You are on page 1of 16

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/45460036

Learning to Read and Spell in Persian: A Cross-Sectional Study From Grades 1


to 4

Article  in  Developmental Psychology · November 2010


DOI: 10.1037/a0020377 · Source: PubMed

CITATIONS READS
11 1,017

2 authors:

Noriyeh Rahbari Monique Sénéchal


Ottawa Hospital Research Institute Carleton University
26 PUBLICATIONS   91 CITATIONS    66 PUBLICATIONS   6,452 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The adverse effect of pre- and postnatal environment on early childhood cognitive and language development as well as long term effects on children’s literacy and
education. View project

Evaluation of PATHS program View project

All content following this page was uploaded by Noriyeh Rahbari on 12 February 2015.

The user has requested enhancement of the downloaded file.


Developmental Psychology
Learning to Read and Spell in Persian: A Cross-Sectional Study
From Grades 1 to 4
Noriyeh Rahbari, and Monique Sénéchal
Online First Publication, August 2, 2010. doi: 10.1037/a0020377

CITATION
Rahbari, N., & Sénéchal, M. (2010, August 2). Learning to Read and Spell in Persian: A
Cross-Sectional Study From Grades 1 to 4. Developmental Psychology. Advance online
publication. doi: 10.1037/a0020377
Developmental Psychology © 2010 American Psychological Association
2010, Vol. ●●, No. ●, 000 – 000 0012-1649/10/$12.00 DOI: 10.1037/a0020377

Learning to Read and Spell in Persian:


A Cross-Sectional Study From Grades 1 to 4

Noriyeh Rahbari and Monique Sénéchal


Carleton University

We investigated the reading and spelling development of 140 Persian children attending Grades 1– 4 in
Iran. Persian has very consistent letter–sound correspondences, but it varies in transparency because 3 of
its 6 vowel phonemes are not marked with letters. Persian also varies in spelling consistency because 6
phonemes have more than one orthographic representation. We tested whether lexicality effects—an
advantage of words over nonwords—would be affected be reading transparency and spelling consistency.
We found that children became more efficient readers and spellers across grades, with the greatest growth
occurring between Grades 1 and 2. For reading, lexicality effects were present with transparent words
starting in Grade 2, but lexicality effects with opaque words were not yet present in Grade 4. As expected,
the size of transparency effects for reading decreased across grades. For spelling, however, there was no
lexicality effect for either consistent or inconsistent words. Moreover, consistency effects were large and
did not decrease systematically across grades. Most interesting from a developmental perspective was the
finding that both reading transparency and spelling polygraphy affected reading as well as spelling in
Grades 1 and 2, but the word characteristics had differential effects as a function of literacy task in Grades
3 and 4. This pattern highlights the vulnerability of children’s representations and processes during the
early phases of acquisition as well as the rapidity with which representations and processes become
specialized as a function of the literacy task at hand.

Keywords: reading, spelling, acquisition, Persian, consistency

Skilled readers make efficient use of phonological, ortho- to be agreement, however, that phonological and orthographic
graphic, and semantic representations to extract meaning from processes develop interdependently and that phonological recod-
written texts. How children acquire these representations and how ing plays a critical role in the development of well-defined ortho-
these representations become interconnected has been of interest graphic representations (Bowey & Muller, 2005; Cunningham,
for more than 100 years now (Rayner, Foorman, Perfetti, Pesetsky, Perry, Stanovich, & Share, 2002; Kyte & Johnson, 2006; Share,
& Seidenberg, 2001). Advances in our understanding are being 1999).
made, and developmental models are being proposed (e.g., Share, Perfetti (1992) proposed that as children learn to read, the
1999). Most interesting is the fact that we are now testing the limits orthographic representations of words become increasingly spe-
of models elaborated to account for findings in English to other cific (i.e., more correct letters represented in their correct position)
orthographies (e.g., Frost, Katz, & Bentin, 1987; Ziegler & Gos- and become redundant with the phonological representations of the
wami, 2005). In the present study, we examined whether specific words (also see Ehri, 1992). As more and more words become
characteristics of the Persian orthographic system would affect the well-specified orthographically and redundant phonologically (i.e.,
acquisition of reading and spelling in Persian, a language that has
stronger associations between orthographic and phonological rep-
received little attention to date.
resentations), reading becomes less effortful. This framework has
Learning to read in alphabetic languages is complex, as evi-
been used to explain individual differences in reading (Perfetti &
denced by the diversity of models elaborated. Although stage-
Hart, 2001), but it can also be applied to examine cross-linguistic
models of reading propose qualitative shifts in children’s strategy
differences in the acquisition of reading. It is possible that the
use (e.g., Frith, 1985; Gough & Juel, 1991), alternative views
propose gradual quantitative changes as children become increas- acquisition of well-defined orthographic representations varies as a
ingly sensitive to the internal structure of written words as well as function of the transparency of writing systems (e.g., Defior,
to the correspondences between letters or groups of letters in Martos, & Cary, 2002; Ziegler & Goswami, 2005).
words and their pronunciation (Rayner et al., 2001). There seems Children who learn to read in transparent orthographies—when
letter–sound correspondences are consistent—acquire reading
faster in comparison with children who learn to read orthographies
that have less direct and more equivocal letter–sound correspon-
dences (e.g., Bruck, Genesee, & Caravolas, 1997; Goswami,
Noriyeh Rahbari and Monique Sénéchal, Department of Psychology,
Carleton University, Ottawa, Ontario, Canada.
Ziegler, Dalton, & Schneider, 2003). For instance, children learn-
Correspondence concerning this article should be addressed to Noriyeh ing to read in transparent orthographic systems, such as Finnish,
Rahbari, Department of Psychology, Carleton University, Ottawa, Ontario German, or Dutch, show higher accuracy (Aro & Wimmer, 2003)
K1S 5B6, Canada. E-mail: nrahbari@connect.carleton.ca and faster reading rates (Seymour, Aro, & Erskine, 2003) than

1
2 RAHBARI AND SÉNÉCHAL

children learning to read in less transparent orthographies, such as words have the same orthographic representation and, conse-
English. quently, would require additional contextual information to be
Even though transparent orthographies seem to ease reading read. Homographs can be represented in English with the sequence
acquisition, transparent orthographies also have peculiarities that bd that could represent the words bid, bad, or bed. In the present
can affect performance. For instance, languages such as Arabic research, no ambiguous words were used. Specifically, all stimuli
and Persian can be viewed as transparent languages because they in the present study had a single correct pronunciation, a single
have very consistent letter–sound correspondences; yet, these lan- correct spelling, and a single meaning, as in the example for the
guages do not mark with letters some vowel phonemes. As a Persian word for cold. Moreover, the words consisted of a single
consequence of this peculiarity, many words in these languages are morpheme. Persian words for which all phonemes are represented
represented as a series of consonants. Abu-Rabia has examined the with letters were labeled transparent words, and the words that
issue of the presence and absence of vowels in numerous studies of include vowels not represented by letters were labeled opaque
reading in Arabic (for a review, see Abu-Rabia & Taha, 2006). The words. The term “transparency” rather than consistency was se-
findings are clear: Readers are slower and more error prone when lected to highlight the presence or absence of a vowel letter in
reading words with vowels omitted as compared with reading the Persian that can result in ambiguity, not inconsistency. Transparent
same words that included diacritics placed on consonants to mark and opaque words were matched on acoustic duration, frequency,
the vowel phoneme. The pattern applies to adult and children and initial consonant. The words were also matched on number of
readers as well as to skilled and poor readers. The daily experi- letters with the consequence that opaque words had an additional
ences of Arabic (and Persian) readers, however, include texts consonant.
without diacritics for vowels. Therefore, we contend that it is Another key characteristic of the Persian orthography is the
essential to conduct research comparing real words for which relative inconsistency of the sound–letter correspondences. In Per-
vowels are not marked with real words for which vowels have a sian, there are six phonemes that can be marked by more than one
corresponding letter. Rahbari and Sénéchal (2009) developed such consonant letter. Specifically, the phoneme /z/ can be represented
a paradigm for adult readers of Persian that was used in the present by four different letters; /s/ can be represented by three different
study. letters; and /t/, /h/, /a/, and /gh/ can each be represented by two
In contrast to research on children’s reading development, chil- different letters. Obtaining the correct spelling of words that con-
dren’s spelling has not been studied extensively. Nonetheless, tain these polygraphic phonemes requires reliance on orthographic
there are some cross-orthographic studies showing that children knowledge. In the present study, Persian polygraphic (or incon-
learning to spell in consistent orthographies, such as Czech (Cara- sistent) words were those for which more than one letter can
volas & Bruck, 1993) or German (Wimmer & Landerl, 1997), represent the same phoneme, whereas monographic (or consistent)
show faster rates of spelling development than English children. In words did not have alternative plausible spelling, and each pho-
addition, within-orthographic studies reveal that it is easier to spell neme corresponded to a single grapheme.
a word when each phoneme consistently corresponds to a single In Iran, formal schooling starts when children are 6 years of age,
graphemic representation as opposed to words with two or more and elementary schools go to Grade 5. Some schools have kinder-
plausible spellings (Caravolas, Kessler, Hulme, & Snowling, 2005; garten, but it is not mandatory to attend, and emergent literacy is
Juul, 2005). Thus, children’s spelling accuracy appears to be not part of the kindergarten program. From Grade 1 onward,
affected by phoneme-to-grapheme consistency. In the present re- reading is taught with the same series of books in schools through-
search, the effect of spelling consistency was tracked across out the country. The first-grade book contains lessons introducing
Grades 1– 4. Presumably, a consistency effect should be stronger one or two letters along with sentences, and eventually passages,
in the earlier grades because orthographic representations are not containing words with the featured letters. In contrast to English,
yet well defined. in which both letter names and sounds are taught to children, in
Persian only letter sounds are formally taught. In the first half of
Reading and Spelling in Persian the first-grade book, diacritics are used to mark missing vowels,
but diacritics are usually omitted in the second half of the book.
The Persian orthography has consistent letter–sound correspon- The most typical teaching approach in Iran is a combination of
dences because each grapheme has a single pronunciation, but phonics and whole-word instruction. The structure of the books for
Persian is not entirely transparent because three of its six vowels Grades 2–5 is the same across grades in that each lesson starts with
are not represented with letters (for a thorough description of a passage, followed by comprehension questions and vocabulary
Persian, see Baluch, 2006). Specifically, the vowels /æ/, /e/, and /o/ definitions among other activities. Writing is taught with its own
are not represented by letters or diacritics after the first year of series of books that focus on spelling, handwriting, as well as
reading instruction (with the exceptions of poetic texts or to sentence or short essay writing depending on grade level. When
disambiguate a word in a text). In contrast, the vowels /â/, /i/, and spelling to dictation in Grade 1, children are not required to use
/u/ are always represented by letters as a fixed part of the words. diacritics.
Hence, words, or syllables within words, that have unmarked Given that the same books are used for all children, we were
vowels are spelled as a sequence of consonants. Further, these able to assess children’s relative exposure to transparent and
words can have a unique orthographic representation when a single opaque words as well as to monographic and polygraphic words.
word corresponds to the sequence of consonants. For example, the To do so, we analyzed 12 texts from Grades 2–5. For each grade,
consonant sequence srd represents the word cold in Persian and a text from the beginning, middle, and the end of the reading book
can only be pronounced correctly as /særd/. Alternatively, a con- was selected, and all one- and two-syllable words were extracted.
sonant sequence can be ambiguous or homographic when multiple Of the 529 one- and two-syllable words found, 27% were trans-
READING AND SPELLING IN PERSIAN 3

