Content uploaded by Lesly Wade-Woolley
Author content
All content in this area was uploaded by Lesly Wade-Woolley on Mar 12, 2014
Content may be subject to copyright.
http://jlr.sagepub.com/
Journal of Literacy Research
http://jlr.sagepub.com/content/25/4/383
The online version of this article can be found at:
DOI: 10.1080/10862969309547827
1993 25: 383Journal of Literacy Research
Esther Geva, Lesly Wade-Woolley and Michal Shany
Different Orthographies
The Concurrent Development of Spelling and Decoding in Two
Published by:
http://www.sagepublications.com
On behalf of:
Literary Research Association
can be found at:Journal of Literacy ResearchAdditional services and information for
http://jlr.sagepub.com/cgi/alertsEmail Alerts:
http://jlr.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
http://jlr.sagepub.com/content/25/4/383.refs.htmlCitations:
What is This?
- Dec 1, 1993Version of Record >>
by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from by guest on October 11, 2013jlr.sagepub.comDownloaded from
Journal
of
Reading Behavior
1993,
Volume
25, No. 4
THE CONCURRENT DEVELOPMENT
OF
SPELLING
AND DECODING
IN
TWO DIFFERENT ORTHOGRAPHIES
Esther Geva
and
Lesly Wade-Woolley
Ontario Institute
for
Studies
in
Education
Michal Shany
Beit Berl College
of
Education
ABSTRACT
The hypothesis that differences between first language
(LI) and
second language
(L2) reading
and
spelling profiles could
be
accounted
for by
lack
of
proficiency
in
the
L2 or
differences
in
orthographic complexity
was
explored
in a
longitudinal
study
of
45 children acquiring reading
and
spelling skills concurrently
in
English
(L1)
and
Hebrew
(L2). The
children were tested
in
Grades
1 and 2 on
literacy
measures
in
both languages. Neither
of
these explanations alone sufficed
to ex-
plain
the
development
of
reading
and
spelling
in the two
languages.
The
less
complex Hebrew orthography facilitated subjects' decoding performance,
but
failed
to maintain that facilitation
in
spelling. Depressed second language effects were
apparent
in
spelling
but not in
decoding, which actually favoured
the
subjects'
L2.
Developmental findings showed that, despite
L1-L2
differences
in
ortho-
graphic complexity
and
language proficiency,
the
profiles
of
emergent spelling
in
both languages
are
strikingly similar.
The
rate
of
acquisition
of
conventional spell-
ing, however, differentiates
L1
from
L2
performance.
The Role of Language Proficiency
There is ample evidence to show that reading comprehension in second lan-
guage (L2) is related to L2 proficiency. The role of language proficiency needs to
be examined also in considering the development of basic decoding skills in L2.
An intuitively appealing hypothesis is that individuals would read more accurately
383
384 Journal of Reading Behavior
in their first language (LI) than in their L2. Two complementary perspectives on
language proficiency, varying in level of specificity, can be conceptualized as con-
tributing to the development of basic decoding skills in L2. According to a global
perspective, it could be argued that language profiency can be mapped onto compo-
nents such as fluency, syntactic and lexical knowledge and sensitivity to
psycholinguistic appropriateness of utterances (Carroll, 1981; Geva & Ryan, 1993;
McLaughlin, 1991). Alternatively, viewed from a perspective focusing on cogni-
tive processing, it should be possible to map linguistic proficiency onto more basic
individual differences in underlying processes, such as auditory discrimination,
phonological representations, memory for phonemes and for verbal information,
and speed of lexical access. All these components, regardless of level of specificity,
should be more developed in LI than in the early stages of L2 development, and
should therefore lead to relatively better reading skills in LI than in L2.
The Role of Orthographic Depth
According to the "Orthographic Depth Hypothesis" (Frost, Katz, & Bentin,
1987;
Katz & Frost, 1992; Turvey, Feldman, & Lukatela, 1984) there are differ-
ences between alphabetic orthographies in terms of how systematically spelling
and pronunciations can be mapped onto each other. A shallow orthography allows
a simple one-to-one correspondence between letters and sounds. Conversely, a deep
orthography, although still abiding by the alphabetic principle, employs a more
complex set of relationships between letters and sounds. Relatedly, frequency and
lexicality effects in word recognition (e.g., word/nonword judgments) are larger
with deep orthographies such as unvowelized Hebrew and English than with shal-
low orthographies such as Serbo-Croatian (Peeremen, 1992). The effects of ortho-
graphic depth on beginning readers has been recognized by educators for several
decades; early studies using modified teaching orthographies showed a facilitating
effect when compared with readers' performance on traditional English orthogra-
phy (Gillooly, 1973). It is possible that the facilitating effects of a shallow orthog-
raphy will be observed in reading and spelling regardless of whether it is the learners'
LI or L2. One hypothesis, therefore, is that individuals will learn to read and spell
accurately faster when a shallow orthography is involved; in fact, the regular struc-
ture of the orthography may override whatever processing difficulties may be im-
posed by weaker proficiency in the L2.
