Content uploaded by Robert Deacon
Author content
All content in this area was uploaded by Robert Deacon on Aug 01, 2018
Content may be subject to copyright.
Available online at www.ejal.eu
Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
EJAL
Eurasian Journal of
Applied Linguistics
The Causes of English Spelling Errors by Arabic
Learners of English
Robert Joel Deacon a
*
a Nagoya University, Furo-cho, Chikusa-ku, Nagoya 464-8601, Japan
Received 11 August 2016
Received in revised form 11 February 2017
Accepted 7 April 2017
Abstract
This study investigates the possible cause(s) of English spelling errors by Arabic learners of English
(ALEs). Studies show that ALEs make significantly more English spelling errors than other English
second-language learner groups. Studies also show ALEs make more errors with vowels. The omission of
short vowels in Arabic writing has been proposed to cause vowel blindness in English, resulting in the
poorer spelling performance. This study evaluates this claim by comparing the distribution of short and
long-vowel errors and vowel and consonant error types from handwritten texts by ALEs. While this study
found more vowel than consonant errors, only the distribution of vowel graph-choice and insertion errors
significantly differed from the number of consonant errors by subcategory. Graph-choice errors, not
omission errors, were exceedingly the most common error type. Vowel length was not significantly
associated with either vowel omission or graph-choice as expected under the vowel blindness hypothesis.
The results, thus, did not indicate a missing vowel orthographic transfer effect as the primary reason for
ALE orthographic production difficulty in English. Instead, this paper proposes an underdeveloped
lexical-orthographic-representation hypothesis to account for both the degree and range of errors found.
This study also found that low and high proficiency groups only significantly differed in consonant graph-
choice and silent-graph error categories, with the advanced group performing better. These results
suggest that ALE spelling skills are not markedly improving with the advancement of other writing
skills and that ALEs may need explicit spelling instruction, especially to connect vowel phonemes with
multiple graphemes.
© 2017 EJAL & the Authors. Published by Eurasian Journal of Applied Linguistics (EJAL). This is an open-access
article distributed under the terms and conditions of the Creative Commons Attribution license (CC BY-NC-ND)
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Keywords: Arabic ESL; orthographic competence; orthographic transfer; spelling; vowel blindness
1. Introduction
1.1. Arabic orthographic difficulty in English
Orthographic difficulty by Arabic Learners of English (ALEs) is a topic of much
discussion. ALEs reportedly have messier handwriting and poorer spelling skills than
several other groups studying English as a second language (ESL) (Thompson-Panos
& Thomas-Ruzic, 1983). Studies have found that ALEs perform significantly worse
than other ESL groups on tests measuring spelling skill in terms of accurate graph
*
Robert Deacon
E-mail address: deacon.r@ilas.nagoya-u.ac.jp
http://dx.doi.org/............................................
2 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
recognition/attention (Hayes-Harb, 2006; Ryan & Meara, 1991) and production
(Dunlap, 2012; Fender, 2008). Given this problem, several studies have investigated
ALE spelling error types (Bowen, 2011; Dunlap, 2012; Haggan, 1991). Studies have
also tested where spelling errors are more likely to occur (Fender, 2008) and be
recognized (Saigh & Schmitt, 2012). When these studies have separated vowel errors
from consonant ones, the results show that ALEs make more vowel errors.
Dunlap (2012) had 88 ESL participants, who spoke either Arabic, Spanish, Korean,
or Chinese as their first language (L1), record an oral response from a computer
prompted question and then transcribe their recorded message. Spelling errors were
then categorized and counted from the transcriptions for each language group. The
results showed that ALEs created more total errors than the other groups and more
vowel errors than consonants.
†
Haggan‟s study (1991) tallied and compared spelling
errors made by 1st and 4th year Arabic English majors on their end-of-the-semester
handwritten examinations.
‡
The results showed that selecting an incorrect vowel
graph (i.e., choosing the wrong letter for vowel graphs) to be a common problem (177
cases out of 405 total errors). Bowen (2011) surveyed ALE teachers to create an error
database and found that 89% of the vowel letters as compared to 43% of the consonant
letters were incorrect from 250 randomly selected misspelled words. Of these vowel
errors, right-vowel wrong-place (i.e., vowel transposition or misordering errors) were
more common than addition, deletion, or other vowel type errors (2011, p. 92). Haggan
(1991), conversely, found few letter misordering errors for either vowel or consonants:
only 12 cases, which was less than .03%. Instead, choosing the wrong letter for vowel
graphs mapping to the vowel schwa (64 cases) and incorrectly selecting vowel graphs
for other vowel phonemes (113 cases) were disproportional, frequent errors. It is
difficult, however, to compare the results of these studies directly because each study
categorized spelling errors differently. Nevertheless, these studies suggest that ALE
orthographic difficulty centers on vowels.
To explain some of the orthographic difficulty that ALEs exhibit, Thompson-Panos
and Thomas-Ruzic (1983) suggested that the omission of short vowels in the Arabic
writing system results in the omission of vowels in English writing. Ryan and Meara
(1991) coined the term vowel blindness to likewise describe why Arabic students were
less likely to notice words with missing vowels in their study. Adopting this
hypothesis, Hayes-Harb (2006, p. 335) concluded that the results of her study indicate
that Arabic speakers attempt to visually process words in English much as they do in
Arabic, creating a condition whereby vowel graphs are given less attention than
consonant graphs. Subsequently, this vowel blindness hypothesis is often discussed as
the cause for ALE spelling difficulty (see Alsadoon & Heift, 2015; Bowen, 2011;
Dunlap, 2012; Saigh & Schmitt, 2012; Taylor, 2008), and it is cited as the reason why
vowel errors are more common for ALEs (Bowen, 2011; Dunlap, 2012). Alsadoon and
Heift (2015), for one, specifically target vowel blindness in their research designed to
†
Dunlap does not give the frequency of more discrete vowel categories.
‡
The exam papers were reported to be written “spontaneously” without the aid of dictionaries, and on a common topic
(p.47).
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 3
improve ALE spelling ability, implying that vowel blindness is a major obstacle and
prominent cause for ALE poor spelling skills.
The results of some studies have suggested that vowel blindness is a valid condition
for literate Arabic (Hayes-Harb, 2006; Ryan & Meara, 1991; Saigh & Schmitt, 2012)
and Hebrew speakers (Koriat, 1984). To examine the possible effects of vowel
blindness on spelling recognition and production, Saigh and Schmitt (2012) tested if
ALEs notice omitted or incorrect graphemes representing tense vowels more than lax
vowels. Tense vowels such as [i] and [u] generally have a longer duration period than
lax vowels, making these tense vowels more like non-omitted Arabic long vowels.
They selected 40 frequent words with short vowels and another 40 words with long
vowels and embedded them in sentences. Each vowel occurred in three conditions: a
correct, incorrect, and omitted vowel condition. Incorrect vowels were represented by
a different vowel grapheme. 24 native Arabic speaking participants were then
instructed to mark each test sentence as either correct or to cross out and correct any
encountered misspelled words. The results showed that the participants often failed
to recognize incorrect or missing long vowels (i.e., about 1/3 of the errors were not
noticed and another 1/3 were not accurately corrected), but the results also showed
that the failure rate for short vowels was significantly greater (i.e., over 40% were not
noticed and nearly 1/2 were not accurately corrected). While, this provides evidence
that vowel quality affects ALE spelling accuracy, the ability of vowel blindness to
explain the degree and range of spelling mistakes by ALEs is still largely unclear.