parent, 34% were opaque, and the remaining 39% were two- months] SD ⫽ 3 months), 35 children (53% girls) in Grade 2 with
syllable words that included one transparent and one opaque a mean age of 8;0 (SD ⫽ 3 months), 35 children (54% girls) in
syllable. Moreover, 51% of the 529 words extracted were poly- Grade 3 with a mean age of 8;11 (SD ⫽ 4 months), and 34 children
graphic words, and 49% were monographic words. Hence, we (53% girls) in Grade 4 with a mean age of 10;6 (SD ⫽ 5 months).
concluded that elementary school children were as likely to read Another child in Grade 4 did not complete all required tasks, and
transparent and opaque words as well as monographic and poly- her partial data were not included. The children were recruited
graphic words. from six classrooms for each grade located in three public schools:
Reading and spelling in Persian children and adults have been two schools for girls and one school for boys.
examined in a limited number of studies (Arab-Moghaddam & Parental education ranged from not having completed high
Sénéchal, 2001; Gholamain & Geva, 1999). For instance, Rahbari, school to completing postgraduate degrees. The distribution of
Sénéchal, and Arab-Moghaddam (2007) reported that the unique educational levels was as follows: 16% of mothers and 19% of
contribution of phonological skills (18% and 17% for reading and fathers had not completed high school, 63% of mothers and 43%
spelling, respectively) was substantially higher than that of ortho- of fathers received a high school diploma, 20% of mothers and
graphic skills (3% and 6% for reading and spelling, respectively) 32% of fathers completed college or university, and 1% of mothers
in second-grade children’s reading and spelling. In contrast to and 6% of fathers completed postgraduate studies.
reading in childhood, adult readers of Persian seem to rely on According to their classroom teachers, all participating students
orthographic representations, as is suggested by the presence of had above average reading and spelling ability and had no cogni-
frequency effects (Baluch, 1993; Baluch & Besner, 1991; Nassaji tive or psychological difficulties. Also, all children were native
& Geva, 1999) and lexicality effects (Rahbari & Sénéchal, 2009). speakers of Persian and had normal or corrected to normal vision.
Taken together, the findings reviewed above suggest that phono- Moreover, although it was not possible to match children across
logical recoding plays an important role in Persian children’s grades on the basis of their IQ scores because of the unavailability
reading but that orthographic representations are salient in Persian of a standardized IQ test, all children who participated in the
adults’ word naming. The extant research, however, is limited present study had general averages of 85% in all subjects, includ-
because it does not represent the developmental pattern of skills ing reading and spelling. Note that there is a formal general
used to read—the main goal of the present research. academic examination every 3 months in Iranian schools.

The Present Study Measures


The main objective of the present study was to document the The measures used in the present study were identical to those
progression of Persian reading and spelling across the first four in Rahbari and Sénéchal’s (2009) study with adults, and it is their
years of instruction. In the present study, children read and spelled description that is generally reproduced here with permission from
words and nonwords to test for the presence of lexicality effects. If Elsevier. Participants read and spelled the same 56 words and 38
orthographic representations are readily available, real words nonwords. For both reading and spelling tasks, latencies and error
should be read and spelled faster and more accurately than non- rates were measured. Latencies were the primary dependent mea-
words because nonwords do not have an orthographic representa- sure because of the high accuracy rates found in other research on
tion and, consequently, require phonological processing. Of spe- Persian (Rahbari et al., 2007) and because latencies have been
cial interest was the description of the effect of the orthographic measured successfully in young elementary school children, in-
peculiarities of Persian on the magnitude of lexicality effects. It cluding first-grade children (e.g., Kerek & Niemi, 2009; Pinheiro,
was predicted that the magnitude of the lexicality effect should 1995). Whenever possible, latency analyses were supplemented
increase with grade levels but that the developmental course with error analyses. An explanation of the process of word selec-
should be affected by the transparency of grapheme–phoneme tion and word type as well as the construction of nonwords
correspondences for reading and by the consistency of phoneme– follows.
grapheme correspondences for spelling. Hence, we tested for lex- Word stimuli.
icality effects for reading transparent and opaque words separately, Word type. Participants read and spelled the same 56 real
and for spelling monographic and polygraphic words separately. words to provide an equivalent measure of reading and spelling
We also indexed the developmental course of transparency effects (see Appendix A). The word types were obtained by crossing
(transparent vs. opaque words) for Persian reading and consistency reading transparency with spelling consistency. Reading transpar-
effects (polygraphic vs. monographic words) for Persian spelling. ency included two levels: Transparent words were those for which
It was expected that the size of the transparency effects for reading all phonemes were represented with letters, whereas opaque words
and that the size of the consistency effects for spelling would were those that included vowel phonemes that were not repre-
decrease across grades. sented with letters. Spelling consistency also included two levels:
monographic words included phonemes with unique consistent
Method orthographic representations, whereas polygraphic words included
phonemes with multiple plausible letters and, therefore, inconsis-
Participants tent orthographic representations. This crossing of reading trans-
parency and spelling consistency yielded 14 words that were
A total of 140 monolingual Persian children attending Grades transparent and monographic, 14 words that were transparent and
1– 4 in Tehran, Iran, participated in this study. There were 35 polygraphic, 14 words that were opaque and monographic, and 14
children (57% girls) in Grade 1 with a mean age of 7;0 ([years; words that were opaque and polygraphic. Thus, the 56 words
4 RAHBARI AND SÉNÉCHAL