Orthographic depth is typically used to describe a factor that influences the
process of decoding. The fact that it influences spelling as well is generally tacitly
assumed; however, it must be made clear that the level of orthographic depth does
not necessarily remain the same for decoding and spelling. The translation of pho-
nemes to graphemes and graphemes to phonemes may require different rules
(Venezky, 1970). In consequence, it is possible for the same language to be shallow
for decoders but deeper for spellers. When there is more than one graphemic can-
didate for a given sound, the speller must evaluate each possibility; this is an added
Spelling and
Decoding
385
dimension that contributes to a higher processing load for spelling than for decod-
ing. It follows logically that the greater the number of candidates for a given
spelling, the greater the chance of error on the part of a novice speller. This sug-
gests that the orthographic depth hypothesis needs to be refined by taking into
account a "depth differential" between the processing demands of decoding and
spelling tasks. This depth differential may be equally applicable to beginning spellers
in both their first language (LI) and second language (L2). The implications of
this refinement of the orthographic depth hypothesis will become clearer in the
comparison of features of Hebrew and English orthography below.
English Orthography
It is generally agreed that the orthographic cipher of English is complex. En-
glish has 26 letters which map onto more than 36 phonemes, and in order to deter-
mine the phonemic value of some graphemes one needs to consider more than one
letter. For example, as Gough, Juel, and Griffith (1992) point out, in English the
phonemic value of b, d, f, I, n, r, v, r can be determined in word-initial positions
without considering other letters. For all the other letters, at least one additional
letter must be identified before the value of the first phoneme in a word can be
determined. Sometimes the reader needs to consider at least four more letters be-
fore the first phoneme can be unambiguously determined (e.g., "chord" vs. "chore").
At other times, the addition of one grapheme can radically change the pronuncia-
tion of a letter elsewhere in the word, as in the case for the "silent e" long vowel
marker. In addition, there are idiosyncratic words in English, like "yacht," which
need to be learned individually since they cannot be pronounced either by rule or
analogy on the basis of their spelling.
One of the chief causes of the depth of English orthography is its
morphophonemic nature (Chomsky & Halle, 1968; Venezky, 1970). Both phono-
logical and morphological information is encoded in the English writing system,
and phonological regularity is often sacrificed for the sake of the preservation of
meaning. For instance, health and heal have little phonological similarity, yet their
semantic connection is obvious from their written forms.
Hebrew Orthography
Hebrew, like English, is an alphabetic orthography, but is read from right to
left. Unlike English, Hebrew script in its vowelized version can be described as
"shallow," in that there is a direct and systematic correspondence between the twenty
two consonants and the associated phonemes (Frost & Bentin, 1992). Hebrew is an
example, however, of a language where decoding depth and spelling depth are not
identical. Five consonants have a different final form, employed when they appear
at the very end of a word. Therefore, for these consonants, spellers must consider
two orthographic candidates for a given phoneme, as opposed to the one-to-one
386 Journal of Reading Behavior
relationship that holds in reading. Additionally, Hebrew orthography allows cer-
tain silent consonants in word final position (i.e., aleph n , heh 71 , ayin ¿i ).
When decoding, these consonants receive a null phonological representation; the
shallowness is preserved since these are always silent. When spelling, the indivi-
dual must be aware, a priori, that these silent consonants are present, since no phon-
ological stimulus signals their presence. For example, in the Hebrew word >&*
(/imaJ,
mother), the decoder pronounces a word that begins and ends with a vowel
sound, but the speller must remember that the orthographic representation includes
silent consonants at the beginning and the end. If, however, an emergent speller
omits the silent consonants in phonetic spelling (e.g., $H ), the output can still be
decoded unambiguously as /ima/. Therefore, even at its deepest, vowelized Hebrew
is shallower than English.
The difference between vowel sounds and vowel letters is paramount here.
Hebrew vowel graphemes, in the form of dots and dashes, are placed below, above,
or to the left of the consonants. These vowels are quite close to the cardinal vowels
in their phonetic realization (Berman, 1978). Although it is a common practice for
printers to omit vowels from texts designed for fluent readers, children (and L2
learners) learn to read first with the vowels intact. Likewise, expert spellers omit
vowels when they write.
Unlike English, the pronunciation of words in Hebrew very rarely requires
readers to consider exceptions associated with specific letter strings, idiosyncratic
spellings, or position in the word and vowel combinations. The formation of letter-
sound associations and word decoding requires the learner to master fewer rules,
and does not entail the creation of a large repertoire of exception words read on the
basis of analogy rather than rule (e.g., have vs. gave). For this reason, accurate
word decoding in Hebrew can be learned easily, and can be executed in a linear and
systematic fashion (Geva & Siegel, 1991). Modern Hebrew includes borrowed words,
but their orthographic realization is, on the whole, transparent. It is possible to
take advantage of Hebrew's regularity and learn to decode without any oral profi-
ciency in or semantic knowledge of the language. This does not mean, of course,
that lexical, syntactic and semantic information does not contribute to efficient
lexical access and to comprehension (Shimron & Navon, 1982).
The Relationship Between Early Reading and Spelling in LI
There seems to be convergence among researchers as to subtle steps in the
development of early reading skills by young children learning to read English as
their LI (Perfetti, 1992). For instance, Gough et al. (1992) point out that the acqui-
sition of the "cipher" (knowledge of spelling-sound correspondence rules) gener-
ally increases the child's ability to recognize words, and that it does not take place
in a uniform fashion across different words (see also Perfetti, 1992). The acquisi-
tion of the cipher influences the probability that the child will commit reading
errors,
because English contains many words that cannot be successfully decoded
Spelling and
Decoding
387
on the basis of simple letter-sound correspondences. Cipher reading also influ-
ences the nature of errors. When children begin to internalize the cipher they are
more likely to commit nonword errors than real word errors.
