The development of phonemic and orthographic spelling patterns

and the reduction of/t/ to a tongue flap

in words like water, butter, and letter. Most children in the phonetic spelling stage write

truck as , dragon as , and water as (Treiman, 1985;

1993). These errors suggest that children analyze spoken words into the constituent

phonemes but their phonemic classifications do not match symbols dictated by the

conventional writing system.

Letter reversals like for her are also frequently found during the phonetic

spelling stage and reflect linguistically-based errors (Treiman, 1997b). For children, the

spoken form of her contains an initial consonant followed by a syllabic /r/. Young

children frequently omit the vowel altogether when spelling such words, as discussed

earlier. In other cases, a child may remember having seen an e in the printed form of her

or may realize that this word, like all other English words, must contain a vowel.

Because the spoken form of the word does not indicate where the vowel should be

placed, the child may arbitrarily misspell the word as . Errors that include a final








e may be particularly common because many words, like came and home, end with a

"silent" e. Thus, an error such as for her may be an invention that reflects both

phonology (the syllabic /r/ sound in the word) and orthographic knowledge (the

knowledge that all English words must be written with a vowel letter and that several

English words end in the letter .) It is largely a coincidence that such errors of letter

reversals tend to contain all of the letters of the conventional spelling.

The final stage of spelling development is called the morphemic, or within-word,

pattern stage. To reach this stage, children must be able to recognize the constancy of

many morpheme spellings and learn to consider meaning, sound and syntax to spell

correctly (Bryant, Nunes, & Bindman, 1997; Nunes, Bryant, & Bindman, 1997). The

transition into correct spelling requires children to absorb certain basic concepts about the

structure of English orthography. Children learn, either explicitly or implicitly, that

many vowels are spelled with digraphs, that some letters can be doubled, that spellings

change when endings are added and that much phonetic detail is not represented in

writing. This stage of spelling acquisition lasts well into the secondary school years and

constantly undergoes refinement with exposure to the printed word, reading skills and

understanding of the grammatical structure of the English language. At this stage, writers

rely to a large extent on the use of educated analogies. Thus, once a novel word is

analyzed semantically and syntactically, it is spelled by analogy to another word of the

same phonetic structure as well as syntactic class. According to Ehri (1989), word

features are cross-referenced in memory. When one aspect of a word image is activated

in memory, the other known features are also activated along with other words that share

those features. Accumulating this knowledge base is a protracted developmental process.

Words are learned in stages and word concepts are elaborated gradually with experience.








This stage is also characterized by increasing use of language-specific orthographic

rules, such as the twinning of a consonant when adding a suffix or the deletion of a final e

when adding a suffix that begins with a vowel. As a note of caution, it is incorrect to

assume that children in the earlier stages of spelling acquisition are totally unaware of

orthographic rules Orthographic rules may be classified by degree of complexity based

on the frequency of their application to vocabulary items (Carney, 1994). For instance,

the vowel lengthening magic -e rule in monosyllables such as home, lake, and kite is a

simpler rule of orthography compared to the consonant-syllabic I-silent e rule in words

like cradle, title and nozzle. Cassar and Treiman (1997) conducted a series of

experiments on kindergarteners and first-graders using nonsense words containing both

legal and illegal combinations of consonants. The experiments probed children's ability

to recognize acceptable spelling patterns based on positional letter constraints. Examples

of the test items included word pairs such as "nuck-ckun" and pess-ppes". Results of the

experiments indicated that as early as in first grade, young spellers are sensitive to simple

and highly frequent orthographic rules. Cassar and Treiman's study (1997) not only

emphasized the fact that the acquisition of spelling is a continuous learning process but

also suggested that every stage represents the use of multiple spelling strategies which

integrate the phonological and orthographic elements of spelling. The extent of this

integration is not however equal among the two elements at any stage. The phonetic

stage is, for instance, heavily loaded with strategies of sound properties whereas the

morphemic stage is largely characterized by orthographic rules that predict the spelling of

morphemes.

Although little is known about the nature and sequence of later stages of spelling

development, there is no doubt that middle-grade and high-school children continue to








refine their concepts of orthography and their ability to associate spelling patterns with

speech patterns (Moats, 1995). As children learn more words, and have more examples

in memory of redundant spelling patterns, they can rely increasingly on analogy

strategies to spell (Ehri, 1989; Goswami, 1988). Children also become increasingly more

aware of opaque phonological relationships among derived words (Templeton, 1992),

and employ this knowledge, as opposed to visual recall, to generate a correct spelling.

Finally, advanced spellers learn to use multiple sources of linguistic information about

words from their orthographic, phonological, semantic and etymological knowledge to

facilitate spelling memory, recall and decoding.