ALEs also struggle with capitalization (Thompson-Panos & Thomas-Ruzic, 1983)
and choosing the correct consonant graphs in English (Bowen, 2011; Dunlap, 2012;
Haggan, 1991). In Haggan‟s (1991) study, consonant doubling errors (54 cases) and
other consonant errors (47 cases), were also relatively common. Silent <e>
§
misspellings were also problematic (36 cases). Furthermore, Saigh and Schmitt (2012)
found that their ALE participants caught missing vowels significantly more often
than incorrect vowels, suggesting that ALEs are aware of the importance of
representing the vowel position. Saigh and Schmitt (2012) also found that neither the
missing nor incorrect vowel-condition had a significant effect on a participant‟s ability
to spell the target word correctly. While ALEs paid attention to vowel graphs, they
had difficulty choosing the correct vowel graph in most cases. Additionally, ALEs
underperformed in comparison to other ESL groups when spelling words containing
both short and long vowels (see Fender, 2008).
Fender (2008) compared ALEs‟ ability to spell different word types with a group of
non-Arabic ESL participants to gauge the acquisition of more complex spelling
patterns. The study consisted of 37 ESL participants: 16 Arabic ESL students and 21
ESL students from Korea, China, and Japan. Three different spelling conditions were
created to evaluate each group‟s acquisition of English spelling rules from simple to
more complex words: a within word, syllable-juncture, and derivational spelling
condition. Monosyllabic words that had short, long, or complex vowels (digraphs and
§
<> indicates orthographic units, // phonemes, [] phonetic units, {//} morphemes, and {} the target spelling.
4 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
or diphthongs) composed the within word condition (e.g., cut, strange, cook, train).
Multisyllabic words consisting of doubled consonants, long vowels with open syllables,
and short vowels with closed syllables composed the syllable juncture condition (e.g.,
written, babies, kitchen), and multisyllabic words with derivational affixes made up
the derivational condition (e.g., responsible, education). The results showed that the
non-ALE group performed significantly better in all three conditions. The results also
showed that ALEs made more errors with multisyllabic words and words containing
derivational affixes as opposed to monosyllabic words with both short and long
vowels: “the problem [was] especially acute among the Arab ESL participants who
seem[ed] to struggle with orthographic complexity” (p. 34). Inexplicable spellings are
also frequently cited in the literature (e.g., “oniouns” for “audience” in Dunlap, 2012,
p. 26). The problem, consequently, appears to be larger than a short-vowel omission
transfer effect.
While it is possible that vowel quality contributes to some of the doubling errors in
Haggan (1991) and Fender (2008) (i.e., the coda consonant in monosyllabic words with
single-graph short vowels is doubled when suffixed with {-ing}, {-ed} etc… (e.g., hop
hopping), there is no known argument to the author‟s knowledge explaining how
vowel blindness specifically causes many of the other frequent error types reported.
Vowel blindness has simply been assumed to cause short-vowel errors, and thus its
outcome has not been clearly articulated. It is unclear why vowel blindness would
cause more graph-choice than graph-omission errors.
In addition to not fully understanding the cause of ALE orthographic difficulty, the
extent that proficiency in written English addresses the cause(s) of the orthographic
difficulty is not evident. When error types were compared between proficiency groups,
Haggan (1991) found that advanced ALEs performed significantly better than ALE
remedial students on consonant-doubling errors following an affix (e.g., swiming
{swimming}), and unnecessary silent <e> additions (e.g., withe {with}). Advanced
ALEs, however, made more other consonant errors (39 cases) than the remedial group
(only 8 cases), but the difference was not significant.
**
The results mainly showed
insignificant improvement with both consonants and vowels.
The depth of the English orthographic system may play a substantial role in graph
choice errors (cf. Fender, 2008; Taylor, 2008) and may cause vowel blindness to
appear more significant than it actually is. That is, even if vowel blindness is a valid
condition, it is possible that its effect is relatively minor when accounting for the
overall spelling production problem. Thus, while it is clear that ALEs have
orthographic problems in English, the cause or causes of this issue remain
insufficiently described and demonstrated.
1.2. The depth of the English orthographic system and vowels
The English orthographic system consists of 26 individual graphs derived from the
Roman alphabet. It reads from left to right, top to bottom. The system is deep because
**
Other comparisons did not reach significance or there were too many subcategories with zero counts for the chi-
square analysis used.
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 5
the mapping of phonemes to graphemes is irregular (Frost, Katz, & Bentin, 1987),
making it difficult for both native (Seymour, Aro, & Erskine, 2003) and non-native
speakers (Haggan, 1991) to learn.
English graphemes create several difficulties for ESL learners to overcome: graphs
link to many different phonemes ( <y> any /i/, syllabus //, shy /a /, year /j/ ); some
phonemes link only to a digraph ( // <sh>; /θ/ <th> ); some phonemes link to
both a graph and digraph ( /f/ <f>, <ph>, <gh> ); some phonemic contrasts have
no distinguishable graph or digraph contrast ( /ð/, /θ/ <th> ); some graphs or
digraphs will be assigned no value ( live <e>, height <gh>); some graphs or digraphs
may systematically change their surface, phonetic value when obtaining a morphemic
value ( <s> as {/plural/} cats [s], dogs [z], boxes [z], <ed> as {/past/} tugged
[d], trucked [t]). Stress placement, syntactic category, and the presence or absence of
other non-local graphs can affect the value of a given graph (finite vs. infinite, live
snails vs. to live, hop vs. hope). Furthermore, only 5 of the 26 letters are exclusively
used to represent ~11 vowels and ~8 diphthongs.
††
In contrast, 21 letters represent
English‟s ~24 consonants. Single vowel-graphs are used for long-vowels (to /u/, me /i/),
diphthongs (bacon /e/, bicycle /a/), and short-vowels (put //, mat /æ/). Digraphs are
also used for long-vowels (spoon /u/, feet /i/), diphthongs (trail /a/, pie /a/), and short-
vowels (certain //, book //). The English system is, thus, a deep system because of the
number of digraphs and the variable mapping of phonemes to graphemes, especially
for vowels.
The Arabic writing system consists of 28 primary graphs that are not derived from
the Roman alphabet. It reads from right to left, top to bottom. The Arabic system is
relatively shallow with more consistent grapheme to phoneme correspondences
(GPCs).
The Arabic orthographic system is a consonantal script or Abjad. Its 28 graphs
principally represent consonants: the short vowels /i/, /u/, and /а/ are not generally
present in written Arabic. They do appear, however, in the Quran and in texts for
language learners (Abu-Rabia, 1997; Fender, 2008). If present, short vowels are only
indicated by a diacritic mark: the graph < >, which represents the sounds [b/p], is
written as <
> for [bi], <
> for [bu] and <
> for [ba]. This means these short
vowel sounds either have secondary status or are not indicated in writing. As Arabic
learners become more proficient, though, they easily “fill in the missing vowels”, as
these short vowels often reflect grammatical information that can be gathered from
the greater context (Hayes-Harb, 2006, p. 2). Thus, individually, written words
usually do not display their full phonemic value.