included a total of 28 transparent words and 28 opaque words for Nonword stimuli. Children read and spelled the same 38
analyzing reading transparency, or, for analyzing spelling consis- nonwords. All nonwords were transparent in terms of reading and
tency, 28 monographic and 28 polygraphic words. Half of the were monographic in terms of spelling. That is, all nonwords had
polygraphic words contained two polygraphic letters, and half vowels marked with letters, and none included polygraphic pho-
contained one polygraphic letter. nemes. Also, nonwords were decodable and had a legal orthogra-
Word selection and characteristics. The words were selected phy (see Appendix B). The nonwords had the same CVC ortho-
from the Persian schoolbooks used in Grades 1–5. All words had graphic structure as the transparent words. The construction and
a single correct pronunciation, a single correct spelling, a single selection of nonwords was as follows. First, nonwords were con-
meaning, and a single morpheme. The 56 items were three-letter structed from real words by changing the initial and/or the final
monosyllabic words. The transparent and opaque words had the consonant or by replacing a long vowel with another long vowel.
same number of letters but had different orthographic structures: The goal was for the nonwords to have little resemblance to real
transparent words were CVC, whereas opaque words were CCC Persian words to force participants to rely on phonological pro-
with an unmarked vowel pronounced after the initial consonant cessing rather than an analogy strategy to read and spell them. The
(i.e., /CVCC/). Most important, the transparent and opaque words list of nonwords was then examined by three educated native
were matched on acoustic duration to ensure that the differences in speakers of Persian to assess nonword resemblance to real words.
coda complexity (/VC/ vs. /VCC/) did not result in longer pronun- If a nonword was identified as similar to a real word in Farsi or any
ciations. Word types were also matched on frequency (see Table 1) other minority language in Iran, it was replaced, and the evaluation
and initial consonant onset. There were no statistically significant was repeated.
differences for acoustic duration or frequency when comparing Nonword reading accuracy was the number of items for which
transparent and opaque words as well as when comparing poly- participants blended correctly all decoded letters into their corre-
graphic and monographic words. sponding sounds, and nonword spelling accuracy corresponded to
Because objective word-frequency norms are not available for the number of items for which participants wrote correctly the
Persian, the procedure used to index word frequency is described sequence of letters that captured the phonology of pronounced
in detail. First, 300 one- and two-syllable words were selected nonwords.
from Persian school books for Grades 1–5. Second, 10 teachers
judged the frequency of each item in written or spoken Persian on Procedure and Apparatus
a 7-point scale ranging from 1 (infrequent words) to 7 (very
frequent words). This method of word-frequency assessment has Participants were tested individually in a quiet room in their
been previously used by other researchers (e.g., Baluch & Besner, schools. The reading tasks were administered at least 1 week
2001; Frost et al., 1987; Ziegler, Ferrand, & Montant, 2004). The before the spelling tasks to prevent any visual or auditory residual
resulting frequency for each word was averaged across teachers. effects. Task administration was not counterbalanced because a
Third, items rated at the end points of the rating scales (i.e., 1 or previous study showed that children who spelled words first per-
7) were excluded to ensure that items were not too familiar or too formed better on reading, whereas the reverse was not true (Rah-
rare. Fourth, words in each category were matched as closely as bari et al., 2007). Likewise, the word list was presented before the
possible on frequency level. Fifth, to ensure that the selected words nonword list for both reading and spelling to prevent participants
were within the spoken vocabularies of first-grade children, we from extending their use of phonological recoding to reading and
asked three first-grade teachers to verify the familiarity of words spelling real words (Baluch & Besner, 1991). There was a 5–10-
for first-grade children. If a word was identified as unfamiliar to min pause between the presentation of words and nonwords for
first-grade children, it was replaced with a familiar word, and the children in Grades 2– 4 in both reading and spelling tasks, but
evaluation was repeated. Finally, 15 skilled readers rated the children in Grade 1 performed all tasks (reading words, reading
frequency of each item in the final list of 56 words as a second nonwords, spelling words, spelling nonwords) in separate sessions.
source of frequency ratings. The frequency ratings from teachers In the present study, data for children in Grade 2 onward were
and skilled readers were highly correlated, r(54) ⫽ .70, p ⬍ .01. collected in the middle of the school year, whereas data for
The averaged frequencies obtained from teachers and acoustic children in Grade 1 were collected at the end of the school year
durations are reported for each word in Appendix A and across because children in Grade 1 typically complete letter learning at
word types in Table 1. the end of the academic year.

Table 1
Mean (and Standard Deviation) Frequency and Duration (in Milliseconds) of Words as a
Function of Reading Transparency and Spelling Consistency

Reading transparency Spelling consistency

Variable Transparent Opaque Monographic Polygraphic

Frequencya 3.48 (0.89) 3.51 (0.67) 3.48 (0.92) 3.29 (0.61)


Duration 731 (0.09) 737 (0.09) 738 (0.09) 730 (0.08)
a
On a 7-point rating scale ranging from 1 (rare) to 7 (frequent).
READING AND SPELLING IN PERSIAN 5

The procedure used in the present study was identical to that in hearing the word to be spelled. Also, measuring spelling onsets
Rahbari and Sénéchal’s (2009) study, and it is their description that avoided any confounds with differences in calligraphic speed.
is reproduced here with permission from Elsevier. The presenta- Given that orthographic representations for inconsistently spelled
tion order of the 56 words was random because studies have shown words would be less well-specified, spelling onsets for polygraphic
that blocking stimuli by type might induce the use of certain words should be slower than for monographic words. Finally, the
processes or strategies (e.g., Kinoshita, Lupker, & Rastle, 2004; instructions for nonwords stipulated that the stimuli were nonsense
Raman, Baluch, & Besner, 2004). Moreover, the random presen- words and that participants should read or spell them as fast and as
tation of transparent and opaque words is closer to an ordinary correctly as possible.
reading situation because the Persian orthography typically con-
tains a mixture of transparent and opaque words. Results
We administered both reading and spelling tasks on a computer
using the SuperLab program, Version 4.0. For reading, all stimuli Preliminary analyses confirmed the absence of outliers and
were presented in a 24-point font size on the center of the com- speed–accuracy tradeoffs. All reading and spelling latencies were
puter screen. A visual ready signal, a “⫹” sign, was presented at within three standard deviations from the participant and the item
the center of the computer screen for 500 ms followed by an item means. Moreover, there were strong positive correlation coeffi-
that remained on the screen until a participant read the item. The cients between latencies and error rates, rs(92) ⫽ .73 and .75, p ⬍
experimental sessions began by the presentation of practice items, .01, for reading and for spelling, respectively.
either six words or six nonwords, to familiarize the participants We performed analyses of variance (ANOVAs) using partici-
with the procedures for each task. Participants were instructed to pants (F1) and items (F2) as the unit of analysis separately for
read aloud each item that appeared on the monitor as quickly and reading and spelling. These analyses were performed both on
correctly as possible. A timer began recording the reading latencies latencies and error rates. Latency analyses were based on correct
when the stimulus appeared on the monitor and ended when the mean latencies. That is, responses for mispronunciations (5%) or
experimenter pressed either a yes or a no button to indicate correct misspellings (9%) were excluded in the latency analyses. In addi-
or incorrect responses, respectively. Therefore, correct and incor- tion, the association between word frequency and latencies was
rect responses were also recorded. The experimenter was trained examined to provide support for variance analyses. As proposed by
not to look at the computer screen but to listen carefully to a Perfetti (1992), children build more specific orthographic repre-
participant and to press the yes or no button as soon as the sentations of words on the basis of their reading experience for that
participant responded. Although using voice-key activation is the word, and as such, it is assumed that orthographic processes are
standard in word-reading research, pilot research clearly showed sensitive to the frequency of each word, whereas phonological
that using a voice detector with young participants tested in processes are not. Throughout the Results section, ps ⬍ .01 unless
schools proved highly unreliable. otherwise indicated, and effect sizes (␩2) as well as 95% lower and
Audio-recordings of the words were presented via the computer upper confidence intervals are reported where appropriate. Anal-
for the spelling tasks. Each participant and the experimenter heard, yses were conducted to test (a) developmental trends, (b) lexicality
at the same time, the auditory stimulus through headphones. Par- effects, and (c) transparency and polygraphy effects.
ticipants were asked to write down the stimulus on a sheet of paper
as fast and as correctly as possible after hearing the stimulus. There Reading Performance
were six practice items for the word and the nonword tasks. The
computer program was set up to record the spelling latencies Mean latencies and error rates for reading transparent words,
between the auditory presentation of the stimuli until the experi- opaque words, as well as nonwords are presented in Table 2.
menter pressed a keyboard key to stop the timer to indicate that the Latency analyses. To analyze children’s reading speed, we
participant started writing the target stimulus. When the participant conducted a 4 ⫻ 3 mixed-factor ANOVA with grade (four levels:
finished writing an item, the experimenter pressed another key on Grades 1– 4) as the between-subjects factor and stimuli (three
the keyboard to present the next auditory stimulus. Spelling onset levels: transparent word, opaque word, and nonword) as the
was measured on the basis of the assumption that phonological and within-subject factor. As indicated in Table 2, older children read
orthographic representations would be activated immediately after faster than did younger children. Indeed, the ANOVA revealed a

Table 2
Reading Mean Correct Latencies in Milliseconds (and Standard Deviation) and Mean
Percentage of Errors (and Standard Deviation) for Nonwords, Transparent Words, and Opaque
Words as a Function of Grade