Theories on the Process of Spelling

Experimental and naturalistic studies of spelling errors and slips of the pen have

contributed to a large extent towards our understanding of the mental processes involved

in the act of spelling. The traditional school of thought (Fry, 1985) regards the act of

spelling as a process of retaining and storing visual orthographic images learned by

memorization of letter sequences The psycholinguistic school of thought (Read, 1986;

Treiman, 1993) describes the acquisition of spelling as a creative process and stresses the

importance of phonological awareness as a basic prerequisite to successful spelling skills.

Knowledge of orthographic rules is required to refine advanced spelling patterns and

develops in later stages of spelling acquisition with exposure to print. Currently, there

exist two competing theories to explain the mental processes involved in spelling. The

dual-route theory proposes that there are two separate neural pathways that can be

recruited during spelling. The connectionist theory emphasizes the extent to which

different aspects of word knowledge are tapped simultaneously and in parallel during the








spelling process. This theory also emphasizes the interdependence of one type of

linguistic knowledge on another.

Many cognitive neuro-psychologists have presented evidence for a dual-route theory

of reading and spelling from laboratory experiments and detailed case study analyses of

individuals with brain lesions (Caramazza, Miceli, Villa, & Roman, 1987; Ellis, 1993;

Frith, 1980a). The dual-route theory of spelling is based on the conceptual premise that

children pass through distinct stages of spelling development, and that impaired spelling

development can be characterized in terms of inability to progress beyond a particular

stage. In Ellis' (1993) and Barry's (1994) dual-route model of spelling processes, there

are two major modules that communicate in parallel via a semantic processor. The

phonological module (also known as the assembled route) stores information about the

sound structure of words. It is specialized to analyze words into their component

phonemes and then assign each phoneme to its corresponding grapheme. Spelling via the

phonological module is regarded as a slow process. It is mainly used when encoding

novel words or pseudo-words following a regular phonological structure. Irregular words

would be spelled with phonologically plausible spelling patterns.

The visual-orthographic module (also known as the lexical or word-specific route)

stores information about graphemes and grapheme sequences in words. In its capacity as

a storehouse for word spellings, the visual-orthographic module holds the spelling

patterns of all memorized words irrespective of their degree of phoneme-grapheme

predictability. The spelling of a particular word can be retrieved from the visual-

orthographic module and produced in response to a meaning cue and/or a pronunciation

cue. If a meaning cue alone triggers the recall of a visually-learned spelling, a lexical

error may arise since the target word may be substituted by a synonym, as in large for








huge. If the retrieval of a word from orthographic memory is triggered by a sound cue

without full connection with a meaning cue, a homophone might be written, as two for

too. If the retrieval of a word from orthographic memory is stimulated by a partial

phonological cue, a word with similar but slightly different phonemic structure might be

written, as rich for ridge. Finally, if the letters of a word are not completely known, a

visually similar word might be retrieved from the orthographic processor with omission,

or reversal, of letters, as in the following slips of the pen made by literate adult writers

(wuold for would; poneme for phoneme, and niht for night) Once a spelling has been

retrieved either lexically, or assembled, it is deposited in a presumed graphemic output

buffer which holds spellings while output processes such as writing, typing, or oral

spelling, are being implemented.

The dual-route theory for spelling considers the visual-orthographic module as the

major spelling processor. Consequently, it accounts adequately for slips of the pen and

the spelling errors observed in cases of acquired dysgraphia (Caramazza et al., 1987;

Posteraro, Zinelli, & Mazzucchi, 1988; Shallice, 1981). However, studies on

developmental spelling provide increasing evidence of an integration between

phonological and visual-orthographic knowledge, an aspect which violates the

dissociation between information pathways in the dual-route theory. In reaction to this

weakness in the foundations of the dual-route theory, connectionist models of reading

and spelling processes were generated (Seidenberg & McClelland, 1989; Van Orden,

Pennington, & Stone, 1990). Connectionist models emphasize that language learning

depends on the extraction and recall of relationships among events. Seidenberg and

McClelland (1989) developed the distributed, developmental model of word recognition

and naming to depict the interactive nature between the involved processors. Hoffman








and Norris (1989) adapted and refined this word recognition model to developmental

spelling. According to the parallel, distributed processing model of Hoffman and Norris

(1989), the process of spelling begins at the lexical level. All relevant knowledge is

activated in memory when a word is spelled and various aspects of word knowledge

interact with and influence one another constantly (Adams, 1990; Ehri, 1980). According

to the transformational grammarians Chomsky and Halle (1968), a lexical item is

characterized by a set of identities The semantic identity specifies the meaning of a

word; the syntactic identity specifies the part of speech represented by a word; the

morphological identity specifies inflectional and derivational endings; and the

phonological identity specifies the acceptable sound patterns. The connectionist model

for spelling incorporates an additional orthographic identity to explain the rule-governed

associations between sounds and letters (Simon, 1976). The building blocks of this

orthographic identity are alphabet letters which, like words, are specified multi-modally,

including correspondences among alternative alphabetic codes such as upper versus

lower case, or print versus cursive.