A phonemic distinction exists between short and long vowels in Arabic. Unlike the
short vowels, the long vowels are not omitted. The consonant letters 'alif <>, yā‟<>,
and wāw <> are also used to represent the long vowels: /a:/, /i:/, and /u:/. Despite this
and the letter tā marbūṭa < > (which is a silent graph in modern Arabic), graphemes
and phonemes in Arabic correspond very closely, almost 1:1 (cf. Saigh & Schmitt,
††
~ means approximately: the exact number of vowels and diphthongs naturally depends on the dialect of English.
6 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
2012; Watson, 2007). Accordingly, because Arabic is more consistent, ALEs may not
be accustomed to matching variable graphemes with a particular phoneme as is
required in English.
As discussed, ALEs appear to struggle with vowels more than consonants, but
vowel-graph errors are also the most common error type for several other ESL groups
(Bebout, 1985; Dunlap, 2012) and for L1 English children (cf. Mullock, 2012). The
reason vowel errors are more common for most groups arguably results from the
depth of the English orthographic system: GPCs vary greatly with vowels in English
because of the small inventory of single graphs and the large number of vowel
digraphs and pronunciation differences.
Since the depth of the English orthographic system also creates an obstacle for
other ESL groups who use shallow orthographies, the difference in performance by
ALEs should be the result of other factors. The question, then, is whether vowel
blindness or something else coupled with the depth of the English orthographic
system is the cause (or a primary reason) for ALE spelling difficulties in English.
1.3. Why English spelling may be particularly problematic for ALEs
The depth of the English orthographic system should be a nearly equal problem for
several ESL groups. Spelling, however, was significantly less problematic for Spanish
ESL students (i.e., Spanish like Arabic utilizes a shallow orthography) and Korean,
Chinese, and Japanese ESL students, who do not use a Romanized script (see Dunlap,
2012, and Hayes-Harb, 2006). Other factors, such as phonological and morphological
differences between English and Arabic and the state of L1 literacy and education in
much of the Arabic world (cf. Taylor, 2008), may also contribute to the English
spelling problem, making the depth of the English system more challenging for ALEs.
The morphology of Arabic may play a role in ALE spelling errors. Arabic roots are
identified by a consonantal pattern. The script is mostly represented with different
consonant clusters that compose a particular root pattern (eg., k-t-b = something to do
with books/writing). This arguably creates a lot of repetition for Arabic readers by
limiting the visual variance of a particular root. This perhaps allows Arabic readers to
connect orthographic form to meaning more easily. Similarly, word length could
contribute to spelling mistakes. Words in Arabic tend to be short: “less than six
character long” (Randall & Meara, 1988, p. 135). This suggests the number of letters
needed to be stored for accurate word recognition and production in Arabic is more
limited than in English.
L1 literacy and education is another complicating factor that ought to be considered
when accounting for ALE spelling errors and reading and writing difficulty in
English. L1 literacy skills affect the quality of subsequent language learning (Carson,
Carrell, Silberstein, Kroll, & Kuehn, 1990; Carrell, 1991; Saiegh-Haddad & Geva,
2010). Reading and writing education in much of the Arabic world has often lacked
proper attention (cf. Taylor, 2008). Fender (2008) similarly suggested that many ALEs
are also weaker readers in their L1. The situation of diglossia within the Arabic world
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 7
contributes to this problem, because Arabic students must learn to read and write in a
language that is different from the language spoken at home (Abu-Rabia, 2000).
Accordingly, ALEs are not as skilled in reading and writing in Arabic. This suggests
that ALEs lack practice with word recognition and practice connecting
semantic/phonological forms to orthographic forms in writing.
One or a combination of these issues may be hindering ALEs‟ acquisition of GPCs
in English. Without this skill, ALEs will subsequently be more susceptible to errors
from orthographic depth and have weaker word recognition ability, resulting in
spelling problems and slower and less accurate reading and writing ability.
1.4. The importance of orthographic competence for ALEs
In addition to poor spelling skills, ALEs have also exhibited poorer reading and
writing skills than other ESL groups (cf. Fender, 2008; Randall & Meara, 1988;
Taylor, 2008). ALEs, nevertheless, have performed nearly the same or better on
listening and speaking tasks (Fender, 2008). Poor spelling skills likely contributed to
the discrepancy between ALE reading and writing skill and listening and speaking
skill.
‡‡
While much ESL research on reading instruction has focused on top-down
strategies, ESL learners who are weak readers in their L1 and or those who have
different L1 orthographic systems may not have the ability to decode a text even after
being given sufficient background information (Taylor, 2008, p. 31). Cultural gaps
cause guesses and inferences to be less successful, greatly hindering comprehension:
“the closer the match between their prior knowledge and the new knowledge, the
more accurately [students] comprehend” (Wang, Martin, & Martin, 2002, p. 98).
Clearly, there is a gap between the culture of the Arabic world and that of much of the
English texts ALEs encounter. Framed as such, ALEs must utilize bottom-up reading
comprehension strategies, and perhaps must do so more than other students.
Orthographic competence or awareness is a key component of writing speed and
accuracy and reading speed and comprehension (cf. Fender, 2008; Perfetti, 1997;
Perfetti & Hart, 2002; Saigh & Schmitt, 2012). The Lexical Quality Hypothesis
(Perfetti, 1992; Perfetti & Hart, 2002) states that efficient word retrieval relies on “a
fully specified orthographic representation (a spelling) and redundant phonological
representations” (p.190). From a bottom-up perspective, it is believed that weak
readers possess weak word recognition skills in both their L1 (Perfetti & Hart, 2002)
and a second language (Fender, 2008; Nassaji, 2003; Randall, 2009). The ability to
deconstruct words into phonemes and graphemes is limited. Consequently, poor
spellers are likely to be slow readers and to have lower reading comprehension skills
than better spellers.
Nassaji (2003) found that better graphophonic and word recognition skills (in
addition to better semantic/lexical processing skill) accurately separated stronger ESL
‡‡
Different processing strategies likely also play a role in ALE reading difficulty (see Randall and Meara, 1988).
8 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
readers from weaker ones. Fender (2008), noted that this and other research support
the idea that “a single orthographic lexicon serves both English word recognition and
spelling production”, meaning those with poor orthographic representations have
difficulty with both comprehension and production (p. 22). Accordingly, identifying the
primary cause(s) for ALE orthographic difficulty may aid in the development of more
effective corrective measures to improve ALE spelling skills, which may in turn
improve ALE reading and writing skills.
1.5. Proposal and research design
This work argues that ALEs have a larger, more fundamental problem with
orthographic competence in English than the vowel blindness hypothesis alone can
explain. This general deficit may be the true cause or a contributing factor for many of
the results attributed to vowel blindness because vowel graph errors should be more
difficult for any learner who is weak with GPCs. Subsequently, it suggests that this
problem is inherently linked with ALE reading and writing difficulty.
This study proposes an underdeveloped orthographic representation hypothesis
(URH) which states that ALEs are mostly relying on phonological representations and
a limited set of GPCs to spell words in English (see Fender, 2008, for a similar idea).
ALEs comparatively lack orthographic representations for whole-word forms. This
hypothesis places many ALEs near the partial alphabetic developmental stage of
spelling described by Ehri (1997), whereby breaking words into phonemes and
representing these with letters or the appropriate graph/digraph is difficult. The
prediction is that ALEs will have problems with both consonants and vowels and that
errors will increase as GPC and phonemic variation increases. Accordingly, graph
choice errors will be the most common category overall but more common for vowels.