Latencies % Error

Grade Nonword Transparent Opaque Nonword Transparent Opaque

1 1,814 (327) 1,848 (384) 3,030 (571) 5.11 (4.77) 1.62 (3.16) 22.55 (10.60)
2 1,592 (232) 1,455 (248) 2,044 (482) 1.80 (2.83) 1.12 (1.89) 14.69 (9.93)
3 1,414 (237) 1,240 (204) 1,685 (354) 0.23 (0.98) 0.41 (1.44) 12.55 (9.50)
4 1,311 (185) 1,058 (152) 1,317 (225) 0.23 (0.99) 0.31 (1.84) 4.31 (3.81)
6 RAHBARI AND SÉNÉCHAL

significant main effect of grade, F1(3, 135) ⫽ 75.66, ␩2 ⫽ .63; The ANOVAs revealed significant main effects of transparency
F2(3, 91) ⫽ 573.29. Follow-up analyses showed that both linear for each grade, Fs1(1, 34 [1, 33 for Grade 4]) ⫽ 428.72, 181.24,
and quadratic trends were significant. Although children read 109.06, and 101.03; ␩s2 ⫽ .93, .84, .76, and .75; 95% CIs [1,094,
faster with each passing grade, the quadratic trend is due to the fact 1,317; 531, 729; 329, 530; and 244, 333]; Fs2(1, 52) ⫽ 170.06,
that the greatest increase in speed occurred between children in 55.62, 62.42, and 38.36 for Grades 1, 2, 3, and 4, respectively. As
Grade 1 (M ⫽ 2,231 ms) and the remaining children (Ms ⫽ 1,697, presented in Table 2, transparent words were read significantly
1,446, and 1,229 ms for Grades 2, 3, and 4, respectively). faster than opaque words in all grades. However, the magnitude of
There was also a significant main effect of stimuli, F1(2, 135) ⫽ the differences between transparent and opaque words decreased
576.66, ␩2 ⫽ .81; F2(2, 91) ⫽ 93.41, as well as a Grade ⫻ Stimuli as the grades progressed (ds ⫽ 1,182, 589, 445, and 259 ms for
interaction, F1(6, 135) ⫽ 90.76, ␩2 ⫽ .67; F2(6, 91) ⫽ 80.39. Grades 1, 2, 3, and 4, respectively).
These results suggested that children’s reading speed differed on The main effect of polygraphy was significant by participants
the basis of stimuli type across the grades, and, consequently, for children in Grades 1 and 2 only, that is, children in Grades 1
lexicality effects were tested for each grade separately. Lexicality and 2 read monographic words faster than polygraphic words,
effects were examined with contrasts comparing nonwords with Fs1(1, 34) ⫽ 113.41 and 78.11; MSEs ⫽ 1,822.66 and 581.01;
transparent and opaque words, respectively. Indeed, lexicality ef- ␩s2 ⫽ .76 and .60; and 95% CIs [39, 189; and 51, 144] for Grades
fects were present for transparent word reading from Grade 2 1 and 2, respectively. It makes intuitive sense that spelling incon-
onward. As indicated in Table 2, starting in Grade 2, children read sistency would affect reading at a time when detailed orthographic
transparent words significantly faster than nonwords, Fs1(1, 34 [1, representations are being established. The interaction between
33 for Grade 4]) ⫽ 27.25, 55.35, and 175.23; 95% CIs [75, 172; transparency and polygraphy was also significant by participants
117, 206; and 207, 283]; Fs2(1, 64) ⫽ 9.15, 35.38, and 89.88 for only, Fs1(1, 34 [1, 33 for Grade 4]) ⫽ 257.09, 88.61, 143.32, and
Grades 2, 3, and 4, respectively. Children in Grade 1 read trans- 106.04; ␩s2 ⫽ .88, .73, .81, and .76 for Grades 1, 2, 3, and 4,
parent words as fast as nonwords (F1,2 ⬍ 1). Moreover, the respectively. As indicated in Table 3, transparency effects appear
examination of the effect sizes as well as the differences between to be larger for polygraphic words than for monographic words.
transparent word and nonword reading revealed that the magnitude Nonetheless, analyses of simple main effects revealed the presence
of lexicality effects increased across grades (␩s2 ⫽ .06, .45, .63, and of significant transparency effects, in all grades, for both mono-
.84, and ds ⫽ 34, 137, 174, and 253 ms for Grades 1, 2, 3, and 4, graphic and polygraphic words (see Table 4).
respectively). The greatest period of growth was between Grades In a subsequent analysis, we examined whether transparent and
1 and 2. opaque words were sensitive to word frequencies by computing
As shown in Table 2, opaque words were never read faster than the correlation between frequency judgments and mean latencies
nonwords in any grade, although the effect sizes reflecting the for each word. It is assumed that more frequent words are more
differences between opaque word and nonword reading decreased likely to have been read more frequently and, consequently, more
at each grade (␩s2 ⫽ .92, .67, .65, and .05, and ds ⫽ 1,216, 452, likely to have better defined orthographic representations. Consis-
271, and 6 ms for Grades 1, 2, 3, and 4, respectively). It is tent with this view, more frequent words were related to faster
important to note the remarkable decrease in differences across reading times for transparent words, r(26) ⫽ ⫺.54, as well as for
grades, such that by Grade 4, children read opaque words as fast as opaque words, r(26) ⫽ ⫺.56.
nonwords. Again, the greatest period of growth was between Error rate analyses. As shown in Tables 2 and 3, the pattern
Grades 1 and 2. of error rates in reading was similar to that of latencies. However,
The next series of analyses tested for the presence of transpar- we were unable to conduct ANOVAs to measure lexicality or
ency effects. Because these analyses excluded nonwords, the anal- transparency effect on errors because the error rates for transparent
yses acknowledged the fact that the real word stimuli were elab- word and nonword reading was very low. In fact, for transparent
orated by crossing transparency and polygraphy. Hence, a 2 words, 82% of children—and for nonwords, more than 70% of
(transparent vs. opaque words) ⫻ 2 (monographic vs. polygraphic children—read all stimuli correctly. In contrast, only 10% of
words) within-subject ANOVA was conducted for each grade. For children were able to read all opaque words correctly. Nonetheless,
ease of comparison, the results are reported as a function of grade. the pattern of error rates presented in Table 2 supports the strong

Table 3
Reading Mean Correct Latencies in Milliseconds (and Standard Deviation) and Mean Percentage of Errors (and Standard Deviation)
as a Function of Grade, Polygraphy, and Transparency

Latencies % Error

Monographic Polygraphic Monographic Polygraphic

Grade Transparent Opaque Transparent Opaque Transparent Opaque Transparent Opaque

1 1,910 (436) 2,776 (598) 1,786 (360) 3,332 (649) 2.48 (4.56) 16.33 (11.03) 0.82 (2.30) 28.77 (15.82)
2 1,496 (295) 1,875 (435) 1,414 (219) 2,216 (568) 1.22 (2.73) 13.67 (10.44) 1.01 (2.53) 15.71 (12.82)
3 1,252 (220) 1,644 (370) 1,227 (204) 1,728 (393) 0.61 (2.66) 9.59 (8.11) 0.20 (1.20) 15.51 (13.47)
4 1,063 (165) 1,313 (257) 1,053 (147) 1,319 (239) 0.00 (0.00) 3.15 (4.71) 0.63 (3.67) 5.46 (5.84)
READING AND SPELLING IN PERSIAN 7

Table 4
Transparent Versus Opaque One-Way Analysis of Variance Results by Subject (Item), Effect
Size, and 95% Confidence Intervals for Monographic and Polygraphic Words for Each Grade

Monographic Polygraphic

Grade F1(2) ␩2 95% CI F1(2) ␩2 95% CI

1 153.80 (37.80) .82 [724, 1,009] 390.12 (174.60) .92 [1,387, 1,705]
2 106.31 (2.24) .76 [304, 453] 130.41 (52.10) .77 [658, 944]
3 99.60 (19.07) .79 [309, 473] 115.55 (56.50) .77 [406, 596]
4 78.71 (22.96) .70 [192, 307] 77.53 (16.99) .71 [204, 327]

Note. Degrees of freedom for subject analyses are 1 and 34 for Grades 1–3; 1 and 33 for Grade 4; and 1 and
26 for item analyses. The subscript 1 refers to analyses by subject, and the subscript 2 refers to analyses by items.