Connectionist models, like dual-route models, pose that there are separate modules

responsible for storage and retrieval of phonological, orthographic, and semantic

information during the reading and spelling process. Unlike dual-route models where

information across modules is generally not recruited, a connectionist model proposes

that all aspects of word knowledge stored in the different modules are activated and used

simultaneously. This activation depends on the extraction and recall of relationships

among the different lexical identities. Thus, when a word is spelled, all relevant

knowledge is activated in memory, initiating an interaction between the different aspects

of word knowledge. Furthermore, each module consists of neural connections which,







21
when activated, are strengthened with practice and exposure to print (Hoffman & Norris,

1989). A beginning speller thus possesses the simplest forms of connectivity for spelling

relationships. As connections become stronger through exposure to, and practice with

print, the associations between phonemes and graphemes are retrieved more easily. In

contrast to the beginning speller, the experienced writer knows the spelling of a set of

lexical items and the frequency with which these words are used, determines the

association weights between sounds and symbols for that person.

A connectionist model for spelling attempts to account for the reason children's

earliest attempts at writing reflect phonological patterns from the beginning stages of

speech production. Ehri (1992b) reconceptualized the parallel distributed processing

model for reading and spelling and introduced her visual-phonological route model, or

amalgamation theory, to support the importance of phonological processing in reading

and spelling. Whereas a speller analyzes a word through the arbitrary connections

established between pronunciation and meaning using the visual pathways of both the

dual-route and the connectionist models, the amalgamation theory proposes that a word is

spelled through systematic connections established between phonological,

morphological, orthographic, semantic and syntactic features. In the early stages of

spelling acquisition, these systematic connections are established through knowledge of

letter-sound correspondences. The connections influence one another and through a

process of amalgamation, they become strongly bonded as orthographic images. It is

these images that are recalled every time a word is spelled. In the case of a novel or

nonsense word, a similar sounding orthographic image is recruited, but it is then amended

through systematic analysis of word knowledge. Ehri suggests that children who spell

poorly demonstrate greater discrepancy between phonological and orthographic








processing than good spellers. Since the poor spellers cannot recall an orthographic

image from the orthographic processor, they remain overly dependent on the sound

structure of the target word. In Ehri's opinion (1984, p. 145), print provides children

with a "schema for conceptualizing and analyzing the structure of speech."

In summary, based on Ehri's amalgamation theory (1992b), more current models

have been developed to explain how children learn to spell, first accurately, then rapidly

(Brown & Loosemore, 1995; Perfetti, 1992, 1997). These models stress the use of

multiple sources of information required for reading and spelling. Two pathways have

been suggested, a visual-orthographic pathway and a phonological-orthographic pathway.

In general, developmental spelling studies indicate that children recruit information from

both pathways thus enhancing the connections between phonemic and orthographic

segments of words, and eventually build in memory a store of systematic orthographic

word images. In the early stages, the development of accurate orthographic images

depends heavily on phonological awareness; in later stages, it depends on the ability to

integrate different features of word knowledge such as phonology, morphology, syntax

and orthographic rules. Most children acquire phonological awareness through indirect

instruction such as nursery rhymes and sounding-out word games. However, some

children have difficulty assimilating the concepts of letter-sound correspondences. These

children often exhibit difficulty in learning to read and write and require explicit

instruction in phonological awareness in order to build orthographic images in memory.

Phonological Processing as the Foundation for Learning to Spell

Phonological processing

Research has established unequivocally that spelling is a multifaceted form of

linguistic representation that integrates phonological, morphological, semantic, and








orthographic knowledge (Fischer, Shankweiler, & Liberman, 1985; Henderson, 1990;

Read, 1986). The core of this integrated framework of linguistic information that

embodies English spelling is represented by the alphabetic principle. This alphabetic

principle describes the conventional letter-sound associations of the language and lies at

the core of what is frequently called phonological processing. Phonological processing is

defined as "an individual's mental operations that make use of the phonological or sound

structure of oral language when he or she is learning how to decode written language"

(Torgesen, Wagner, & Rashotte, 1994, p. 276). Torgesen et al. further note that one of

the most frequently studied phonological processing skills is phonological awareness.