This hypothesis does not exclude the possibility that ALEs are also less familiar with
derivational spelling rules or that short-vowel omission has an effect, but claims
insufficient whole word representation and GPCs are the core problem.
If vowel blindness is primarily responsible for ALE orthographic difficulty, then the
following strong hypothesis may be made: pronounced vowel omission errors will be
more frequent than extra vowel insertion errors, consonant omission errors, and silent
<e> errors; short-vowel omission errors will be more frequent than long-vowel
omission errors. Furthermore, as a weaker corollary, short-vowel graph choice errors
are expected to be more frequent than long-vowel ones.
To evaluate these claims, errors were categorized into vowel and consonant
graph/digraph error types and subtypes and tallied to see whether there were
significant distribution differences. Pearson's chi-square tests, as used in Dunlap
(2012) and Haggan (1991), were used to show significant differences between error
categories and proficiency levels and between types of error categories. If vowel type
errors attributable to vowel blindness constitute a larger percentage of the overall
errors, this would suggest that addressing vowel blindness (as done in Alsadoon &
Heift, 2015) is a priority when attempting to improve ALE spelling mistakes in
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 9
English. If vowel length does not appear to influence spelling error rates, this
suggests an alternative cause such as insufficient knowledge of English GPCs.
2. Method
2.1. Error categorization
While there is no standard method for categorizing orthographic errors, as
mentioned in the introduction, some previous studies (Bowen, 2011; Dunlap, 2012;
Haggan, 1991) have examined types of spelling violations by ALEs.
§§
Bebout (1985)
devised a discrete system which endeavors to universally categorize spelling error
types by learners of English, but only Haggan (1991) attempts to directly use it to
describe ALE spelling errors. All three studies categorized errors differently and
Bebout did not design the system to investigate the cause of ALE errors: Bebout‟s
system ignores vowel length as a variable.
This study proposes a way to categorize errors by ALEs to investigate vowels and
vowel length on omission errors, and GPC accuracy. It borrows from Bebout‟s system,
but the organization directly contrasts consonant and vowel type errors as done in
Dunlap (2012). It also deviates from Bebout‟s system by not using several unattested
error subcategories reported in Haggan (1991) and by focusing more on graphemes
(i.e., graphs and or digraphs as a single unit) rather than letters. This study also
eliminated several vague other categories, balancing error categories between vowel
and consonant type errors to compare the frequency of each. This design was
important under the premise that vowel blindness should effect the distribution of
vowel errors differently than consonant ones, especially for omission type errors.
This study divided omission errors into silent and salient categories, unlike Bowen
(2011) and Dunlap (2012). Silent omissions (e.g., <tim> {time} / hav {have}) are not the
same as short-vowel omissions in Arabic, making their connection to vowel blindness
less straightforward. Since silent graphs do not directly link to a phonemic value,
their omission is arguably the result of incomplete orthographic, lexical knowledge.
Like other studies, this study also examined metathesis (transposition) errors to
check whether writing direction in Arabic interferes with the order of graphs in
English. This was to compare the effect of one orthographic variable with another:
linear direction vs. omission. Unlike other studies, however, this study accounted for
transpositions involving only vowel graphs, consonant graphs or a combination of two
in order to examine if one type was more common.
This study also did not count form and morphological/pattern/rule type errors as
done in Dunlap (2012) in order to separate punctuation, word use, and morpho-
syntactic grammatical knowledge from word form and grapheme knowledge. The form
<musics> would not be counted as an error because the derivation of this word is
possible (e.g., The musics of the world emotionally unite us. „types of music‟) and
§§
Fender (2008) tested conditions where spelling errors were more likely to occur rather than types of spelling errors.
10 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
<deers> would not be counted as an error because this likely reflects
morphosyntactic/lexical knowledge instead of spelling accuracy.
Vowel and consonant segment violations (graph/digraph errors) were divided into
six major categories: graph choice, salient omission (a pronounced graph/digraph is
missing), silent omission (an unpronounced graph/digraph is missing), insertion (a
graph or digraph is inserted), and metathesis error categories. A single word could
contain multiple graph errors of one or several categories (e.g., chouc_latte {chocolate}
= 3 graph errors representing 3 different categories). The target word was determined
by context (e.g., “when I was a small shild” {child}) and when the target word could
not be clearly determined, the misspelling was not counted.
***
Table 1. Examples of General Error Categories
†††
Error Types
Vowel Error Examples Consonant Error Examples
1. Graph/Digraph Choice:
seviral (several), pai (pie)
cources (courses),plak (black)
2. Salient Omission:
inter_sted (interested)
hear_ (heart)
3. Silent Omission:
leag_e (league),
hai_t (height)
4. Insertion:
prefefers (prefers)
driviting (driving)
5. Metathesis:
thier (their)
[ingore]‡‡‡ (ignore)
6. Metathesis CV:
starnge (strange)
Table 1 line (1) demonstrates graph choice errors. These are errors where the
student failed to pick the correct graph or digraph, choosing instead the wrong graph
within the correct sequence. Line (2) type errors are ones where the student failed to
produce a graph or digraph that satisfied all the sound segments of the target word.
Line (3) type errors consisted of failing to produce a graph or digraph that has no
associated phonological value. Line (4) type errors involved inserting an extra graph.
With these errors, it is impossible to determine if the student intended for the graph
to be pronounced or silent. In most cases, however, the addition would result in an
additional syllable. Line (5) type errors are presumably the hardest to categorize
because a variety of things are involved. Nevertheless, if the produced letters
appeared out of order from the target form, these errors were counted as metathesis
errors, regardless if the graph was silent, salient, or a digraph. Finally, Line (6)
demonstrates metathesis CV errors. Since this error type involves both a vowel and a
consonant, it was categorically both a vowel and consonant type error.
This design also better accounts for the use of digraphs. It interpreted errors
involving a digraph as a single error rather than as two errors or as a transposition
error as done in Bowen (2011). Accordingly, this study interpreted <pai> {pie} as
applying <ai> for /a/ as opposed to <a> for <i> and <i> for <e>: two potentially
separate errors. Relatedly, <fainlly> {finally} is both a digraph error and a short-vowel
omission error in this study, instead of 1 transposition error, where <a> is moved
***
There were only a few cases where the target word could not be reasonably determined (e.g., adure, advistar {target
unknown}).
†††
All examples are ALE misspellings from this study except <ingore>. This example is from Haggan (1991) because
no consonant metathesis errors were found in this study.
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 11
ahead of <i>, as done in Bowen (2011, p. 92). This approach is arguably better because
the errors were generally accounted for with fewer assumptions (see Limitations
section for more on this).
2.2. Error subcategorization
Vowel blindness only ostensibly explains short-vowel error types. Unlike previous
studies that categorized produced spelling errors, this study divided omission errors
into long /e i a o u/
§§§
and short / æ / vowel errors. Another complicating
factor is the different rate of vowel occurrence: short vowels are more frequent,
creating more opportunity for errors. According to Cruttenden (2014), short vowels
create roughly 67% of vowels found in texts in general British English (short vowels:
// ~26%; // ~21%; / / ~ 7%; / æ / ~ 4%; // ~ 4%; // ~ 4%; // ~ 1.5 %), a figure which is
similar to general American English (p. 158-159). This percentage was used to adjust
the theoretical expected outcome when comparing long and short-vowel error counts
with a chi-square test of goodness-of-fit. This study likewise divided graph choice
errors into long and short subcategories to examine the possible weak effect of vowel
blindness. If vowel length influences graph-choice errors, we would expect short-vowel
errors to account for more than two thirds of the total errors because short vowels
make up approximately two thirds of the vowels.