transparency effects obtained for latency analyses. That is, chil- There was also a significant effect of stimuli, F1(2, 135) ⫽
dren’s error rates were substantially higher for opaque word read- 96.47, ␩2 ⫽ .43; F2(2, 91) ⫽ 24.87, with faster spelling speed for
ing (M ⫽ 14%) than for transparent word reading (M ⫽ 1%). In monographic words (M ⫽ 2,024) followed by nonwords (M ⫽
contrast, Persian polygraphy did not influence reading accuracy, as 2,099) and polygraphic words (M ⫽ 2,329). Moreover, there was
children read polygraphic words (M ⫽ 8%) almost as accurately as a significant Grade ⫻ Stimuli interaction, F1(6, 135) ⫽ 4.64, ␩2 ⫽
monographic words (M ⫽ 6%; see Table 3). .11; F2(6, 91) ⫽ 10.69. This expected interaction was further
Given the higher level of errors on opaque words, it was examined by testing for lexicality effects at each grade. Although
possible to examine statistically children’s progress across grades, spelling onsets for monographic words were slightly faster than
and this analysis revealed a significant grade effect, F1(3, 135) ⫽ those for nonwords at each grade, the differences were never
24.35, MSE ⫽ 79.68, as well as a significant linear trend across significant (all ps ⬎ .15). Moreover, children never spelled poly-
grades. Children read opaque words more accurately with each graphic words faster than nonwords (see Table 5). Contrary to the
passing grade, starting at 77% accuracy in Grade 1 to reach 95% reading results, these analyses failed to reveal any lexicality effects
accuracy in Grade 4. for spelling monographic or polygraphic words.
The next series of analyses tested for the presence of polygraphy
Spelling Performance effects. Because these analyses excluded nonwords, the analyses
Mean latencies, standard deviations, and error rates for spelling acknowledged the fact that the real-word stimuli were elaborated
polygraphic words, monographic words, and nonwords are pre- by crossing polygraphy and transparency. Hence, a 2 ⫻ 2 ANOVA
sented in Table 5. was conducted with polygraphy (monographic vs. polygraphic)
Latency analyses. To evaluate children’s spelling speed, we and transparency (transparent vs. opaque words) as within-subject
conducted a 4 (grade) ⫻ 3 (stimuli: polygraphic words, mono- factors. There was a strong polygraphy effect on children’s spell-
graphic words, and nonwords) mixed-factor ANOVA. There was a ing at each grade, that is, monographic words were always spelled
significant grade effect, with older children spelling faster than significantly faster than polygraphic words, Fs1(1, 34 [1, 33 for
younger children, F1(3, 135) ⫽ 29.71, ␩2 ⫽ .41; F2(3, 91) ⫽ Grade 4]) ⫽ 72.82, 154.88, 108.71, and 277.39; ␩s2 ⫽ .68, .82, .76,
235.19. Trend analyses across grades yielded significant linear and and .89; 95% CIs [314, 628; 218, 346; 161, 274; and 201, 279];
quadratic differences. The linear trend revealed that children’s Fs2(1, 52) ⫽ 62.94, 34.05, 28.13, and 28.44 for Grades 1, 2, 3, and
overall spelling speed improved over time, whereas the quadratic 4, respectively. However, the magnitude of the differences be-
trend was likely due to the fact that there was no difference in tween monographic and polygraphic words tended to decrease as
performance between Grades 3 and 4 (Ms ⫽ 2,499, 2,187, 1,976, the grades progressed, albeit not consistently (ds ⫽ 471, 282, 218,
and 1,974 ms for Grades 1, 2, 3, and 4, respectively). and 241 ms for Grades 1, 2, 3, and 4, respectively). The main effect

Table 5
Spelling Mean Correct Latencies in Milliseconds (and Standard Deviation) and Mean
Percentage of Errors (and Standard Deviation) for Nonwords, Monographic Words, and
Polygraphic Words as a Function of Grade

Latencies % Error

Grade Nonword Monographic Polygraphic Nonword Monographic Polygraphic

1 2,381 (391) 2,312 (293) 2,783 (511) 4.44 (3.92) 2.11 (3.53) 36.76 (9.16)
2 2,147 (299) 2,066 (246) 2,348 (326) 2.43 (2.50) 2.86 (3.25) 27.33 (11.80)
3 1,915 (268) 1,900 (236) 2,118 (273) 0.93 (1.90) 1.71 (2.60) 23.24 (11.27)
4 1,939 (268) 1,870 (200) 2,111 (231) 1.01 (1.93) 0.39 (1.10) 11.86 (8.05)
8 RAHBARI AND SÉNÉCHAL

of transparency on children’s spelling was significant for Grade 1 evidence of the differential difficulty for reading and spelling
and Grade 2 only, that is, children in Grades 1 and 2 spelled posed by the characteristics of Persian. As expected, the words that
transparent words faster than opaque words, F1(1, 34) ⫽ 34.65, children most often read correctly but spelled incorrectly were
␩2 ⫽ .35, 95% CI [193, 310]; F2(1, 52) ⫽ 23.80; and F1(1, 34) ⫽ polygraphic (80%, 83%, 83% and 82% of errors for Grades 1, 2,
7.52, p ⫽ .01, ␩2 ⫽ .16, 95% CI [178, 283]; F2(1, 52) ⫽ 28.69 for 3, and 4, respectively), and the words that children most often
Grades 1 and 2, respectively. The Polygraphy ⫻ Transparency spelled correctly but read incorrectly were opaque–monographic
interaction was not significant at any grade level. (80%, 75%, 81%, and 86% of errors for Grades 1, 2, 3, and 4,
The examination of the correlation coefficients between word respectively). Moreover, the correlations between reading and
frequency and spelling latencies revealed a different pattern across spelling transparent–polygraphic words, r(137) ⫽ .07, p ⬎ .05,
word type. Word frequency was associated with spelling poly- and opaque–monographic words, r(137) ⫽ .10, p ⬎ .05, were not
graphic words, r(26) ⫽ ⫺.46, p ⫽ .03, but not with spelling significant. Taken together, these results suggest that the opacity of
monographic words, r(26) ⫽ ⫺.04, p ⫽ .67, despite similar words typically affected reading, not spelling, and that the polyg-
distributions of latencies and frequencies across word type. These raphy of words typically affected spelling, not reading. Further
findings support the notion that children must retrieve some or- support comes from the association between reading and spelling
thographic information to select the correct letter among an array the easiest categories of words as well as the hardest categories of
of plausible choices when spelling polygraphic words. words. There was a significant correlation between reading and
Error rate analyses. The pattern of error rates was consistent spelling easy transparent–monographic words, r(137) ⫽ .25, p ⬍
with that obtained for latencies. As indicated in Table 5, children’s .05, as well as reading and spelling hard opaque–polygraphic
error rates were substantially higher for spelling polygraphic words, r(137) ⫽ .48, p ⬍ .01. These correlational results, however,
words (M ⫽ 25%) compared with monographic words (M ⫽ 2%). should be interpreted with caution given the high accuracy rates
In fact, 80% of children made no error for monographic words, for some words.
whereas only 3% of children spelled polygraphic words entirely
correctly. A one-way ANOVA on polygraphic words yielded a
significant grade effect, F1(3, 135) ⫽ 37.78. Follow-up analyses Discussion
revealed a significant linear trend showing that children’s spelling
accuracy improved with schooling. Further, examination of the In the present study, we investigated the developmental change
polygraphic words in Table 6 revealed that opaque polygraphic in Persian children’s reading and spelling across the first four years
words were particularly difficult for children. of instruction. We found the children became more efficient read-
Finally, examination of the types of spelling errors across grades ers and spellers across grades but that the greatest growth occurred
revealed that almost all of children’s spelling errors were substi- between Grades 1 and 2. Growth, however, was affected by the
tution errors. For polygraphic words, the misspelled words were orthographic characteristics of Persian. For reading, lexicalization
phonologically correct and sounded like the target words, but was present starting in Grade 2 for words that were transparent, but
children selected a wrong letter for a polygraphic phoneme that lexicalization was not yet present in Grade 4 for opaque words. For
resulted in a pseudohomophone. For the few spelling errors on spelling, no lexicality effects were found for either monographic or
monographic words, children generally substituted a grapheme polygraphic words in any grade. As expected, there were strong
that had a similar sound to the target phoneme. transparency effects for reading, and there were strong polygraphy
effects for spelling. Although the size of the transparency effects
decreased for reading across grades, polygraphy effects for spell-
Supplemental Error Analyses
ing did not decrease systematically. Most interesting was the
Examination of children’s reading and spelling errors revealed finding that children in Grades 1 and 2 were the most sensitive to
that although there were words that children could read and spell interference of word complexities across literacy tasks. Specifi-
correctly or those that they were not able either to read or spell, cally, polygraphy affected reading performance, and transparency
there were words that children were able to read but could not spell affected spelling performance for these children, but these effects
and those that they were able to spell but could not read. It is those were no longer present by Grade 3. Each of these findings is
two latter types of words that we examined to provide converging discussed in turn.