With the development of the cipher, the child's ability to generate spellings
increases, but it does not happen across the board in an all or none fashion. Juel,
Griffith and Gough (1985) found a significant correlation between a decoding task
and a spelling task. Gough, Juel, and Roper/Schneider (1983) maintain that the
acquisition of the cipher enhances correct spelling, and that the quality of children's
spelling errors is affected by the acquisition of the cipher. Support for this comes
from Gough et al. (1983) who found that cipher readers make more phonetic spell-
ing errors than "code" readers, who rely on a selective association of words based
on distinctive visual cues in written words. Furthermore, Juel et al. (1985) found
that code readers perform poorly on pseudoword reading, and in spelling often
include one salient correct letter of a target word, but all other letters may be ran-
dom.
A complementary position holds that children pass through an intermediate
stage between the code and cipher stages proposed by Gough et al. (1992) to de-
scribe how children read words. In this stage, "phonetic-cue reading" (Ehri, 1992),
only a few relevant cues are considered by the child; these important cues link the
spelling to the pronunciation of the word. The crucial element in this process is the
linkage of the grapheme with its phonemic value, which for most English letters, is
easily recoverable from the letter name
itself.
Bradley and Bryant (1985) argue that children use different strategies for reading
and spelling. Gough et al. (1992), however, maintain that the cipher is the basis for
skilled reading and skilled spelling. They agree, however, that in spite of a high
degree of consistency between spelling and reading, which can be attributed to
their common basis (i.e., the cipher), spelling cannot be seen as simply a reflection
of reading, since "polygraphy is not the same as polyphony" (p. 47). Willows (1993)
points out that the task demands involved in generating correct spellings are heavier
than those involved in decoding.
Early Reading/Spelling Acquisition in L2: Theoretical and Methodological
Issues
Not much is known on developmental stages in the acquisition of basic read-
ing skills in L2. For this reason the relevance of the above discussion for compar-
ing reading acquisition across different orthographies and for studying L2 reading
acquisition requires careful examination. It is not completely clear on what basis
one can determine the equivalence of task demands implicated in decoding iso-
lated words in different languages. For example, word frequency is generally taken
to reflect stronger lexical representation (Seidenberg, 1992), and frequency effects
in LI readers are well-documented. However, one cannot assume that the fre-
quency of a word in one language mirrors that of its translation in another Ian-
388 Journal of Reading Behavior
guage. Furthermore, L2 learners by definition have limited lexical knowledge in
their second language, so measures of word frequency which are derived from
distributional characteristics in the target language cannot be used reliably to cal-
culate L2 processing factors.
Likewise, nonword error patterns in L2 decoding and spelling cannot be un-
ambiguously attributed to a particular reading stage, as Gough et al. (1983) sug-
gest for LI reading. For the less proficient L2 learner, there may be no qualitative
difference between successfully decoding an unfamiliar word, and making decod-
ing errors that result in a nonword, since nonnative speakers are accustomed to
encountering unknown words which have the status of nonwords.
Additionally, knowing the names of L2 letters may not be as good an index of
knowledge of those letters' phonemic values, an observation often associated with
research on English as LI (Ehri, 1992), simply because of the lack of early child-
hood experience with letter-naming activities typical of North American prereaders
(Lundberg & Hoiem, 1991; Naslund & Schneider, 1991). Evidence for this comes
from a recent study by Geva, Shany and Himel (1992), which found similar levels
of decoding accuracy in both English (LI) and Hebrew (L2), despite the fact that
children made no letter-naming errors in English, compared to a significant inci-
dence of such errors in their second language.
Another point that must be raised is the assumption of phonological compe-
tence in the L2. One of the challenges facing children learning a second language
is the lack of phonological representation of L2-specific sounds. The theoretical
link between phonology and reading skills is well established; however, most of
the research that informs this position has focused on English native language
readers. In studying differences between LI and L2 reading profiles, one may need
to take into consideration the effects of interlanguage phonologies. It is possible
that, as in English, the necessary cue resides in the letter name, but if there is no
corresponding representation in the learner's L2 phonological repertoire, then er-
rors on these elements may be expected to reflect this lack.
Finally, existing literacy skills in the LI may influence the course of reading
acquisition in the second language. Transfer effects of LI skills to L2 may be more
substantial when older literate learners are considered. In this study, possible ef-
fects due to previous literacy instruction were minimized by investigating the par-
allel development of decoding and spelling skills in children who were acquiring
these skills concurrently in their LI and L2.
Notwithstanding the caveats listed above, an economical account of L2 read-
ing development favours similar developmental stages in both LI and L2 (at least
where alphabetic languages are concerned). Thus one would expect, for instance,
that the progression from cipher to code stages of reading offered by Gough et al.
(1992) would be evident in both LI and L2. A complementary hypothesis derived
from the theoretical perspective that language proficiency is the primary influence
on literacy acquisition might predict differences in rate of reading and spelling
Spelling
and
Decoding
389
acquisition due to L1-L2 proficiency differences, with an advantage to LI due to
higher proficiency in LI. In such a scenario, the transition from stage to stage may
be slower in the L2 than in the LI.