Phonological awareness is generally defined as one's sensitivity to, or explicit awareness

of, the phonological structure of the words in one's language. This explicit awareness

allows an individual to recognize that a spoken word consists of smaller components such

as syllables and phonemes and that these units can be consciously manipulated to decode

a written word, or analyze a word into its sound components prior to spelling it.

Phonological awareness is consequently measured by tasks requiring children to identify,

isolate, or blend individual phonemes in words (p. 276).

A developmental hierarchy of phonological awareness skills has been identified in

the literature on reading and spelling acquisition (Goswami & Bryant, 1990; Morais,

1991; Tunmer & Hoover, 1992; Tunmer & Rohl, 1991; Yopp, 1988). At the earliest

levels of the hierarchy, phonological awareness skills require the identification and

manipulation of syllabic units within a word (e.g., say 'football' without the 'ball'). The

next level of phonological awareness skills requires the ability to focus on components of

sounds that rhyme. Tasks at this intrasyllabic awareness level focus on the identification

and isolation of onsets and rimes in words (e.g., find the odd man out: 'bag, hag, nine,








tag'). The highest level of phonological awareness skills, referred to as phonemic

awareness, requires explicit manipulation of individual phonemes. Tasks of phonemic

awareness involve the identification of sounds in words, segmentation of words into

constituent sounds and blending sounds to form words or nonwords. At the phonemic

awareness level, an individual is capable of consciously manipulating phonemes and, in

the case of spelling, can represent individual phonemes with the corresponding

conventional graphemes.

Phonemic awareness and its role in spelling

Spelling requires the explicit ability to segment a word into its phonemic components

and then select the alphabetic symbol that corresponds to each sound in the word (Morris

& Perney, 1984; Tangel & Blachman, 1992). Several studies that examined the spelling

characteristics of children have shown that there exists a relationship between degree of

phonological awareness and spelling success in preliterate and literate children (Tunmer

& Rohl, 1991; Wagner & Torgesen, 1987). Treiman's extensive research on the spelling

errors of kindergarten and first grade children (1997) shows a strong relationship between

phonemic awareness and spelling development. Treiman found that children's early

spelling errors are not idiosyncratic attempts at writing but reflect the use of phonetic

strategies; as a result, children produce spellings such as 'jragon' for 'dragon' where the

word initial consonant cluster is affricated due to articulatory positioning.

Studies on the spelling characteristics of poor spellers also suggest overall weakness

in phonological awareness skills (Bruck & Treiman, 1990; Foorman & Francis, 1994;

Goswami & Bryant, 1990; Liberman, Rubin, Duques, & Carlisle, 1985). Bruck and

Treiman (1990) compared the phonological awareness skills of poor spellers to younger,

spelling-level matched controls. The results of the study suggested that the poor spellers








exhibited weaker phonological awareness skills than expected given their performance on

standardized spelling tests. The poor spellers in the study performed significantly worse

than the control group on tasks of phoneme deletion as well as phoneme recognition.

Although children in both groups made errors, the poor spellers exhibited greater

difficulty on the tasks and missed a higher number of test items. Stuart and Masterson

(1992) studied the relationship between prereading phonological abilities and later

reading and spelling skills in a group of 10-year-old children who were initially assessed

as 4-year-old prereaders. The results of the study indicated that children's scores on

phonological awareness tasks as prereaders significantly predicted their spelling

performance six years later. Children with good early phonological awareness obtained

consistently higher scores on tasks of nonword spelling and their errors were more

phonologically-driven than those made by children with poor early phonological

awareness.

The influence of phonological awareness on early spelling development is

highlighted by the results of explicit instruction in phonemic awareness. Ball and

Blachman (1991) investigated the effect of training phoneme segmentation and

instruction in letter names and letter sounds on word recognition and spelling. Their

study included 89 kindergarten children whose mean age was 5.71 years. Subjects were

divided into three treatment groups. The phoneme awareness group received training in

segmenting words into phonemes and in letter-name-sound correspondences; the

language group received training in letter-name-sound correspondences only; the control

group received no training. Results of the study showed that children in the phoneme

awareness group performed significantly better on reading and spelling than children in

the language and control groups respectively. Children in the phoneme awareness group








were able to generalize trained items to novel ones. Children in the language group did

not significantly improve in phoneme segmentation skills, reading skills or spelling skills

when compared to the control group. Tangel and Blachman (1992, 1995) conducted a

longitudinal study to trace the effect of explicit instruction in phonemic awareness on the

spelling skills of first grade children. Children in the intervention and control groups

were matched on age, sex, race, phoneme segmentation abilities, and word identification

test scores. The children in the intervention group received training in say-it-move-it

phoneme activities, segmentation activities, and letter-name and letter-sound activities

four times a week over a period of 11 weeks. The results of the study showed that

children in the intervention group outperformed the control group on tasks of phoneme

segmentation, letter-sound correspondence, and phonetic reading of regular words and

nonwords. Furthermore, children in the intervention group exhibited significantly better

spelling performance than the control group. Tangel and Blachman attributed this

significant difference to the combined effects of intensive and consistent training in

letter-sound associations and manipulations during the early stages of spelling

acquisition.