Table 6
Spelling Mean Correct Latencies in Milliseconds (and Standard Deviation) and Mean Percentage of Errors (and Standard Deviation)
as a Function of Grade, Polygraphy, and Transparency

Latencies % Error

Monographic Polygraphic Monographic Polygraphic

Grade Transparent Opaque Transparent Opaque Transparent Opaque Transparent Opaque

1 2,189 (339) 2,434 (282) 2,618 (468) 3,072 (614) 1.71 (3.37) 2.48 (4.30) 17.33 (9.46) 56.19 (14.03)
2 1,943 (235) 2,193 (286) 2,234 (328) 2,516 (396) 1.14 (2.55) 4.57 (5.55) 10.28 (9.88) 44.38 (17.93)
3 1,795 (255) 2,009 (254) 1,965 (258) 2,331 (348) 0.19 (1.13) 3.24 (3.21) 9.71 (8.76) 36.76 (16.75)
4 1,745 (212) 1,994 (204) 1,948 (202) 2,292 (304) 0.20 (1.14) 0.59 (1.92) 5.30 (5.38) 18.43 (12.61)
READING AND SPELLING IN PERSIAN 9

The findings in the present study show the gradual progression ms in Grade 1, and it dropped substantially to 259 ms in Grade 4.
toward more efficient literacy skills across Grades 1– 4, with the The reported difference in Persian adult readers was 100 ms
greatest period of growth between Grades 1 and 2. Specifically, the (Rahbari & Sénéchal, 2009). These findings are consistent with the
averaged difference in word reading speed across Grades 2, 3, and evidence from studies in other languages showing that inconsis-
4 was 281 ms, but it was 690 ms between Grades 1 and 2. tencies/irregularities have a larger effect on young children’s as
Similarly, the average difference in spelling speed across Grades 2, compared with older children’s reading performance (e.g., Alegria
3, and 4 was 108 ms, but it was 341 ms between Grades 1 and 2. & Mousty, 1996; Weekes, Castles, & Davies, 2006).
Although Persian is generally viewed as a transparent orthography The decrease over time in the advantage of nonwords over
(Baluch, 2006), it is interesting that the pattern of greatest growth opaque words coupled with the decrease over time in the magni-
is similar to that found in longitudinal research in English (Juel, tude of the transparency effect suggest that children are gradually
1988) as well as French (Sprenger-Charolles, Siegel, Bechennec, building orthographic representations for reading opaque words.
& Serniclaes, 2003), Spanish (Defior et al., 2002), and Dutch This interpretation was further supported by a moderately strong
(Aarnoutse, Van Leeuwe, Voeten, & Oud, 2001). The key differ- association between opaque-word reading latencies and word fre-
ence is that results in Persian pertain to latencies, and those for the quency, r(26) ⫽ ⫺.56.
above-mentioned languages pertain to accuracy. In the present research, opaque and transparent words were
In the present study, we found a differential growth pattern for matched in the number of letters such that the orthographic struc-
lexicality effects across word types. First-grade children read ture for transparent words was CVC, and it was CCC for opaque
transparent Persian words as fast as transparent nonwords, but words. As a consequence, the opaque words had a coda structure
from Grade 2 onward, children read transparent words signifi- that was more complex, even though the words were matched on
cantly faster than transparent nonwords. Therefore, the availability initial consonant and duration (i.e., naming latencies were con-
of one-to-one correspondences for transparent words helped Per- trolled across word types). Given that syllabic structure is known
sian children to learn to read these words quickly, which, in turn, to affect reading acquisition in other languages (e.g., Sprenger-
facilitated lexicalization. In addition, academic experience also led Charolles & Siegel, 1997), it raises the possibility that the differ-
to more efficient use of orthographic representations, as the size of ence between transparent and opaque words might be due to
lexicality effects increased with each passing grade. The acquisi- differences in coda complexity. Reading performance differences
tion of efficient orthographic representations seemed more diffi- between opaque and transparent words cannot, however, be sub-
cult when there was not a one-to-one relation between letters and sumed entirely to syllabic complexity because of the well-
sounds, as was the case for opaque words. Children never read established findings that reading the same words with or without
opaque words faster than nonwords. Nonetheless, the differences diacritics to mark vowels still produces transparency effects in
between opaque word and nonword reading decreased systemati- children and adult readers (Abu-Rabia & Taha, 2006; Ravid &
cally across the grades, such that fourth-grade children read Shlesinger, 2001). Moreover, adult readers of Persian read opaque
opaque words as fast as nonwords. Most important, the presence of word faster than nonwords with a CVC structure, yet they read
a lexicality effect for opaque words in young adults (Rahbari & opaque words slower than transparent words (Rahbari & Sénéchal,
Sénéchal, 2009) and the absence of one for the fourth-grade 2009). The issue of coda complexity, nonetheless, should be in-
children in the present study suggest that opaque word reading in vestigated further in future research.
Persian reaches lexicalization after Grade 4. The spelling results show similarities and differences with those
In the present research, children read transparent words faster for reading. Similar to the reading results, children in higher grades
and more accurately than words in which the orthographic repre- spelled both words and nonwords faster and more accurately than
sentation excluded a vocalic phoneme (i.e., opaque words). The did children in lower grades. These results are consistent with the
three-letter words were matched on number of letters, consonant findings from many studies showing a significant effect of grade
onset, frequency, and duration. As such, the present results add, levels in children’s spelling (e.g., Laxon, Coltheart, & Keating,
under more controlled conditions, to the limited body of research 1988; Weekes et al., 2006). Contrary to the reading results, how-
on Persian reading in children (Arab-Moghaddam & Sénéchal, ever, there was no lexicality effect either for monographic or for
2001; Baluch & Danaye-Tousi, 2006; Baluch & Shahidi, 1991; polygraphic words. In particular, the lack of a systematic incre-
Rahbari et al., 2007) and adults (Baluch & Besner, 1991; Rahbari ment in the difference between word and nonword spelling across
& Sénéchal, 2009). The present findings also extend, under more grades suggests that children used the same phonological pro-
natural reading conditions, previous research in Arabic (Abu- cesses to spell both words and nonwords in all grades. The results
Rabia & Taha, 2006) and Hebrew (Ravid & Shlesinger, 2001) that are consistent with studies showing that phonological processing
shows that adult and children readers are slower and more error plays an important role in spelling consistent orthographic systems
prone when reading words with vowels omitted as compared with (Porpodas, 2001; Wimmer & Hummer, 1990). However, the pat-
reading the same words with diacritic marks for the vowel pho- tern of findings was more complex. The examination of the cor-
neme. relation coefficients between word frequency and word type
The novelty of the current findings, however, is to show the showed a different pattern across word types. The finding that
developmental progression over the first four years of schooling. spelling monographic words was not sensitive to word frequencies,
We found that the magnitude of the transparency effect decreased r(26) ⫽ ⫺.04, is consistent with the interpretation that phonolog-
systematically with reading experience. That is, the differences ical processing is involved in spelling. However, the significant
between transparent and opaque words observed for younger chil- correlation between polygraphic word spelling and word fre-
dren tended to be smaller in older children’s reading. For instance, quency, r(26) ⫽ ⫺.46, suggests that phonological processing alone
the difference between transparent and opaque words was 1,182 does not produce accurate spellings of words with inconsistent
10 RAHBARI AND SÉNÉCHAL