At the same time, the role of differences in orthographic depth cannot be ig-
nored. A hypothesis derived from the theoretical position that orthographic depth
is the primary influence of literacy acquisition might predict differences in the rate
of acquisition favoring the less complex orthography, regardless of language profi-
ciency. Thus, Hebrew, as the shallower orthography, would be acquired faster than
the English, despite its status as the subjects' less proficient second language. As
well as considering orthographic depth across languages, one must also take the
task-specific "depth differential" into account. For example, it is generally agreed
that English is equally deep for spellers and decoders; therefore, one might expect
a similar level of performance on spelling and reading measures in English, since
both require an elaborated knowledge of complex orthographic patterns. In He-
brew on the other hand, knowledge of orthographic complexity associated with
silent letters and homophones is challenged to a much greater degree in spelling
than in decoding. As a result, one might expect learners to perform better on He-
brew decoding than on Hebrew spelling tasks.
Whereas the relationship between reading and spelling has been amply re-
searched in first language studies, there is no corresponding theory in the arena of
second language. A parsimonious account of these processes in L2 favours similar
developmental stages in both LI and L2 (at least where alphabetic languages are
concerned). At the same time, the development of a more realistic model also re-
quires an examination of differential language proficiency and orthographic com-
plexity.
METHOD
Subjects
Children participating in the study came from an elementary, English-Hebrew
private day-school in a large metropolitan area. The classes in the school range
from junior kindergarten (age 4) to Grade 8. The majority of children in the school
come from middle-class homes in which English is the primary language. Begin-
ning in Grade 1, children are exposed each day to a regular English program, as
well as to 2-3 periods of Hebrew studies. The Hebrew program includes language
and reading as well as cultural content. The academic program is fairly demand-
ing; the school day starts at 9:00 in the morning and continues to 4:00 in the
afternoon. Within each class the English and Hebrew programs are taught by dif-
ferent teachers. Reading instruction includes a combination of phonics and whole
language methods.
390 Journal of Reading Behavior
A letter was sent to all the children in the 3 Senior Kindergarten (SK) classes
(age range 5 to 6). Only those children whose parents signed a consent form which
described the nature of the study, the anonymity of participants, the confidentiality
of individual scores and that they were free to withdraw from the study at any time,
took part in the study. Results reported here are based on the performance of chil-
dren with English as their first or best language (as determined by information
provided by teachers as well as a brief student interview questionnaire). Of the 90
children in that age group, the parents of 55 children provided their consent. Of
the original 55 subjects, 2 had Hebrew as their best language, and their data was
not included in subsequent analyses. In addition, 3 of the children tested in SK did
not continue attending this bilingual English-Hebrew day-school and 2 moved to
another city. Three additional children left the school after Grade 1. Of the 55
children in the original SK cohort, complete data for all three years were available
for 45 children (22 boys and 23 girls; average age 66.18 months).
Measures
Independent Variables
Language. The subjects in this study served as their own controls since they
performed on parallel tasks in English and Hebrew.
Grade level. Due to the longitudinal nature of this study, development is likely
to be a factor in subjects' performance; therefore, dependent measures were taken
on all tasks in both Grades 1 and 2.
Receptive vocabulary. The Peabody Picture Vocabulary Test-Revised (PPVT)
(1981) was used as a measure of receptive vocabulary.
Intervening Variables
The following variables were conceptualized as component processing skills,
and used as predictors in multiple regressions analyses:
Phoneme segmentation. The Auditory Analysis Test (Rosner & Simon, 1971;
hereafter, Rosner) is a measure of phonological awareness and segmenting ability,
in which the child is asked to segment and delete syllables and phonemes from
familiar words. Since this task was administered in SK, prior to any systematic
instruction of Hebrew, only an English version was administered.
Phonological recognition. A task developed by Olson, Kliegl, Davidson, and
Foltz (1985) and revised by Siegel (1993) was administered to examine children's
developing repertoire of English words they can recognize on the basis of assembled
phonological information. In this task (hereafter EPHONO), the child is asked to
Spelling
and
Decoding
391
identify which of two pseudowords in a stimulus pair (e.g.,joak-joap) "sounds like
a real word." A parallel task (hereafter HPHONO ) was developed in Hebrew with
the real words drawn from the repertoire of words taught in the Grade 1 program
(e.g., spinning top, /s'vivon/-/s'vivod/). There are 26 stimulus pairs in EPHONO
and 20 pairs in HPHONO. Results are based on percent correct.
Visual recognition. A task developed by Olson et al. (1985) and revised by
Siegel (1993) was administered to examine children's developing repertoire of words
they can recognize from visual memory. The task (hereafter EVISUAL) consists of
26 homophonic word pairs, where one of the items is spelled correctly (e.g.,
rain-rane).
Subjects must identify which of the items in each stimulus pair is spelled
correctly. A parallel task consisting of 20 stimulus pairs was developed for Hebrew
(hereafter HVISUAL). All the correctly spelled words consist of words taught in
the Grade 1 program (e.g., /torah/-/tora/). Results are based on percent correct.
