all of the number words correctly. For these children, the
direct recognition
units underlying
number
word reading
may have had inadequate
phonological
underpinning,
and so could not easily be used as a basis
for analogies to the nonsense words. Such inadequate
phonological
underpinning
could be a consequence
of poor phonological
skills. Although
the phonological
skills of individual
children were not measured
in this study, the link between
phonological
ability and nonsense
word reading is well-established
(e.g., Baddeley, Ellis, Miles, & Lewis, 1982; Frith & Snowling,
1983; Olson, Davidson,
Kliegl, & Foltz, 1985).
Our proposal that it is phonological
underpinning
at the onset-rime
level that is
important
in establishing
direct recognition
units may apply to German,
too, but
such graphemic
units may only emerge later in reading development.
There is no
reason to assume that German children are not aware of onsets and rimes from
the beginning
of reading,
but they initially rely on grapheme-phoneme
correspondences
when decoding the orthography.
However,
a continued
reliance on
assembled
pronunciations
would not allow German
children
to become
fast
readers and accurate spellers, whereas the establishment
of recognition
units for
written words would. Take the example of reading the word Strand (beach). The
first time that a German
child sees this word, he or she may assemble
the
pronunciation
for the initial consonant
cluster “str-” by a laborious
process of
grapheme-phoneme
conversion.
The phonologically
able child, however,
may
code the graphemic
cluster Str- as the onset, setting up a direct recognition
unit
for Strand that includes this coding. This recognition
unit would then allow the
fast decoding of other words that begin with the same onset by analogy to Strand,
like StruJ3e, Strafe, Stroh, or Strung.
There is evidence
for this part of our developmental
model too. Wimmer,
Landerl, and Schneider (in press) investigated
the relationship
between onset and
rime sensitivity
and reading
development
in German,
using the oddity task
developed
by Bradley and Bryant (1985). The onset-rime
measures,
which were
102
H. Wimmer, lJ. Goswami
taken
before
the German
I Cognition 51 (1994) 91-103
children
went to school,
early reading development
in German.
German
depends on grapheme-phoneme
were only weakly
predictive
of
This is not surprising,
as early reading in
translation.
However,
the onset-rime
measures did become predictive later in the reading process. Both reading speed
and spelling accuracy were highly correlated with these measures at around 9-10
years of age. This suggests that word recognition
units that are underpinned
at the
onset-rime
of learning
level are indeed
to read.
established
for German
children,
but at a later phase
In summary,
our model proposes that phonological
skills at the onset-rime
level play a crucial role in structuring
the recognition
units underlying
direct
access in any orthography.
The timing of direct access will depend on orthographic consistency
and instructional
bias. As our results show, English and
German
children
differ in the point at which they begin using direct access
strategies in reading. In the early phases of learning to read German,
the highly
transparent
orthography
encourages
word recognition
via grapheme-phoneme
conversion.
This strategy confers the obvious advantage
of allowing a child to
read almost any new word that is encountered.
Less transparent
orthographies
such as English make the use of grapheme-phoneme
translation
to recognize
words less reliable, and encourage
a reliance on direct access. In the later phases
of reading, where the goal is fast reading for meaning,
these early differences
in
reading strategies
will diminish.
In order to achieve fast reading for meaning,
children learning to read in any alpahabetic
orthography
need to develop direct
word recognition
strategies
and stop assembling
pronunciations
via graphemephoneme
translation.
The use of direct recognition
ensures automaticity.
So it is
the nature
of the orthography
that determines
initial differences
in reading
strategy,
and the nature of the task that ensures eventual convergence
towards
direct access.
References
Baddeley,
A.D.,
Ellis, N.C., Miles, T.R.,
& Lewis, V.J. (1982). Developmental
and acquired
dyslexia: a comparison.
Cognition,
11, 185-199.
Bradley, L., & Bryant, P.E. (1985). Rhyme and reason in reading and spelling. Ann Arbor: University
of Michigan Press.