phoneme– grapheme correspondences. To spell inconsistent graph- gesting the involvement of phonological processing. At the same
emes correctly, children use other sources of information, such as time, there was a significant correlation coefficient between word
their orthographic knowledge acquired from exposure to words. frequency and reading opaque words as well as between word
Similar results were obtained for young adult spellers (Rahbari & frequency and spelling polygraphic words, suggesting the involve-
Sénéchal, 2009). This is also supported by evidence from Rahbari ment of orthographic processing for these words. If one accepts
et al. (2007), who reported that the unique contribution of ortho- that orthographic processes are also involved in opaque-word
graphic skills was almost twice as high when spelling polygraphic reading and polygraphic-word spelling, then one has to explain
(7%) than monographic (3%) words. Moreover, the unique con- why opaque words are more difficult to read and why polygraphic
tribution of phonological skills was three times smaller when words are more difficult to spell than words without mismatches
spelling polygraphic words (8%) as compared with monographic between phonology and orthography. Rahbari and Sénéchal (2009)
words (24%). proposed that these differences might be related to a selection
The significant polygraphy effect obtained is in accord with mechanism for opaque word reading in Persian. Recall that Persian
evidence from cross-orthographic as well as within-orthographic opaque words appear as a sequence of consonant letters, and,
studies showing that spelling consistent words/orthographies is consequently, a Persian reader has to select a correct vowel among
easier than spelling inconsistent words/orthographies (e.g., Cara- three plausible vowels. This selection mechanism might slow
volas & Bruck, 1993; Juul, 2005; Wimmer & Landerl, 1997). down the reading of opaque words. Rahbari and Sénéchal found
However, the magnitude of the polygraphy effect for latencies did some support for this interpretation in that Persian adults read
not decrease dramatically across Grades 2, 3, and 4 (ds ⫽ 282, opaque words faster than nonwords (i.e., a lexicality effect), but
218, and 241 ms, respectively). Findings in other orthographies opaque words were still read significantly slower than transparent
demonstrate that spelling consistency effects based on accuracy do words. The present findings suggest that a similar selection mech-
decrease with academic experience in French-speaking (Alegria & anism might be at play for spelling. That is, when spelling poly-
Mousty, 1996) or English-speaking (e.g., Cassar & Treiman, 1997; graphic words, the speller has to choose a letter among two or
Laxon et al., 1988; Weekes et al., 2006) children. The current more plausible alternatives for the same phoneme, and this selec-
findings raise the interesting possibility of a relatively stable effect tion might take time as compared with when a one-to-one corre-
on latencies that is not as sensitive to spelling experience. The spondence exists between sounds and letters.
possibility of a stable effect is also supported by a comparison with The third theoretical contribution concerns the vulnerability of
adult spellers. Rahbari and Sénéchal (2009) reported a 154-ms the developing orthographic representations during acquisition.
advantage for monographic words over polygraphic words. In the The present findings clearly show that both a mismatch between
present research, however, children read and spelled very short letter–sound correspondences as well as a mismatch between
words. It is also possible that when the length of a word is short, sound–letter correspondences affected both reading and spelling
such as the one-syllable words in the present study, a speller may during the initial phases of acquisition. That is, children in Grades
be able to quickly select appropriate graphemes for the phonemes 1 and 2 had more difficulty reading and spelling words that
kept in working memory. With longer words, however, it may be contained sound-to-letter mismatches as well as those that con-
more efficient to retrieve orthographic representations. These in- tained letter-to-sound mismatches. The findings for Grade 2 are
teresting possibilities could be examined further in future research. particularly interesting because these children had high accuracy
The reading and spelling results obtained in the present research rates and had reached lexicalization for reading transparent words
contribute in three ways to our understanding of the development (i.e., reading transparent words faster than nonwords). These find-
of accurate and well-defined orthographic representations. First, ings for Grade 2 suggest that although children might seem to have
the use of latencies revealed that reading and spelling fluency was well-defined orthographic representations in place, the influence of
still progressing in Grades 3 and 4. In fact, fourth-grade students spelling characteristics on reading and vice versa suggests that
still had not reached adult reading and spelling speed for these their orthographic representations are still vulnerable to interfer-
familiar three-letter words. This raises the interesting theoretical ence. This cross-over effect was no longer present by Grade 3.
question of determining when literacy reaches asymptotic levels. There are limitations in the present study that are noteworthy.
Actual models of word reading acquisition based on reading ac- First, in the present research, we did not examine how semantic
curacy seldom consider word reading growth beyond the early processes interact with phonological and orthographic processes.
elementary school years, but the present findings suggest the need One can certainly imagine the importance of semantic processes in
to document the development of word reading and spelling beyond reading opaque words that are homographic (as in the example of
Grade 4. bd that can represent bid, bed, bud). The present findings are
The second contribution concerns the difficulty in building therefore limited to opaque words that have a single pronunciation
well-defined orthographic representations that are redundant pho- and a single orthographic representation. Second, we did not
nologically when there are mismatches between the phonology and examine a potentially important feature of the Persian orthographic
orthography of a language. In the present research, the mismatch system. Persian, unlike English, is written with a mixture of
for reading concerned vocalic phonemes that were not represented cursive and disjointed letters, that is, some of the words in Persian
by a grapheme (i.e., opaque words) and the mismatch for spelling have a mandatory spelling in which all or some of the letters have
concerned consonantal phonemes that could be represented by to be written in a joined fashion, whereas other words or letter
more than one grapheme (i.e., polygraphic words). Although quite combinations have to be written in a disjointed fashion. This
different, these mismatches slowed the reading and spelling pro- characteristic of Persian writing, which is independent of polyg-
cess. That is, there were reliable transparency effects for reading as raphy effects, may also affect spelling latencies, in particular if one
well as stable consistency effects for spelling at each grade, sug- measures the time to spell entire words, and warrants further
READING AND SPELLING IN PERSIAN 11

examination. In the present study, however, spelling latencies were impact on dyslexic and unimpaired children’s memory of words. Annals
based on onsets, not whole words. Third, a more sensitive assess- of Dyslexia, 56, 319 –334.
ment of latencies might reveal small, but significant, differences Baluch, B., & Shahidi, S. (1991). Visual word recognition in beginning
between spelling latencies for monographic words and nonwords. readers of Persian. Perceptual and Motor Skills, 72, 1327–1331.
In conclusion, the current findings that experience and word Bowey, J. A., & Muller, D. (2005). Phonological recoding and rapid
characteristics influence children’s reading and spelling develop- orthographic learning in third-graders’ silent reading: A critical test of
the self-teaching hypothesis. Journal of Experimental Child Psychology,
ment qualify the view that literacy skills are acquired as a sequence
92, 203–219.
of stages (Frith, 1985; Gough & Juel, 1991). In fact, the current
Bruck, M., Genesee, F., & Caravolas, M. (1997). A cross-linguistic study
pattern of findings for reading and spelling are in accord with an of early literacy acquisition. In B. A. Blachman (Ed.), Foundations of
alternative view of development whereby gradual quantitative reading acquisition and dyslexia: Implications for early intervention
changes in proficiency occur as children become increasingly (pp. 145–162). Mahwah, NJ: Erlbaum.
sensitive to the internal structure of written words as well as to the Caravolas, M., & Bruck, M. (1993). The effect of oral and written language
correspondences between letters or groups of letters in words and input on children’s phonological awareness: A cross-linguistic study.
their pronunciation (Perfetti, 1992; Rayner et al., 2001; Share, Journal of Experimental Child Psychology, 55, 1–30.
1999). Moreover, the obtained differential pattern of results for Caravolas, M., Kessler, B., Hulme, C., & Snowling, M. (2005). Effects of
reading and spelling highlights the importance of studying their orthographic consistency, frequency, and letter knowledge on children’s
development concurrently (Hanley & Kay, 1992; Weekes & Colt- vowel spelling development. Journal of Experimental Child Psychology,
heart, 1996). Finally, and most interesting from a developmental 92, 307–321.
perspective, is the finding that both reading transparency and Cassar, M., & Treiman, R. (1997). The beginnings of orthographic knowl-
spelling polygraphy affected reading as well as spelling in Grades edge: Children’s knowledge of double letter in words. Journal of Ex-
perimental Psychology, 89, 631– 644.
1 and 2, but the word characteristics had differential effects in
Cunningham, A. E., Perry, K. E., Stanovich, K. E., & Share, D. L. (2002).
Grades 3 and 4. Specifically, in Grades 3 and 4, transparency
Orthographic learning during reading: Examining the role of self-
effects were limited to reading, and polygraphy effects were lim-
teaching. Journal of Experimental Child Psychology, 82, 185–199.
ited to spelling. This pattern highlights the vulnerability of chil- Defior, S., Martos, F., & Cary, L. (2002). Differences in reading acquisi-
dren’s representations and processes during the early phases of tion development in two shallow orthographies: Portuguese and Spanish.
acquisition as well as the rapidity with which representations (and Applied Psycholinguistics, 23, 135–148.
processes) may become specialized as a function of the literacy Ehri, L. C. (1992). Reconceptualizing the development of sight word
task at hand. reading and its relationship to recoding. In P. B. Gough, L. C. Ehri, &
R. Treiman (Eds.), Reading acquisition (pp. 105–143). Hillsdale, NJ:
References Erlbaum.
Frith, U. (1985). Beneath the surface of developmental dyslexia. In K. E.
Aarnoutse, C., Van Leeuwe, J., Voeten, M., & Oud, H. (2001). Develop- Patterson, J. C. Marshall, & M. Coltheart (Eds.), Surface dyslexia:
ment of decoding, reading comprehension, vocabulary, and spelling Neuropsychological and cognitive studies of phonological reading (pp.
during the elementary school years. Reading and Writing: An Interdis- 301–330). London, England: Erlbaum.
ciplinary Journal, 14, 61– 89. Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recog-
Abu-Rabia, S., & Taha, H. (2006). Reading in Arabic orthography: Char- nition and orthographical depth: A multilingual comparison. Journal of
acteristic, research findings, and assessment. In R. M. Joshi & P. G. Experimental Psychology: Human Perception and Performance, 13,
Aaron (Eds.), Handbook of orthography and literacy (pp. 365–376). 104 –115.
Mahwah, NJ: Erlbaum. Gholamain, M., & Geva, E. (1999). Orthographic and cognitive factors in
Alegria, J., & Mousty, P. (1996). The development of spelling procedures concurrent development of basic reading skills in English and Persian.
in French-speaking, normal, and learning disabled children: Effects of
Language Learning, 49, 183–217.
frequency and lexicality. Journal of Experimental Child Psychology, 63,
Goswami, U., Ziegler, J. C., Dalton, L., & Schneider, W. (2003). Nonword
312–338.
reading across orthographies: How flexible is the choice of reading
Arab-Moghaddam, N., & Sénéchal, M. (2001). Orthographic processing
units? Applied Psycholinguistics, 24, 235–247.
skills in reading and spelling in Persian/English bilinguals. International
Gough, P. B., & Juel, C. (1991). The first stages of word recognition. In L.
Journal of Behavioral Development, 25, 140 –147.
Rieben & C. Perfetti (Eds.), Learning to read: Basic research and its
Aro, M., & Wimmer, H. (2003). Learning to read: English in comparison
implications (pp. 47–56). Hillsdale, NJ: Erlbaum.
to six more regular orthographies. Applied Psycholinguistics, 24, 619 –
634. Hanley, J. R., & Kay, J. (1992). Does letter-by-letter reading involve the
Baluch, B. (1993). Lexical decision in Persian: A test of the orthographic spelling system? Neuropsychologia, 30, 237–256.
depth hypothesis. International Journal of Psychology, 28, 19 –29. Juel, C. (1988). Learning to read and write: A longitudinal study of 54
Baluch, B. (2006). Persian orthography and its relation to literacy. In R. M. children from first through fourth grades. Journal of Educational Psy-
Joshi & P. G. Aaron (Eds.), Handbook of orthography and literacy (pp. chology, 80, 4, 437– 447.
365–376). Mahwah, NJ: Erlbaum. Juul, H. (2005). Knowledge of context sensitive spellings as a component
Baluch, B., & Besner, D. (1991). Visual word recognition: Evidence for of spelling competence: Evidence from Danish. Applied Psycholinguis-
strategic control of lexical and nonlexical routines in oral reading. tics, 26, 249 –265.
Journal of Experimental Psychology: Learning, Memory, and Cogni- Kerek, E., & Niemi, P. (2009). Learning to read in Russian: Effects of
tion, 17, 644 – 652. orthographic complexity. Journal of Research in Reading, 32, 157–179.
Baluch, B., & Besner, D. (2001). Basic process in reading: Semantics Kinoshita, S., Lupker, S. J., & Rastle, K. (2004). Modulation of regularity
affects speeded naming of high-frequency words in an alphabetic script. and lexicality effects in reading aloud. Memory & Cognition, 32, 1255–
Canadian Journal of Experimental Psychology, 55, 63– 69. 1264.
Baluch, B., & Danaye-Tousi, M. (2006). Spelling transparency and its Kyte, C. S., & Johnson, C. J. (2006). The role of phonological recoding in
12 RAHBARI AND SÉNÉCHAL