Dependent Variables
English word recognition. To assess children's ability to read isolated words in
English, the word recognition subtest of the Wide Range Achievement Test-Revised
(WRAT-R) (Jastak & Jastak, 1984) was used. This test (hereafter EWORD) con-
sists of 75 unrelated words. It begins with highly familiar and short words, but as
the child progresses through the list the words become less frequent, longer and
orthographically more complex. In keeping with the administration directives of
this test, responses arrived at through slow syllable-by-syllable decoding were scored
as incorrect. Testing stops when the child commits 10 consecutive errors.
The content validity of the WRAT has been questioned; critics have charged
that the items have not maintained pace with changes in curricula. Given that the
predominant philosophy in current reading instruction does not favour the use of
basal readers, it becomes increasingly difficult to assume that any one standardized
test of word recognition can fulfill the content validity requirements across differ-
ent reading programs. Additionally, some concerns have been raised about the test
standardization and norming properties of the WRAT and its applicability for di-
agnostic purposes (Conoley & Kramer, 1989). Yet, since one of the key research
questions in this study revolves around the development of reading from Grade 1 to
Grade 2 and the subjects act as their own controls, these criticisms, although valid
in themselves, may be less crucial.
Hebrew word recognition. To test children's ability to read words in their L2,
a test developed by Geva (see Geva & Siegel, 1991) based on the vocabulary taught
in the Grades 1-5 program in the bilingual Hebrew day-school system was used.
This test (hereafter HWORD) consists of 60 unrelated words with the vowels in-
tact. Two variables were manipulated in the design of this test: familiarity and
word length. The first 30 words come from the school curriculum and should be
familiar to all children even in Grade 1. The familiar list was constructed in two
392 Journal of Reading Behavior
steps:
The first involved the selection of common, recurring words in the curricu-
lum identified during classroom observations by the test developer. The second
step involved teacher judgments of word familiarity. Only words judged to be fa-
miliar in the second step were included. The remaining 30 words are highly infre-
quent in everyday Hebrew, a fact confirmed by judgments of 4 Hebrew native speakers
with graduate degrees, and do not appear in the curriculum.
Within each familiarity condition, the first 10 are one-syllable words, the next
10 are two-syllable words, and the last 10 are three- or four-syllable words, for
example, /ner/ (candle), /shanah/ (year), and /maxberet/ (notebook) in the familiar
set, and /tel/ (archaeological digging
site),
/meitzar/ (strait), and /anafah/ (heron)
in the unfamiliar set, respectively. Administration procedures are identical to those
described above for EWORD. Based on data of 245 Grades 1-5 children, Geva
and Siegel (1991) report that the Cronbach alpha for this test is 0.92.
English pseudoword decoding. The Word Attack subtest of the Woodcock Read-
ing Mastery Test-Revised (Woodcock, 1987) was used to assess children's ability
to employ various orthographic rules to decode pseudowords. This test (hereafter
EPSEUDO) consists of 50 pseudowords which comply with English orthographic
rules.
Testing stops when the child makes 6 consecutive errors. Pseudoword decod-
ing is taken to be a reliable measure of phonological skills. Results are based on
percentage scores.
Hebrew pseudoword decoding. To test the ability of children to employ word
attack skills in Hebrew, a pseudowords task (hereafter HPSEUDO) was employed
(Geva & Siegel, 1991). The test consists of 44, two-syllable pseudowords. Admin-
istration procedures are identical to those described above for EPSEUDO. Results
are based on percent correct. The Cronbach alpha for this test is >95.
English spelling. The Developmental Spelling Task (Ferroli & Shanahan, 1987)
was used to find out about children's development and knowledge of word ele-
ments in English, as revealed in their spelling and error patterns. The list consists
of 12 simple and frequent words which reflect various aspects of English orthogra-
phy (e.g., long vowels, morphological endings).
Hebrew spelling. A parallel task in Hebrew was created for this study. In an
attempt to minimize the effects of lack of familiarity with word meanings and
exposure to the written form of words in Hebrew, the Hebrew spelling task was
developed from the range of vocabulary explicitly taught in the Grade 1 curricu-
lum. It includes 15 words which address various orthographic principles in He-
brew such as the marking of plurality (e.g., /tsva'im/, colors), the spelling of the
/ax/ syllable in word final position (e.g., /tapuax/, -apple), and the spelling of the
/ts/
sound in word final position (e.g., /mits/, juice).
Three scoring systems, varying in level of specificity, were used in this study
to encode various levels of spelling knowledge in both English and Hebrew. First,
an overall measure of absolute accuracy was taken. Words were counted as cor-
Spelling and
Decoding
393
rectly or incorrectly spelled, according to the rules of conventional spelling, and
the percentage of correctly spelled words was calculated. These measures will be
referred to hereafter as ESPELL% and HSPELL%.
The second scoring scheme locates spellings along a continuum conceptual-
ized by Ehri (1986). This model of emergent spelling comprises three stages; a
score of 0 was given to spellings that were reflective of the "semiphonetic" stage;
characteristic errors at this stage may involve lack of medial vowels, incomplete
spellings, or arbitrary extra letters (e.g., dlake for lake; stk for stick). Spellings that
were typical of the "phonetic" stage received a score of 1. Children spelling at this
level adhere rigorously to a "one letter per sound" principle, and consequently
represent both long and short vowels by one letter. They may also tend to add
nonarbitrary letters that they hear when they stretch the word out as they spell it
(e.g., deragin for dragon). Spellings reflective of the final stage in this scoring
scheme, the "morphemic" stage, receive a score of 2. At this stage, children start to
become cognizant of particular spelling patterns, word families and morphemic
encodings. For example, they may show an understanding that the past tense is
spelled ed, regardless of the fact that it sometimes has a phonological realization of
N, as in
peeked.