Ehri, L.C. (1992). Reconceptualizing
the development
of sight word reading and its relationship
to
recoding.
In P. Gough, L. Ehri, & R. Treiman
(Eds.), Reading acquisition (pp. 107-143).
Hillsdale,
NJ: Erlbaum.
Ehri, L.C., & Wilce, L.S. (1985). Movement
into reading: is the first stage of printed word learning
visual or phonetic?
Reading Research Quarterly,
20, 163-179.
Frith, U. (1985). Beneath the surface of developmental
dyslexia. In K.E. Patterson,
J.C. Marshall, &
M. Coltheart
(Eds.), Surface dyslexia (pp. 301-330).
Hillsdale, NJ: Erlbaum.
Frith, U., & Snowling, M. (1983). Reading for meaning and reading for sound in autistic and dyslexic
children.
British Journal of Developmental Psychology, 1, 329-342.
H. Wimmer, II. Goswami
I Cognition 51 (1994) 91-103
103
Goswami,
U. (1986). Children’s use of analogy in learning to read: A developmental
study. Journal of
Experimental Child Psychology, 42, 73-83.
Goswami,
U. (1988).
Orthographic
analogies
and reading
development.
Quarterly Journal of
Experimental Psychology, 4OA(Z), 239-268.
Goswami,
U. (1990). A special link between rhyming skills and the use of orthographic
analogies by
beginning
readers. Journal of Child Psychology and Psychiatry, 31, 301-311.
Goswami,
U. (1991). Learning about spelling sequences:
the role of onsets and rimes in analogies in
reading.
Child Development, 62, 1110-1123.
Goswami,
U. (1993). Towards an interactive
analogy theory of reading development:
decoding vowel
sounds in early reading. Journal of Experimental Child Psychology.
Goswami,
U., & Mead, F. (1992). Onset and rime awareness
and analogies
in reading.
Reading
Research Quarterly, 27(Z), 153-162.
nursery
rhymes and reading
in early
Maclean,
M., Bryant,
P. & Bradley,
L. (1987). Rhymes,
childhood.
Merril-Palmer
Quarterly, 33, 255-281.
Marsh, G., Friedman,
M., Welch, V., & Desburg,
G. (1981). A cognitive developmental
theory of
reading acquisition.
In G.E. MacKinnon
& T.G. Waller (Eds.), Reading research: advances in
theory and practice (Vol. 3, pp. 199-221).
New York: Academic
Press.
Olson,
R.K.,
Davidson,
B.J.,
Kliegl,
R., & Foltz, G. (1985).
Individual
and developmental
differences
in reading disability. In G.E. MacKinnon
& T.G. Waller (Eds.), Reading research:
advances in theory and practice (Vol. 4, pp. l-64).
New York: Academic
Press.
Perfetti, C.A. (1992). The representation
problem in reading acquisition.
In P.B. Gough, L.C. Ehri,
& R. Treiman (Eds.), Reading acquisition (pp. 145-174).
Hillsdale, NJ: Erlbaum.
Seymour,
P.H.K., & Elder, L. (1986). Beginning reading without phonology.
Cognitive Neuropsychology, 3(l), l-37.
Stuart, M., & Coltheart,
M. (1988). Does reading develop in a sequence of stages? Cognition, 30,
139-181.
Wimmer,
H., & Hummer,
P. (1990). How German speaking first graders read and spell: doubts on
the importance
of the logographic
stage. Applied Psycholinguistics, II, 349-368.
Wimmer,
H., Landerl,
K., & Schneider,
W. (in press). The role of rhyme awareness
in learning to
read a regular orthography.
British Journal of Developmental Psychology.
Appendix:
the nonsense word lists
English
German
thro
nee
nour
twive
tix
feven
thrine
sen
felve
nei
fei
zwier
siinf
vechs
zieben
dreun
zwehn
sdf