Schemes of Spelling Analysis

A review of the literature on spelling acquisition indicates that accuracy in spelling

forms is largely determined by the writer's awareness of (1) phonological characteristics

of spoken syllables (Hoffinan & Norris, 1989); (2) characteristics of individual phonemes

(Hoffman & Norris, 1989); (3) morphophonemic aspects of word structure (Schlagal,

1986; Sterling, 1983); and (4) orthographic structures (Henry, 1988). By implication,

effective interpretation of spelling can best be achieved by means of a scheme that

incorporates at least some of these factors. Over the years of spelling research, four








major schemes of analysis have been proposed and implemented. These schemes are (1)

serial sequencing of letters, (2) predictable/unpredictable words, (3) phonetic/dysphonetic

error classification, and (4) phonological analysis.

Serial sequencing of letters

In the 40s and 50s, the prevailing school of thought was that children mastered the

English spelling system by rote visual memorization. Learning to spell by this approach

presupposes attentiveness and a good visual memory but little linguistic knowledge.

Children were expected to memorize strings of letters forming the spelling of a word by

visual and/or auditory input, then retrieve this information through a memory recall loop.

Consequently, early studies on spelling (Fry, 1985; Mendenhall, 1930; Spache, 1940)

aimed to identify difficult words, or difficult elements in words, and to compile lists of

words generally misspelled by children. Misspellings were compared letter by letter to

the conventional spelling of the target word, then classified as errors of letter omission,

substitution, addition, or reversal. Such a classification scheme stresses the serial

ordering of letters within a word and assesses children's spelling skills as a function of

their ability to memorize and recall a string of letters as rapidly and as correctly as

possible. This scheme fails to capture the relationships between the spelling of a word

and its pronunciation, thus overshadowing the reasons why children make certain errors,

but outgrow others, in the course of their early years of spelling acquisition-

Predictable/ unpredictable words

This approach to spelling analysis divides lists for dictation into words comprised of

so-called predictable spelling patterns, and those that are unpredictable. Relative success

on these pre-categorized items is compared, with the expectation that children with

reading disabilities will perform differently on the two lists according to their strengths








and weaknesses. This implies that children with poor phonological processing skills are

likely to make more errors on the phonically predictable words relative to unpredictable

words that can be learned by memorization only. Analysis of spelling errors by a

predictable/ unpredictable word approach poses several conceptual problems. First,

predictability is a matter of degree. In English, a word's predictability comprises not

only phonetic transparency, but also morphological structure, word origin, and

orthographic conventions (Fischer, Shankweiler, & Liberman, 1985). Some words

follow patterns that are constrained by word origin and phonemic environment (e.g., the

long vowel sounds in the wordsfind, old, wild). Many words conform to orthographic

rules (e.g., after in words like have, love). The spelling of the majority of words

in English is phonetically opaque, but becomes predictable if analyzed in light of its

morphological structure (e.g., the phonetic variants for the plural morpheme).

Moats (1993) raises an important objection regarding the diagnostic and intervention

value of predictable/ unpredictable words lists. There is little evidence that children who

are poor spellers are more likely to respond differentially to word lists sorted as

predictable and unpredictable. Furthermore, there is no evidence that differences in

performance on two lists are reliable over time in younger children. Neither is there any

external validation that children who respond better on one list respond uniquely to any

given instructional approach.

Phonetic/ dysphonetic error classification

Another approach to analysis of spelling involves classifying errors as either good or

poor phonetic equivalents. Contrary to the traditional view that reading and spelling were

acquired through serial memorization, the phonetic/ dysphonetic error classification

system viewed learning as a creative process. Emphasis consequently shifted from a








visual orthographic task of memory to spelling as a developmental process. In this

system, phonetic accuracy is determined holistically, by judging how well the entire word

represents the sounds, or by applying phoneme-by-phoneme rating systems that specify

how well each speech sound is represented in an individual's spelling attempts. A

phonetic error, such as for cried, was considered a linguistically acceptable

representation of the target word. On the other hand, the spelling for peach

(Treiman, 1993, p. 34) violates all possible representations of the spoken word and is

therefore, a dysphonetic error. Recent work by Treiman (1997, 1998) presents a well-

documented critique on the weaknesses embodied in a two-dimensional phonetic/

dysphonetic error classification scheme. Treiman argues that such a scheme fails to

capture the developmental processes reflected in a child's spelling attempts. A few

salient examples include the reduction of consonant clusters in spelling, or the omission

of the letter in stressed syllables with r-controlled vowels (e.g., car, her, bird).