Conventional spellings received a score of 3.
The third scoring system provided a sublexical analysis to detect specific areas
of difficulty. The spelling items were divided into sublexical units that appeared to
be most interesting on the basis of the preceding scoring. These orthographic ele-
ments were classified into two categories. The first category contained transparent
elements; these are orthographic patterns that are easily recoverable from their
phonological representations, such as "bl" in black and simple short vowels. The
second category contained opaque elements which cannot be unambiguously de-
rived from their pronunciations, but they may be the product of a highly productive
rule (e.g., the "silent e rule" for long-vowel formation in English and final-form
consonants in Hebrew). For scoring purposes, each spelling element in each of the
two categories received a 0 if it was not represented at all, a 1 if the representation
was partial or incorrect, and a 2 if conventional spelling was achieved.
Procedures
The data reported here are part of a longitudinal project concerned with the
development of parallel LI and L2 literacy indices, and the role of various linguis-
tic and cognitive measures (e.g., phonological awareness, verbal ability, memory)
in understanding and predicting this development. Children were tested in late
winter or early spring in each of 3 consecutive years. Children were first tested
when they were in Senior Kindergarten (SK), and then again in Grade 1 and Grade
2.
The PPVT-R and Rosner were administered in SK; The parallel LI and L2
reading and spelling measures, as well as the parallel LI and L2 visual recognition
and phonological recognition tasks, were administered to these children in Grade
1,
and a year later when they were in Grade 2. Testing was done on an individual
394 Journal of Reading Behavior
basis.
Each of the LI and L2 reading and spelling task clusters was administered in
separate sessions by university students. LI testing was done by English native
speakers, L2 testing was carried out by Hebrew native speakers.
RESULTS AND DISCUSSION
The Relationships among Emerging Reading and Spelling Skills in LI and L2
Table 1 provides a summary of the relationships among spelling and reading
indices in Grades 1 and 2, as well as the correlations among these measures, recep-
tive vocabulary (PPVT), the phonological segmenting task (Rosner), and the vi-
sual and phonological recognition tasks. Four general observations can be drawn
from an examination of this table. First, one notes that within each language, even
though the words used to assess word recognition and spelling were not identical,
all tasks were positively and highly intercorrelated. For instance, in Grade 1 the
correlation between ESPELL% and EWORD was r=.82, and with EPSEUDO it
was
r=.81.
Although the overall strength of the correlations remained high for the
Hebrew tasks, there was a drop in the strength of the correlations. For instance, the
correlation between HSPELL% and HWORD was r=.57, and between HSPELL%
and HPSEUDO r=.62. This pattern of strong within-language correlations not only
confirms previous findings reported in the literature on the fact that the emergence
of decoding and spelling skills tend to be strongly interrelated but also extends it to
L2 acquisition.
The second observation concerns the extent to which the same facilitative in-
terrelationships between reading and spelling hold across languages. The answer
to this question is not straightforward. Hebrew spelling correlated equally with
reading measures in both languages. For instance, the correlation between
HSPELL% and EWORD was r=.12, and with EPSEUDO r=.65. However, English
spelling correlated significantly but only moderately with the Hebrew word recog-
nition and pseudoword measures (r=.39, and n=.44, respectively).
Next, one notes that in Grade 2 the within-language correlations among spell-
ing and reading measures were somewhat attenuated. At the same time the cross
language patterns described above continue to be evident, namely, positive and
significant relationships between the Hebrew spelling and the English decoding
tasks,
but virtually no correlation between the English spelling measure and the
Hebrew reading measures.
On the whole, the generally high intercorrelations among English and Hebrew
measures in Grade 1 and their subsequent drop in Grade 2 suggest that as the
children develop there is a gradual "unpacking" and increased differentiation of
LI and L2 linguistic skills. This trend is also evident by an examination of the
correlations among the Rosner and the measures in English and Hebrew. In fact, a
striking result can be seen in the highly predictive role of the Rosner vis-à-vis
Table 1
Intercorrelations Matrix of Reading and Spelling Indices in Grade 1 and Grade 2 (n=45)
a
Grade 1 EWORD
H WORD
EPSEUDO
HPSEUDO
E VISUAL
H VISUAL
EPHONO
HPHONO
ESPELL%
H SPELL %
ROSNER
PPVT
Grade 2 EWORD
H WORD
EPSEUDO
HPSEUDO
E VISUAL
H VISUAL
EPHONO
HPHONO
ESPELL%
H SPELL %
ROSNER
PPVT
E
WORD
1
.554
.855
.588
.734
.492
.707
.494
.823
.722
.615
.039
1
.543
.670
.395
.607
.314
.450
.346
.655
.519
.413
.023
H
WORD
1
.504
.876
.331
.454
.397
.560
.386
.569
.316
.010
1
.565
.664
.481
.403
.477
.554
.293
.619
.374
.009
E
PSEUDO
1
.532
.667
.566
.660
.563
.812
.648
.679
.011
1
.472
.475
.196
.534
.360
.675
.487
.439
.015
H
PSEUDO
1
.319
.516
.461
.637
.437
.618
.385
.033
1
.454
.429
.281
.624
.158
.524
.256
.002
E
VISUAL
1
.428
.609
.242
.719
.565
.477
.015
1
.381
.372
.660
.569
.413
.367
.036
H
VISUAL
1
.404
.432
.520
.563
.301
.008
1
.030
.684
.203
.320
.216
.023
E
PHONO
1
.451
.666
.600
.449
.027
1
.360
.340
.295
.471
.001
H
PHONO
1
.457
.531
.505
.068
1
.172
.409
.381
.006
E
SPELL%
1
.655
.581
.004
1
.520
.251
.007
H
SPELL% ROSNER
1
.409 1
.010 .064
1
.296 1
.016 .064
a.