The study also checked silent <e> errors following a short or long vowel. Silent <e>
errors may be GPC errors when they change the quality of the preceding vowel (e.g.,
cap vs. cape). They may also be a truly silent graph (e.g., have, some, one, because) for
which correct use requires complete lexical knowledge rather than correct phonology
and GPCs. Nevertheless, this study examined the frequency of silent <e> errors
occurring after short and long vowels to see whether fewer omission errors occurred
with long vowels.
Finally, to examine the possible effect of short vowels on a spelling rule, this study,
like Haggan (1991), checked whether doubling errors (i.e., errors involving two
adjacent identical graphs) occurred after the affixes {/-ing/}, {/-ed/} etc. This study also
checked whether a doubling error occurred in a monosyllabic word with a single-graph
short vowel (e.g., cut > cuting), digraph short-vowel (e.g., look > lookking), or long
vowel / glide coda / complex coda (e.g., take > takking; say > sayying; talk > talkking).
This was to see whether doubling errors were likely the product of GPC / word form
errors, vowel blindness or incomplete knowledge of a derivational spelling rule. A
large number of doubling errors after a short vowel could suggest a vowel blindness
effect. On the other hand, if most mistakes were stem internal (e.g., ocur [occur]), this
would suggest limited GPC / word form knowledge.
‡‡‡
[] within a Table indicates an unattested example from this study.
§§§
Rhotic vowels were not simply categorized as long vowels but instead as a vowel and a consonant. The vowel could
be short fur or long here.
12 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
Table 2. Examples of Vowel Error Subcategories
Vowel Error Subtypes
Examples
Graph Choice
Short Graph Error
geless (jealous)
Long Graph Error
butiful (beautiful)
Omission
Short Vowel
inter_sted (interested)
Long Vowel
unus_al (unusual),
Silent <e>, Morphological
happen_d (happened)
Silent <e>, Syllabic <l>
peopl_ (people)
Silent <e>, Short Vowel
som_ (some)
Silent <e>, Long Vowel
mistak_s (mistakes)
Vowel Insertion
Silent <e>, Morphological
reasones (reasons)
Root Final <e>, Short Vowel
sectore (sector_)
Root Final <e>, Long Vowel
companye (company)
Table 3. Examples of Consonant Error Subcategories
Consonant Error Subtypes
Examples
Graph Choice
1. Single for Single
televition (television)
2. Single for Digraph
preftionally (professionally)
3. Digraph for Digraph
mush (much)
4. Digraph for Single
toghothar (together)
Omission
5. Silent Omission Other
gover_ment (government)
6. Doubled Stem Omission
eag (egg)****
7. D. O. - Affix, Multisyllabic Stem
financialy (financially)
8. D. O. - Affix, Monosyllabic, Single Graph, Short Vowel
[weding] (wedding)††††
9. D. O. - Affix, Monosyllabic, Digraph, Long Vowel
realy (really)
Insertion
10. Insertion Other
teatchers (teachers)
11. Doubled Insertion
midell (middle)
12. Other Doubled Insertion at Affix
imottion (emotion)
13. D. I. - Affix, Monosyllabic, Single Graph, Other
[slowwing] (slowing)
14. D. I. - Affix, Monosyllabic, Digraph, Short Vowel
[bookking] (booking)
15. D. I. - Affix, Monosyllabic, Digraph, Long Vowel
[keepping] (keeping)
****
These are counted as a type of doubling error under the premise that there is no phonological cue to differentiate
between one consonant or a doubled consonant: beg and egg.
††††
This was the closest example of this possible error type found. The actual spelling produced was weedding for
wedding.
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 13
2.3. Research material selection
Assessment texts used to gauge ESL writing level proficiency at the end of a
semester at the University of Florida English Language Institute were selected for
this study‟s spelling error analysis. Each text was handwritten by one and only one
student who had been given approximately one hour to write either a few paragraphs
responding to the same, open ended prompts (e.g., describe a happy day in our life) or
an essay on a similar, common prompt.
Two or more instructors at the University of Florida English Language Institute
had independently rated each writing assessment on a proficiency scale of 1 (low) to 6
(high).
‡‡‡‡
Raters looked for writing structure, clarity, coherence, sentence complexity,
grammar, vocabulary, and spelling to make their decisions. Students who had
performed well on this task were placed into a higher level writing class. Since
students were finishing coursework in one proficiency level, if they had performed
well, they were placed into the next level or higher by this assessment. If they had not
performed well, they had to repeat the previous level course.
From these assessments, 20 ALE texts were selected for this study.
§§§§
Each text
had been written by an adult (18 years or older) ALE from Saudi Arabia. Each author
had completed at least four months of intensive English study in the U.S. Each text
was between 200 and 450 words long. These texts were then subdivided into two
groups based upon their proficiency level rating: 12 texts were rated low (10 at level 3
and 2 at level 2) and 8 texts were rated high (6 at level 5 and 2 at level 4). All level 2
and 3 texts were paragraph responses. Level 5 and 4 rated texts were essays
responses except for one level 5 text, which was a paragraph response. The possibility
that errors resulted from simple, careless typos was arguably reduced because the
texts were handwritten.
3. Results
3.1. Main category error results
Table 4 divides the total graph/digraph vowel and consonant errors found from the
20 assessment texts by main category.
Table 4. Main Category Consonant and Vowel Errors
‡‡‡‡
A third rater had scored the assessment if there was a disagreement over level placement.
§§§§
IRB-02 approval was issued via the University‟s review board as the assessments had been originally administered
and gathered for educational purposes other than research.
14 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
Error Type Vowel
Error Number
Error Type Consonant
Error Number
1. Graph Choice
180
1. Graph Choice
42
2. Salient Omission
42
2. Salient Omission
17
3. Silent Omission
36
3. Silent Omission
19
4. Insertion
33
4. Insertion
25
5. Metathesis
5
5. Metathesis
0
Subtotal Vowels
296
Subtotal Consonants
103
6. Metathesis CV
12
Total Errors
411
Figure 1. Proportion of Total Errors
A chi-square test of independence between vowels and consonants by error
categories showed a significant difference, χ2 (4, N = 399) = 18.771, p < .001. Cramer‟s
V further indicated that this difference was strong (V = 0.217). The null hypothesis
that vowel and consonant categories are independent can be rejected. Graph choice, χ2
= 5.51, and insertion, χ2 = 9.06, differences contributed most to this result as
compared to salient omission, χ2 = 0.28, silent omission, χ2 = 2.19, and metathesis, χ2
= 1.74.
The difference between vowels and consonants and graph choice and insertion
errors was also found to be significant: p < .001, Fisher‟s exact test, two tailed. No
significant difference, however, was found between consonant and vowel categories
and silent and salient omission error types: p = .55, Fisher‟s exact test, two tailed. A
chi-square test of goodness-of-fit test did not show a significant difference between
salient vowel omission and vowel insertion errors with an expected even distribution,
χ2 (1, N = 75) = 0.86, p = .35 (corrected for continuity).