Children bring to the act of spelling their awareness of phoneme, syllable and rime units

integrated with a developing knowledge of orthographic conventions. They do not omit

elements from a consonant cluster or a syllable because they do not perceive them, or

cannot articulate them. They omit these elements because they do not possess sufficient

orthographic knowledge and thus resort to strategies like letter names and representation

of whole syllables rather than individual phonemes.

As spelling deficits are primary characteristics of reading disorders, this population

has often been the target of investigations on spelling acquisition. Frith (1980b) and

Nelson (1980) examined the phonetic accuracy of the misspellings of three groups of

children, good readers-good spellers (good), good readers-poor spellers (mixed) and poor

readers-poor spellers (poor). Based on the percentage of phonetically misspelled








phonemes, the mixed and the poor subjects constituted two distinct subgroups of

dyslexics. However, the phonetic/ dysphonetic classification scheme did not reveal any

qualitative difference between normally progressing children and the two subgroups of

poor spellers. Bruck and Waters (1988) refined the phonetic classification scheme by

introducing positional constraint criteria. In their scheme, spelling forms were

considered phonetically acceptable only if they satisfied orthographic conventions

relative to the position of the sound and its associated letter in a syllable. Thus, the

spelling for back is regarded as unacceptable because the digraph

necessarily follows a short vowel in a stressed syllable. Bruck and Waters' (1988)

constrained letter-sound classification system did show that mixed spellers have marked

deficits in letter-sound knowledge compared to good spellers. However, results of the

study failed to distinguish between subtypes of poor spellers.

In addition to its failure to distinguish between subtypes of poor spellers, Moats

(1993) points out another problem with the phonetic/ dysphonetic classification approach.

Such a classification scheme assumes that spelling is a dual route process involving both

a phonological processing system and an orthographic processing system, each of which

may be selectively dysfunctional in a dyslexic child. However, studies on the spelling

skills of dyslexic children suggest that both proficient and non-proficient spellers may

vary along the dimensions of phonological and orthographic processing (Bryant &

Impey, 1986; Moats, 1983; Stanovich, 1988). If phonetic accuracy or inaccuracy were

characteristics of subtypes, these subtypes would represent a normal continuum of

learning style differences rather than types of disorder.

Another problem with the phonetic/ dysphonetic dichotomy in spelling analysis is

the assumption that phonological and orthographic processes are associated with








distinctive aspects of spelling production (Moats, 1993). Diagnostically, such a system

will try to identify children who are on the extremes of continuous distributions in both

processes (Boder & Jarrico, 1982). This assumption clearly belies the present belief that

most children develop phonological and orthographic processing skills in an interactive

and integrated manner (Ehri, 1989),

Phonological analysis

In his study on children's creative spellings, Read (1986) noticed that spelling

patterns not only reflected a child's perception of the phonetic form of a word, but they

also generalized to other similar sounding words. After computing the frequencies of

spellings for selected phonemes, Read found that some phonetic properties were more

salient than others. For instance, children quickly learned that one letter symbolizes two

allophones of the plural -s. Read concluded that spelling is a dynamic process that

parallels linguistic and cognitive growth in children.

Moats (1993) underscores the importance of examining children's spelling errors

along a phonological classification scheme. In her view, a phonetic analysis of spelling

entails counting the number of words or letters that are plausible phonetic equivalents of

the phonemes in the target word. A phonological analysis, on the other hand, matches the

articulatory features of the whole word as a dynamic speech gesture with its spelling

representation. Moreover, phonological processes interact with morphological awareness

and knowledge of orthographic patterns when learning to spell. The purpose of a

phonological spelling analysis scheme is to capture the degree of interaction between

language processes and thereby, to ascertain why certain features are sources of errors. It

is also assumed that a phonological spelling analysis will detect subtle processing deficits








that distinguish the spelling proficiency of dyslexics and normal children (Bailet, 1990;

Carlisle, 1987; Rubin, 1988)

Moats' scheme for a phonological spelling error analysis rests on the following

assumptions (1993). First, the phonological analysis will match the articulatory features

of the spoken word to its spelling representation. Units of analysis will include the word,

the syllable, and the phoneme. This means that the salient articulatory features of the

syllable should be easier to match with symbols than features that are internally complex

(e.g., the closed syllable pat versus the consonant cluster in sprat). A typical outcome of

this assumption is the simplification of syllables in spelling that is not only reminiscent of

children's beginning speech productions (Hoffinan & Norris, 1989) but also recapitulates

the creative nature of spelling development.