O
1
a.
•Correlations above .29 are significant atp<.05. Correlations above .37 are significant
atp<.01.
396 Journal of Reading Behavior
English word recognition and word attack skills in Grade 1 and its dramatic drop
in Grade 2. At the same time, the correlations of the Rosner with the parallel
Hebrew tasks were more moderate but remained fairly stable across both grade
levels.
This suggests that the role played by phonological skills is more stable in
Hebrew than it is in English. In English, phonological processing is necessary but
not sufficient; it is offset by the need to acquire numerous orthographic rules. (See
also complementary results from multiple regressions in Table 3.) Although most
of the research on the Orthographic Depth Hypothesis and the added processing
demands of deep orthographies like English, as compared to more shallow orthog-
raphies, has been carried out with adult readers, these findings strengthen the evi-
dence that ties it in with early spelling and decoding skills in young school children.
Development of LI and L2 Reading and Spelling
As the reader can see from summary statistics in Table 2, in Grade 1 and in
Grade 2 on measures relying on phonological skills, namely word recognition,
pseudoword decoding and the phonological recognition measures, children's scores
were much higher in Hebrew than in English. A series of repeated measures
ANOVAs indicated that language had a highly significant effect on each of these
measures [F(l, 44)=246.86,
p<.001;
F(l, 44)=41.86,
p<.001;
F(l, 44)=17.56,
/?<.001,
respectively]. However, there was no significant difference between lan-
guages on the visual recognition measure [F(l, 44)=1.06, p>.05], which measured
subjects' ability to recognize correct spelling patterns with minimal phonological
Table 2
Grade 1 and 2 Means (and Standard Deviations) of Reading and Spelling in
English and Hebrew (n=45)
Word Recognition %
Pseudoword Decoding %
Phonological Recognition %
Visual Recognition %
Spelling %
Grade
English
29.42
(16.31)
27.96
(18.05)
61.88
(14.49)
60.89
(16.59)
34.88
(25.73)
1
Hebrew
65.19
(29.82)
50.42
(32.59)
73.13
(16.75)
64.38
(13.95)
24.81
(20.74)
Grade
English
52.18
(12.69)
60.19
(14.13)
82.04
(17.59)
87.07
(15.37)
72.09
(21.89)
2
Hebrew
81.86
(14.05)
75.00
(30.87)
86.11
(10.89)
87.00
(10.36)
48.21
(18.10)
Spelling
and
Decoding
397
processing. Naturally, there was a main effect for grade level on each of the above-
mentioned variables; that is, children improved their performance from Grade 1 to
Grade 2 on word recognition, pseudoword decoding and the phonological recogni-
tion measures, [F(l, 44)=30.35,
/x.001;
F(l, 44)=38.70,
/x.001;
F(l, 44)=42.96,
p<.001,
respectively] and also on visual recognition [F(l, 44)=95.22,p<.001]. The
only significant interaction that was found occurred on the spelling task. The in-
crease in spelling performance from Grade 1 to Grade 2 was significantly larger in
English than in Hebrew [F(2, 43)=10.45, p<.01]. In other words, contrary to the
pattern observed on the phonologically based tasks, subjects' spelling performance
in Hebrew consistently lagged behind that in English. These latter findings are
analyzed below in greater detail.
Results showed that, as expected, accuracy scores on children's English Grade
1 word recognition and spelling were not significantly different from each other
[i(44)=-1.99, p>.06]. However, in Hebrew, children were much more accurate in
reading than in spelling [i(42)=10.71, p<.0001]. As indicated above, in Grade 2
they improved on all tasks, but their spelling accuracy of English words was sig-
nificantly higher than their reading accuracy [r(44)=-7.81, p<.001]. In Hebrew, the
advantage of the shallow orthography for reading was maintained; reading scores
were significantly higher than spelling scores [f(42)=14.27,/?<.0001]. At this point,
these intriguing results need to be interpreted with caution since the word recogni-
tion and spelling tasks did not utilize the same items. At the same time, given the
pattern of within-language correlations between these tasks (discussed earlier), it
is very likely that future research, using identical items would yield similar results.