orthographic learning. Journal of Experimental Child Psychology, 93, M. S. (2001). How psychological science informs the teaching of read-
166 –185. ing. Psychological Science in the Public Interest, 2, 31–74.
Laxon, V. J., Coltheart, V., & Keating, C. (1988). Children find friendly Seymour, P. H. K., Aro, M., & Erskine, J. M. (2003). Foundation literacy
words friendly too: Words with many orthographic neighbours are easier acquisition in European orthographies. British Journal of Psychology,
to read and spell. British Journal of Educational Psychology, 58, 103– 94, 143–174.
119. Share, D. L. (1999). Phonological recoding and orthographic learning: A
Nassaji, H., & Geva, E. (1999). The contribution of phonological and direct test of self-teaching hypothesis. Journal of Experimental Child
orthographic processing skills to adult ESL reading: Evidence from Psychology, 29, 294 –305.
native speakers of Farsi. Applied Psycholinguistics, 20, 241–267. Sprenger-Charolles, L., & Siegel, L. S. (1997). A longitudinal study of the
Perfetti, C. A. (1992). The representation problem in reading acquisition. In effects of syllabic structure on the development of reading and spelling
P. B. Gough, L. C. Ehri, & R. Treiman (Eds.), Reading acquisition (pp. skills in French. Applied Psycholinguistic, 18, 485–505.
145–174). Hillsdale, NJ: Erlbaum. Sprenger-Charolles, L., Siegel, L. S., Bechennec, D., & Serniclaes, W.
Perfetti, C. A., & Hart, L. (2001). The lexical basis of comprehension skill. (2003). Development of phonological and orthographic processing in
In D. S. Gorfien (Ed.), On the consequences of meaning selection: reading aloud, in silent reading, and in spelling: A four-year longitudinal
Perspectives on resolving lexical ambiguity (pp. 67– 86). Washington, study. Journal of Experimental Child Psychology, 84, 194 –217.
DC: American Psychological Association. Weekes, B. S., Castles, A. E., & Davies, R. A. (2006). Effects of consis-
Pinheiro, A. M. V. (1995). Reading and spelling development in Brazilian tency and age of acquisition on reading and spelling among developing
Portuguese. Reading and Writing: An Interdisciplinary Journal, 7, 111– readers. Reading and Writing: An Interdisciplinary Journal, 19, 133–
138. 169.
Porpodas, C. (2001). Cognitive processes in first grade reading and spelling Weekes, B., & Coltheart, M. (1996). Surface dyslexia and surface dys-
of Greek. Psychology: The Journal of the Hellenic Psychology Society, graphia: Treatment studies and their theoretical implications. Cognitive
8, 384 – 400. Neuropsychology, 13, 277–315.
Rahbari, N., & Sénéchal, M. (2009). Lexical and nonlexical processes in Wimmer, H., & Hummer, P. (1990). How German-speaking first graders
the skilled reading and spelling of Persian. Reading and Writing: An read and spell: Doubts on the importance of the Logographic stage.
Interdisciplinary Journal, 22, 511–530. Applied Psycholinguistics, 11, 349 –368.
Rahbari, N., Sénéchal, M., & Arab-Moghaddam, N. (2007). The role of Wimmer, H., & Landerl, K. (1997). How learning to spell German differs
visual and phonological processing skills in reading and spelling of from learning to spell English. In C. A. Perfetti, L. Rieben, & M. Fayol
monolingual Persian children. Reading and Writing: An Interdiscipli- (Eds.), Learning to spell: Research, theory, and practice across lan-
nary Journal, 20, 511–533. guages (pp. 295–318). Mahwah, NJ: Erlbaum.
Raman, I., Baluch, B., & Besner, D. (2004). On the control of visual word Ziegler, J. C., Ferrand, L., & Montant, M. (2004). Visual phonology: The
recognition: Changing routes versus changing deadlines. Memory & effects of orthographic consistency on different auditory word recogni-
Cognition, 32, 489 –500. tion tasks. Memory & Cognition, 32, 732–741.
Ravid, D., & Shlesinger, Y. (2001). Vowel reduction in Modern Hebrew: Ziegler, J. C., & Goswami, U. (2005). Reading acquisition, developmental
Traces of the past and current variation. Folia Linguistica, 35, 371–397. dyslexia, and skilled reading across languages: A psycholinguistic grain
Rayner, K., Foorman, B. R., Perfetti, C. A., Pesetsky, D., & Seidenberg, size theory. Psychological Bulletin, 131, 3–29.

(Appendices follow)
READING AND SPELLING IN PERSIAN 13

Appendix A

List of Real Words (58) Used to Test Persian Children’s Reading and Spelling With Frequencies
and Duration (in Milliseconds) in Parentheses

Transparent Opaque
Type of words Persian spelling Pronunciation Meaning Persian spelling Pronunciation Meaning

(2.20,603) dig pot (4.46,692) kshf/kashf discovery


(3.15,829) rig gravel (3.85,658) grg/gorg wolf
(3.35,795) pich screw (2.63,759) knj/kong corner
(3.35,707) kham raw (3.01,739) shrm/sharm shame
(3.01,624) bim fear (3.20,679) plk/pelk eyelid
(5.85,713) pak clean (4.02,642) pshm/pashm wool
(4.45,663) pir old (4.10,768) mrg/marg death
Non-polygraphic (3.36,663) mikh nail (3.95,745) khrj/kharj expense
(3.65,645) kakh palace (3.15,827) kshk/kashk dried whey
(4.55,621) khun blood (3.75,666) pkhsh/pakhsh broadcast
(2.45,590) gij confused (3.82,686) gnj/ganj treasure
(2.80,607) gur grave (2.50,661) kbk/kabk partridge
(3.62,742) kuch migration (3.23,789) chrb/charb oily
(3.75,787) shur salty (4.75,771) jshn/jashn festival

(3.55,665) sim wire (3.25,710) ghrg/gharg sink


(2.82,616) gif funnel (3.45,884) sgf/sagf ceiling
(3.75,718) naz cute (2.93,570) gtb/gotb pole
(4.73,678) raz secret (2.01,732) sbt/sabt registration
(3.01,715) zag crow (2.62,811) shl/sahl easy
(3.45,883) saf smooth (4.25,802) slh/solh peace
(3.95,869) sud benefit (3.15,780) zhn/zehn mind
polygraphic (3.75,803) sut whistle (3.92,786) shrt/shart condition
(3.80,710) chag fat (3.01,766) zbt/zabt record
(5.64,646) bag garden (3.62,715) chtr/chatr umbrella
(2.60,779) tag crown (3.15,710) rbt/rabt connection
(3.54,697) dag hot (2.95,585) ngs/nags deficiency
(3.00,610) gaz gas (3.50,637) grz/garz debt
(2.45,810) tig thorn (3.92,700) gsd/gasd purpose

(Appendices continue)
14 RAHBARI AND SÉNÉCHAL

Appendix B

List of Nonwords (38) Used to Test Phonological Processing

Non-words Non-words

Persian spelling Pronunciation Persian spelling Pronunciation

rib bukh

mig Makh

luj shar

lab fid

razh nal

rab dizh

khim mazh

paj jab

chud shil

dak cham

zhir vash

juj lud

jad rim

nukh daf

dil zhid

lad lakh

khazh rif

rak zhar

dikh zhuk

Received May 21, 2009


Revision received May 19, 2010
Accepted May 25, 2010 䡲

View publication stats

You might also like