Another assumption of the phonological spelling error analysis is that certain classes

of phonemes are more likely to constitute a source of spelling error than others because

of their place or manner of articulation, their position in syllable, their stress, or their

coarticulation with other sounds. Children with subtle compromise of their phonological

systems are expected to make relatively more errors on those aspects of words with high

phonological difficulty, and less errors on those aspects of words that are phonologically

simpler, less accessible, more contrasted, or easier to segment and identify (Moats, 1993).

Finally, a phonological approach recognizes that a person's ability to spell a given

word phonetically might interact with other linguistic factors such as word length,

frequency, grammatical class and morphology (Moats, 1993). Furthermore, a

phonological interpretation of spelling would predict error patterns to change with

acquisition of relevant knowledge as a child progresses through the different stages of

spelling development (Schlagal, 1992). In this light, Moats (1993) stresses that a








frequency count of phonetically accurate spelling errors can broadly characterize a

child's efforts at surface phonetic representation of speech sounds. A phonologically-

based spelling error analysis is more likely to capture the subtle problems persons with

dyslexia face when spelling the linguistically complex and phonologically elusive aspects

of words.

Dyslexia and Spelling

Although dyslexia has been of interest to researchers since Orton's work in the early

1930s, the spelling processes of reading-disabled individuals have been generally

neglected. Orton (1931) commented that, "Usually an inability to spell is treated as of

more or less minor importance in sharp contrast to a reading disability" (p. 169). In more

recent years, Frith (1983) pointed out that children who have problems with both reading

and writing are called "dyslexic" and not "dysgraphic" even though, as a group, dyslexic

children have consistently lower spelling achievement than reading achievement. The

last two decades have witnessed an upsurge of interest in the spelling processes of

reading-disabled children. The Orton Dyslexia Society (1994) Research Committee and

the National Institutes of Health recognize spelling as an important component of both

written language and reading and define dyslexia as

a learning disability, a specific language-based disorder of constitutional origin

characterized by difficulties in single word decoding, usually reflecting insufficient

phonological processing abilities... Dyslexia is manifested by variable difficulty

with different forms of language, often including, in addition to problems reading, a

conspicuous problem with acquiring proficiency in writing and spelling. (p.4)

This definition of dyslexia implies a close relationship between reading and spelling.

Assuredly, research has shown that young children use spelling-sound information for








both reading and spelling (Treiman, 1984; Waters, Bruck, & Seidenberg, 1985; Waters,

Seidenberg, & Bruck, 1984). This finding leads to the assumption that good readers and

good spellers should have better knowledge of spelling-sound correspondences than poor

readers and poor spellers who are deficient in phonological processing (Treiman, 1997).

This view led to four primary research directions in the study of the spelling processes of

dyslexic individuals, (1) in comparison to good readers, poor readers exhibit a higher

number of dysphonetic spelling errors, (2) poor readers exhibit particular difficulty in

spelling novel and nonsense words, (3) poor readers may exhibit better orthographic than

phonetic knowledge in spelling, and (4) poor readers exhibit considerable difficulty

applying morphological knowledge to spelling. The following section reviews the

literature that addresses these research directives.

Phonetic and dysphonetic spelling errors

Phonetic errors are those in which each phoneme is represented with a letter or letter

group that may be used to symbolize that sound in conventional English. In a strict

criterion system, an error is phonetic only if it sounds like the target word when read

aloud. For example, 'tak' for 'take' is not a phonetic error using a strict criterion because

the rules of English call for a final 'e' in this context. In a lax criterion system, an error is

phonetic if each phoneme is represented by a letter regardless of contextual constraints.

In Treiman's (1997) recent review of the literature, she found four studies showing

that children with dyslexia did not produce a higher proportion of dysphonetic errors than

younger children of the same spelling age. Pennington and colleagues (1986) used both

lax and strict scoring systems in their study of dyslexic adults and normally achieving

children. Using a lax criterion, 35% of the dyslexic adults' errors were phonetically

inaccurate, as compared to 33% of the normal children's errors. Using a strict criterion,








difference in performance between the two groups was also not significant, with adult

dyslexics making 75% of dysphonetic errors compared to normal children's 71%. Moats

(1983) found that 39% of the errors made by fourth to eighth grade children with dyslexia

were dysphonetic compared to 44% for typical second grade children. Nelson (1980)

used a lax criterion to compare the spelling errors made by dyslexics and younger normal

children. Her analysis showed that for dyslexics, 35% of the errors were dysphonetic

whereas for the younger normal children, 36% of the errors were dysphonetic. Finally,

Bradley and Bryant (1979) compared dyslexic children with spelling-level matched

younger children. For both groups, at least one phoneme was spelled correctly in 90% of

the errors. Errors in which none of the phonemes were spelled correctly were no more

common among the dyslexics (3%) than among the normal children (5%).