*****
*****
The value of all chi-square tests with 1 degree of freedom was corrected for continuity.
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 15
Table 5. Errors by Proficiency Group
Low Level
High Level
Error type
Vowel
Consonant
Vowel
Consonant
1. Graph Choice
113
34
67
8
2. Salient
Omission
26
12
16
5
3. Silent Omission
17
17
19
2
4. Insertion
18
17
15
8
5. Metathesis
2
0
3
0
Figure 2. Vowel and Consonant Errors by Proficiency Group
A chi-square test of independence did not show a significant difference between low
and high proficiency levels across vowel categories, χ2 (4, N = 296) = 4.28, p = .36.
Similarly, the results did not show a significant difference between low and high
proficiency levels across consonant categories, χ2 (3, N = 103) = 3.627, p = .3. A
significant difference was, however, found using a chi-square test of goodness-of-fit
between low and high levels for consonant silent-omission errors with an expected
even frequency, χ2 (1, N = 19) = 5.7, p < .02 and consonant graph-choice errors with an
expected even frequency, χ2 (1, N = 42) = 6.83, p < .01. No significant difference was
found for addition or salient omission categories p > .05.
3.2. Results of subcategories
Table 6. Graph Choice Error Subtypes
Vowel Error Subtype
Number
Consonant Error Subtype
Number
Short Vowel
109
Single for Single
30
Long Vowel
71
Single for Digraph
9
Digraph for Digraph
1
Digraph for Single
2
A chi-square test of goodness-of-fit only showed a significant difference between
short and long vowels with an expected even frequency, χ2 (1, N = 180) = 7.6, p < .05.
When the expected rate for short-vowel errors was increased to 60%, there was no
16 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
significant difference, χ2 (1, N = 180) = .023, p = .87, (χ2 = 0, p = 1, when corrected for
continuity). When it was increased to 67%, following the frequency from Cruttenden
(2014), the difference was nearing significance, χ2 (1, N = 180) = 3.09, p ≅ .08, with
more long and fewer short-vowel errors than expected, assuming a vowel blindness
effect.
A chi-square test of goodness-of-fit between consonant single for digraph and
digraph for single-graph errors was also found to be nearly significant with an
expected even frequency, χ2 (1, N = 11) = 3.28, p = .07.
Table 7. Salient Omission Subcategories
Vowel Error Subtype
Number
Consonant Error Subtype
Number
Short Vowel
33
Word Final
10
Long Vowel
9
Word Internal
7
A chi-square test of goodness-of-fit also showed a significant difference between
observed and expected long and short-vowel salient-omission errors with an expected
even frequency, χ2 (1, N = 42) = 12.6, p < .001, but not at the proposed 67% rate, χ2 (1,
N = 42) = 2.05, p = .15. No difference was found between word final and internal
consonant omission errors with an expected even frequency, χ2 (1, N = 17) = 0.24, p =
.62. Moreover, comparing salient and silent omissions found no significant difference
with an expected even frequency, χ2 (1, N = 78) = 0.32, p = .57.
Table 8. Silent Omission Error Subcategories
Vowel Error Subtype
Number
Consonant Error Subtype
Number
Other Vowel
2
Other Consonant
2
Short Vowel
Silent <e>
13
Doubled Stem Internal
10
Long Vowel
Silent <e>
16
Doubled, Affix, Multisyllabic
5
Morphological
Silent <e>
2
Doubled, Affix, Monosyllabic, Short
Vowel
1
Syllabic <l>
Silent <e>
3
Doubled, Affix, Monosyllabic,
Long Vowel
1
A chi-square test of goodness-of-fit between silent <e> omission and other vowel
omission errors clearly showed a significant difference with an expected even
frequency, χ2 (1, N = 36) = 26.7, p < .001. No significant difference was found between
silent <e> errors following short and long vowels with an expected even frequency, χ2
(1, N = 29) = 0.14, p = .7.
A significant difference was found between doubled and other omission errors with
an expected even frequency, χ2 (1, N = 19) = 10.32, p = .001. No significant difference
was found between stem-internal doubled omission errors and doubling errors
following an affix with an expected even frequency, χ2 (1, N = 17) = 0.24, p = .62.
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 17
Table 9. Insertion Subtype errors.
Vowel Error Subtype
Number
Consonant Error Subtype
Number
Insertion (Other)
12
Insertion (Other)
11
Short Vowel Final <e>
11
Doubled Stem Internal
12
Long Vowel Final <e>
5
Doubled, Affix Other
2
Morphological Silent <e>
5
Doubled, Affix, Monosyllabic
0
No significant difference was found between other vowel insertion and silent <e>
insertion errors with an expected even frequency, χ2 (1, N = 33) = 1.94, p = .16.
Likewise, no significant difference was found between short and long silent <e>
insertion errors with an expected even distribution, χ2 (1, N = 16) = 1.56, p = .21. No
significant difference was found between doubling and other insertion type errors
with an expected even frequency, χ2 (1, N = 25) = 0.16, p ≅ .69. A significant difference
was, however, found between stem internal and doubling after an affix insertion-
errors with an expected even frequency, χ2 (1, N = 14) = 5.78, p < .02.
4. Discussion
This paper aimed to answer whether vowel blindness or the proposed
underdeveloped orthographic representation hypothesis better explains the types and
frequency of ALE spelling errors and to answer whether vowel errors decrease with
greater proficiency. Subsequently, it aimed to increase our understanding of which
spelling error types improve with overall stronger writing skills and to describe
prominent ALE spelling errors more discretely. This study found a significant
difference between the distribution of vowel and consonant errors with vowel errors
being more problematic than consonants (as similarly reported in other studies). This
study, however, did not find a significant difference in the percentage of consonant
and vowel omission errors or a clear association between vowel length and error rates,
suggesting that vowel blindness is not the core reason for ALE orthographic difficulty.
These results are valuable when considering appropriate pedagogical responses for
the orthographic problem, a problem that likely contributes to ALE reading difficulty.
4.1. Lack of evidence for a specific vowel blindness effect
As articulated here, the strong version of the vowel blindness hypothesis not only
predicts that vowels will be more problematic than consonants, but also that vowel
omission errors will be more frequent than vowel insertion errors. This was not the
case. Vowel omission errors were not found to be significantly different from vowel
insertions errors. As in Haggan (1991), vowel graph-choice errors were the most
common error type. Salient vowel omission errors only accounted for about 10% of the
total errors found. There was also no significant difference between salient and silent
omission errors. In addition, no significant difference was found between short and
long-vowel omission errors when the expected frequency rate of errors was adjusted to
reflect that short vowels occur more often. Likewise, no significant relationship was
found between long and short-vowel graph-choice errors. It is unclear why a vowel
omission effect was not detected. While these findings are technically only a failure to
18 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
reject the null hypothesis, they also suggest that vowel length and error rates are
independent. They indicate that ALEs are just as likely to insert an unnecessary
vowel graph as they are to omit a necessary one, just as likely to omit long vowels as
short vowels, and just as likely to choose the wrong graph for both short and long
vowels when engaged in a writing task. Thus, these results do not support either the
weak or strong version of the vowel blindness hypothesis discussed here.