Prerequisite Linguistic Skills across Orthographies
To determine relevant predictor variables of reading and spelling tasks in each
grade, a series of stepwise multiple regressions was carried out. In each analysis,
experimental tasks tapping component processes were used as predictors. These
included the Rosner as well as the Visual and Phonological Recognition tasks in
both languages. The PPVT was not included since it did not correlate with any of
the reading and spelling measures (see Table 1). A number of observations can be
made from Table 3, which provides a summary of these analyses. First, in Grade 1,
phonological skills (captured by the Phonological Recognition task and the Rosner)
contributed significantly to the variance of those reading tasks that have a high
phonological demand, specifically, pseudoword decoding in both languages and
word recognition in Hebrew. Although phonological skills play a role in word rec-
ognition in English, success depends also on the recognition of specific visual
patterns, as evidenced by the significant proportion of the variance explained by
the Visual Recognition task. Likewise, accurate spelling of sight words in both
languages also requires memory for distinctive visual sequences; to this end one
notes that a substantial proportion of the variance in spelling scores was explained
by the Visual Recognition task. In both languages tasks which present to the child
Table 3
Significant Predictors of Reading and Spelling in Grades 1 and 2 (Stepwise Multiple Regression)"
Dependent
Variables
EPseudo
HPseudo
EWord
HWord
ESpell
%
HSpell
%
Significant
Predictors
Rosner
EPhono
HVisual
HPhono
HVisual
EVisual
EPhono
Rosner
HPhono
EVisual
HPhono
EPhono
HVisual
Grade
1
Cumulative
F
34.15
31.57
28.20
27.39
17.82
46.83
35.62
29.57
18.27
42.89
29.57
22.45
18.16
Adjusted
R
2
Change
45
15
7
39
6
53
10
5
30
51
9
34
11
Total
Adjusted
R
2
67
45
68
30
60
45
Significant
Predictors
EVisual
EPhono
HPhono
EVisual
EPhono
HVisual
EVisual
HPhono
Grade
2
Cumulative
F
12.88
10.67
17.90
25.67
15.73
15.11
20.65
8.72
Adjusted
R
2
Change
22
10
30
38
26
15
32
16
Total
Adjusted
R
2
32
30
38
41
32
16
r
a.
a"
•p<-01
for all
variables.
Spelling
and
Decoding
399
similar challenges were predicted by similar or identical component measures.
It is noteworthy that in comparison with the Grade 1 results, in Grade 2 there
was an overall reduction in the proportion of variance explained by phonological
and visual recognition component skills. This may have been due to the overall
reduction in variability (as noted also by the smaller standard deviations; see Table
2).
At the same time an interesting trend begins to emerge in Grade 2: Hebrew
spelling and reading were predicted mainly by the Phonological Recognition task,
but the English spelling and reading tasks are predicted mainly by the Visual Rec-
ognition task.
In sum we see that as they shift from the cipher stage to the "post" cipher
stage, phonological and visual processes play a complementary role in facilitating
children's performance on reading and spelling in English and Hebrew. Yet, per-
formance on parallel tasks in both languages was not predicted equally by the same
variables. Instead, in support of the Orthographic Depth Hypothesis, one notes that
even at this early stage of their literacy development, visual word recognition plays
a more prominent role in English than it does in Hebrew. As noted above, this
phenomenon seems to be even clearer in Grade 2 than it is in Grade 1.
Spelling Development in LI and 12
In this section we examined in depth the nature of the development of children's
spelling in LI and L2. Figure 1 provides a graphic summary of this development
for the 12 English spelling words (upper panel) and the 15 Hebrew spelling words
(lower panel), respectively. Spelling level means for each item were based on Ehri's
(1986) conceptualization of spelling development (see Method). The mean scores
for each word were arranged in an ascending order on the basis of the Grade 1
performance on the respective tasks. Three general observations can be made. First,
it is clear that in both languages approximations to standard spelling do not evolve
simultaneously for all words. Second, none of the children spelled any of the En-
glish words at the semiphonetic stage, though in Hebrew there were some excep-
tions (i.e., mean score below
1.00).
In fact, in Grade 1 almost all the words in both
languages were spelled at the phonetic stage (i.e., 1.00 to 1.99 range). In Grade 2
children's spelling of English words can be characterized as morphemic (i.e., 2.00
to 2.99 range). This is true for many of the Hebrew words, as well, but it seems that
the spelling of some words still reflects a phonetic approach. Third, it seems that,
although the order of difficulty defined by the Grade 1 performance was main-
tained in Grade 2 within each language, the progression was more erratic in He-
brew. The "peaks" and "valleys" noted in the Hebrew Grade 2 profile, as well as
the observation that some Hebrew words were still spelled at the phonetic stage,
motivated us to carry out a more fine-grained analysis in a search for the source of
these difficulties (see next section).
400
Journal of Reading Behavior
grade
1
grade
2
o
1
S 2,
£ B
I 1 I
.g
•»
<D
grade
1
grade
2
C3
5 2H
es
e
"cS
eg"
1
t
1
Figure 1. Words ranked for emergent spelling level in English (upper panel)
and Hebrew (lower panel) in Grades 1 and 2.
Sources of Spelling Difficulty in LI and L2
In an attempt to uncover the sources of spelling difficulty in English and He-
brew for young learners, we employed an analysis that went beyond the lexical
level. For this purpose, we developed an additional 3-point error scoring system
which was applied to the orthographic elements in sublexical units in each word. It
reflects the degree to which the child's error comes from the omission of a pho-
Spelling and Decoding
401
neme (0 points), partial/incorrect inclusion (1 point) or standard spelling (2 points).
Orthographic elements across the words within each language were classified as
"transparent" or "opaque" (see Method) on the