The findings also failed to show that ALEs are significantly improving on the
spelling of vowels. They, however, did show significant improvement with consonants
on silent graphs and graph choice. The improvement on silent graphs and consonant
doubling errors concurs with Haggan‟s (1991) findings while the improvement on
graph choice is the opposite of what Haggan found. The lack of improvement with
vowels coupled with some improvement with consonants could be taken as evidence
that vowels are being treated categorically different. Nevertheless, while the
differences were not significant, the high level ALEs did improve in every vowel
category except for errors involving silent <e>. Because GPCs are more variable for
vowels, this may have contributed to this result.
4.2. URH and GPCs
As introduced, the URH posits that ALEs are not acquiring lexical orthographic
representations of a similar quality as compared to other groups. The URH predicts
greater vowel errors while also explaining problems with consonant graphs. The
greater GPC variance of vowels is argued to cause the disproportionate number of
vowel errors and make learning vowels more problematic. A vowel-omission transfer
effect may also exacerbate this problem, but much like the infrequency of metathesis
errors in this study and in Haggan (1991), these kinds of orthographic differences
appeared to cause few errors. The URH may then better explain why the short and
long-vowel distinction did not affect the frequency of graph choice or omission errors.
If ALEs have poor orthographic representations, they likely build them from
phonological ones. This is accomplished via GPCs, which appear limited (e.g., chiken
{chicken}) or mismatched (e.g., <sh> for [t] mush {much}). Accordingly, misspelling
consonant digraphs with a single graph was more frequent than misspelling single
graphs with digraphs and this difference was nearly significant. Single graph for
single graph errors were the most common and often reflected using a graph
incorrectly while preserving the correct pronunciation: consentrate (concentrate);
televition (television); engoy (enjoy). This suggests that ALEs may resort to using
simpler, more common GPCs that reflect accurate pronunciation while failing to
notice incorrect word shapes, even in very common words/roots (e.g., vision, much,
enjoy). GPC problems can also explain many of the other consonant graph errors. For
instance, an English phoneme that does not exist in Arabic caused several of the
graph choice errors (e.g., /b/ vs. /p/: proplem {problem}, berfect {perfect}).
English‟s deep orthography and Arabic‟s smaller inventory of phonemic vowels
accounts for the prevalence of vowel graph choice errors. For example, the phoneme
/a/ can correspond with a single graph <i>, two non-local graphs <iCe>, a single
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 19
consonant/vowel graph <y>
†††††
, or a digraph <ai>, <ie>. If ALEs are building from the
phonological level, they may rely on 1 or 2 common graphs/digraphs for this sound
(i.e., this study found that <ai> was often mistakenly used for /a/: maight {might};
taires {tires}; insaid {inside}; orgnaize {organize}. This contradicts the preference to
use a single graph over a digraph. It may be that ALEs tend to use one graph for each
sound: diphthongs perhaps are perceived as two units. There appeared to be some
preference for using a single graph to represent short vowels: alredy {already}; geless
{jealous} did {dead}; famus {famous}. Moreover, <i> was often used for the vowel []/[i]:
thim {them}; seviral {several}; thise {these}.
Ryan and Meara (1991) noted in their study that the position of the deleted
segment within the word influenced the detection rate (i.e., deletions at word edges
were detected more frequently). Similarly, in this study salient vowel omissions
always occurred word internally and most often on unstressed vowels in multisyllabic
words (e.g., dang_rous {dangerous}). In this position, the pronounced vowel is often
reduced or deleted in speech. In addition, sonorant consonants often followed omitted
vowels (e.g., sudd_nly {suddenly}), suggesting that the sonorant consonant is
accounting for the nucleus of the syllable, making the omission of the vowel less
obvious. Thus, while it is not clear why and how letter position would influence the
effect of vowel blindness on short vowels, one may see how these results could emerge
from deriving orthographic representations from phonological ones.
Poor lexical representations likely caused many of the omission/insertion silent <e>
errors and omission/insertion doubled-consonant errors. The difference between
silent <e> omission and other silent vowel omission errors was significant, which
likely only means that silent <e> is a much more frequent silent graph. No significant
difference was found between other vowel insertions and silent <e> insertion errors.
Taken together, this may suggest that despite the prevalence of silent <e>, it is not
part of the lexical representation. GPCs also did not seem to influence the
distribution, as errors did not predictably follow vowel quality: no significant
difference between insertion/omission of silent <e> and vowel length was found. Weak
orthographic representations may, therefore, explain the frequency of these errors. In
addition, the pronunciation of schwa after words like as and child might account for
errors such as ase {as} and childe {child}.
If an orthographically doubled consonant is pronounced differently, the difference is
not very salient (e.g., tomorrow vs. tummy). A significant difference was found
between doubling and other omission errors with there being more doubling errors,
but no significant difference was found between stem internal and doubling omission
errors following an affix. This suggests that doubling is the primary reason for
consonant omission errors, but that it is not associated with an affixation spelling
rule. No significant difference, however, was found between doubling and other
insertion type errors. A significant difference was found between insertion errors
involving stem internal doubling and doubling after an affix with there being more
†††††
One could make the argument that this graph is a vowel graph or a semivowel graph.
20 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
stem internal errors. In fact, this study did not find any cases of overapplying the
doubling rule to digraphs, long vowels, glide codas, or complex codas. This suggests
that doubling is not the primary reason for insertion errors and that when such errors
do occur, they are not likely caused by the over-application of the spelling rule. These
results are different from Haggan‟s (1991) because that study found most doubling
insertion and omission errors to be at the affix/stem boundary.
4.3. Limitations and suggestions for future research
It was not possible to be completely confident about the categorization of every
error. For instance, it is possible that some errors counted as silent omission errors
were actually metathesis errors: does achiev result from achive or achieve? This study
categorized this error as a silent <e> omission error given the evidence from clearer
examples which signal a propensity to make this error type.
Sampling was dictated more by convenience than true randomness. Samples could
only be taken from ALEs attending the English Language Institute who had
completed the assessment. These problems, however, exist in the other studies on this
topic as well.
The educational background (exact length of time studying English), L1 reading
and writing proficiency, and knowledge of other languages was unknown for each
author of the analyzed texts. Likewise, the other studies have not consistently
reported or controlled these variables. One may wish to consider and control these
variables in future research. Future study may want to examine typed errors and the
use of spellcheckers. Such studies might also want to compare freely written ALEs
errors with another group (such as Hebrew) whose L1 writing system also omits
vowels.
5. Conclusions
The results of this study suggest that even if vowel blindness is a valid condition, it
is not the core problem. The results show that both vowel and consonant errors are
problematic across several categories but that vowel errors are much more frequent.
While the distribution of vowel and consonant errors appeared to be significantly
different, the cause of this difference was not omission errors as predicted by the
strong version of the vowel blindness hypothesis. Instead, graph choice and insertion
error frequencies were significantly different, suggesting that graph choice was
especially problematic for vowels and that insertions were relatively problematic for
consonants. Short vowel graph-choice and salient-omission errors were not
significantly greater than long-vowel errors. Assuming vowel blindness more strongly
affects short vowels, a larger percentage of short-vowel graph and omission type
errors should have been found. Likewise, finding more salient omission than silent
omission errors would more clearly indicate vowel blindness, but this study did not
find a significant difference between these two error types. These findings, thus,
suggest vowel blindness is not a useful hypothesis when attempting to explain the
core cause of ALE spelling errors. Instead, spelling from a phonological representation
may better explain the common distribution of consonant and vowel silent / salient
Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22 21
omission errors. GPC errors may better explain the consonant graph-choice errors and
the significantly higher number of vowel errors. GPC errors are more numerous for
vowels because GPCs are more variable for vowels and ALEs may lack the literacy
skills/learning habits to overcome this problem easily. If this conclusion is correct,
teachers should explicitly teach both consonant and vowel GPCs to ALEs but focus
more on the accurate production of different graphemes representing the same vowel
sounds. Improving GPC awareness in ALEs will likely not only improve spelling
accuracy but may also improve reading comprehension and speed. Accordingly, more
research is needed to confirm this and to test effective means for improving GPCs for
ALEs.
Acknowledgements
First, I want to thank the administrators at the University of Florida‟s English
Language Institute for helping provide access to the raw data for this project. I would
also like to give thanks to the anonymous reviewers for their helpful suggestions and
corrections.
References
Abu-Rabia, S. (1997). Reading in Arabic orthography: The effect of vowels and context on
reading accuracy of poor and skilled native Arabic readers. Reading and Writing: An
Interdisciplinary Journal, 9(1), 65-78. http://dx.doi.org/10.1023/A:1007962408827
Abu-Rabia, S. (2000). Effects of exposure to literary Arabic on reading comprehension in a
diglossic situation. Reading and writing, 13(1), 147-157.
http://dx.doi.org/10.1023/A:1008133701024
Alsadoon, R., & Heift, T. (2015). Textual input enhancement for vowel blindness: A study with
Arabic ESL learners. The Modern Language Journal, 99(1), 57-79.
http://dx.doi.org/10.1111/modl.12188
Bebout, L. (1985). An error analysis of misspellings made by learners of English as a first and
as a second language. Journal of psycholinguistic research, 14(6), 569-593.
http://dx.doi.org/10.1007/BF01067386
Bowen, H. (2011). Spelling it out! Accounting for spelling difficulties for Arab learners of
English. In T. Smith (Ed.), Foundations for the future: Focus on vocabulary (pp. 85-98). Abu
Dhabi, United Arab Emirates: HCT Press.
Carrell, P. L. (1991). Second language reading: Reading ability or language
proficiency? Applied linguistics, 12(2), 159-179. http://dx.doi.org/10.1093/applin/12.2.159
Carson, J. E., Carrell, P. L., Silberstein, S., Kroll, B., & Kuehn, P. A. (1990). Reading‐writing
relationships in first and second language. Tesol Quarterly, 24(2), 245-266.
http://dx.doi.org/10.2307/3586901
Cruttenden, A. (2014). Gimson's Pronunciation of English. New York, New York: Routledge.
http://dx.doi.org/ 10.1017/S0025100303231121
Dunlap, S. (2012). Orthographic Quality in English as a Second Language, Ph.D. dissertation,
School of Arts and Sciences, University of Pittsburgh.
Ehri, L. C. (1997). Learning to read and learning to spell are one and the same,
almost. Learning to spell: Research, theory, and practice across languages, 13, 237-268.
Fender, M. (2008). Spelling knowledge and reading development: Insights from Arab ESL
Learners. Reading in a Foreign Language, 20(1), 19-42. Retrieved from
http://search.proquest.com/docview/1705668932? accountid=12653
Frost, R., Katz, L., & Bentin, S. (1987). Strategies for visual word recognition and
orthographical depth: a multilingual comparison. Journal of Experimental Psychology:
22 Robert J. Deacon / Eurasian Journal of Applied Linguistics 3(2) (2017) 1–22
Human Perception and Performance, 13(1), 104. http://dx.doi.org/10.1037/0096-
1523.13.1.104
Haggan, M. (1991). Spelling errors in native Arabic-speaking English majors: A comparison
between remedial students and fourth year students. System, 19(1-2), 45-61.
http://dx.doi.org/10.1016/0346-251X(91)90007-C
Hayes-Harb, R. (2006). Native speakers of Arabic in ESL texts: Evidence for the transfer of
written word identification processes. TESOL Quarterly, 40, 321- 339.
http://dx.doi.org/10.2307/40264525
Koriat, A. (1984). Reading without Vowels-Lexical Access in Hebrew. Attention and
performance, 10, 227-242.
Mullock, B. (2012). An examination of commercial spelling programs for upper primary level
students. Australasian Journal of Special Education, 36(2), 172-195.
http://dx.doi.org/10.1017/jse.2012.14
Nassaji, H. (2003). Higher-level and lower-level text processing skills in advanced ESL reading
comprehension. The Modern Language Journal, 87, 261– 276.
http://dx.doi.org/10.1111/1540-4781.00189
Perfetti, C. (1997). The psycholinguistics of spelling and reading. In Perfetti, C., Rieben, L.,
and F. Michel (eds.), Learning to spell: Research, theory, and practice across languages.
Mahwah, New Jersey: Erlbaum. 21–38.
Perfetti C., & Hart, L. A. (2002). The lexical quality hypothesis. In Vehoeven L., Elbro C.,
Reitsma P. (eds.), Precursors of Functional Literacy. Amsterdam: John Benjamins, 180–213.
http://dx.doi.org/10.1075/swll.11.14per
Randall, M., & Meara, P. (1988). How Arabs read roman letters. Reading in a Foreign
Language, 4(2), 133-145.
Randall, M. (2009). Second Language Reading Proficiency and Word Recognition: The Concept
of Saliency and Its Application Across Different Scripts. Issues in Second Language
Proficiency, 116-131.
Ryan, A., & Meara, P. (1991). The case of the invisible vowels: Arabic speakers reading
English words. Reading in a foreign language, 7, 531-531.
Saiegh-Haddad, E., & Geva, E. (2010). Acquiring reading in two languages: An introduction to
the special issue. Reading and Writing, 23(3), 263-267. http://dx.doi.org/10.1007/s11145-
009-9208-4
Saigh, K., & Schmitt, N. (2012). Difficulties with vocabulary word form: The case of Arabic
ESL learners. System, 40(1), 24-36. http://dx.doi.org/10.1016/j.system.2012.01.005
Seymour, P.H., Aro, M., & Erskine, J.M. (2003). Foundation literacy acquisition in European
orthographies. British Journal of psychology, 94(2), 143-174.
http://dx.doi.org/10.1348/000712603321661859
Taylor, M. (2008). Orthographic and phonological awareness among L1 Arabic ESL learners: A
quasi-experimental study. Retrieved from ProQuest Dissertations and Theses database.
Thompson-Panos, K., & Thomas-Ruzic, M. (1983). The least you should know about Arabic:
Implications for the ESL writing instructor. TESOL Quarterly, 17, 609–623.
http://dx.doi.org/10.2307/3586616
Wang, Y., Martin, M. A., & Martin, S. H. (2002). Understanding Asian graduate students'
English literacy problems. College Teaching, 50(3), 97-101.
http://dx.doi.org/10.1080/87567550209595885
Watson, J. C. (2007). The phonology and morphology of Arabic. Oxford: Oxford University
Press. http://dx.doi.org/10.1017/S0952675703004548
Copyrights
Copyright for this article is retained by the author(s), with first publication rights granted to the Journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons
Attribution license (CC BY-NC-ND) (http://creativecommons.org/licenses/by-nc-nd/4.0/).