Other studies however, have found evidence that children with dyslexia made fewer

phonetic errors than typical children of the same spelling level (Bruck, 1988; Olson,

1985). Strong support for this idea comes from the results of a study by Bruck.and

Treiman (1990) which examined the performance of dyslexics and younger spelling-level

matched children when asked to spell words with consonant clusters. Using a lax

criterion for phonetic legality, the dyslexics produced 36% dysphonetic errors as

compared to 21% for the normal children. Bruck and Treiman attributed this difference

to the nature of the classification scheme. In their opinion, dylexics' spelling errors

followed a primitive level of phonologically-based errors. Although the frequency of

these errors was higher than that made by younger children of the same spelling level, the

errors themselves were qualitatively similar to those of normal younger children.








Spelling of novel and nonsense words

Since poor readers have a phonological processing deficit, it is logical to assume that

they would have particular difficulty reading and spelling novel and nonsense words.

However, the few studies that have examined this assumption have yielded contradictory

findings. Bruck (1988) scored children's spellings of nonsense words using a strict

criterion and found no difference in the relative proportion of errors for children with and

without dyslexia. Two other studies (Martlew, 1992; Siegel & Ryan, 1988) however

reported significant differences between the spelling performance of dyslexics and

younger normal children matched on either reading or spelling level. From these studies,

it appears that dyslexics' spelling errors on both real and nonsense words are qualitatively

similar to those of younger normal children. These errors are invariably phonologically-

motivated and reflect an earlier stage of spelling development. Whether children with

dyslexia make more spelling errors than normal younger children will depend on

individual weaknesses and the nature of the spelling words.

Orthographic knowledge

Poor readers appear to have little difficulty learning permissible letter sequences.

Nelson (1980), for example, found that 82% of the dyslexics' spelling errors were

orthographically legal (e.g., "cack" as opposed to "chak" for "cake") compared to 87%

for the normal group. In another study, Siegel, Share and Geva (1995) compared first to

eighth grade readers with dyslexia to groups of normal children matched for reading

level. The children were shown pairs of words such as "moke/moje" and asked to select

the one that most resembled a real word. The poor readers actually did better than the

reading-matched controls. These studies indicate that poor readers' orthographic

knowledge is at least commensurate with their overall level of reading and spelling and








may even exceed the knowledge younger normal children have about orthographic

sequences.

Morphological knowledge

Higher-level spelling tasks require knowledge of the morphological structure of the

language and its applications to spelling rules. Thus spellers need to recognize that

variations in pronunciation of related words are not necessarily accompanied by

corresponding changes in spelling. The three words "electrical, electricity, and

electrician" are all spelled with the same letter "c" despite differences in pronunciation.

Very few studies have examined morphological knowledge in poor readers. Carlisle

(1987) compared ninth grade poor readers to fourth graders matched for spelling level.

The poor readers performed significantly better than the younger children on an oral test

of derivational morphology, but they had difficulty applying this knowledge to print. For

example, they frequently could spell only one cognate word correctly in a pair like

"magic/magician".

Bruck (1993) compared dyslexic college students to sixth graders matched for

spelling level. The dyslexic students performed more poorly than the sixth graders on all

the spelling tasks except the ones that tapped morphological knowledge. On these tests,

both groups performed comparably. Findings of the studies done by Carlisle (1987) and

Bruck (1993) suggest that most poor readers, including adult college students with

dyslexia, have difficulty applying morphological knowledge to spelling.

In summary, the research comparing spelling abilities of good and poor readers

matched for reading and/or spelling level indicates more similarities than differences

across the two groups. The spelling errors that poor readers make are definitely not

unusual. The errors classified as dysphonetic are typically phonologically-based and








similar to errors made by normal beginning spellers. The only striking difference

between dyslexics and normal readers is the coexistence of primitive phonologically-

based errors with relatively high levels of orthographic knowledge. This apparent

contradiction resolves itself when one considers that dyslexics core problem lies in an

inability to analyze and manipulate the sounds in a word, hence their difficulty spelling

novel or nonsense words. Impairment in phonological processing does not impede the

development and mastery of permissible letter sequences.

The good-reader poor-speller paradox

Clinical cases have uncovered students who display a marked discrepancy between

reading and spelling performance. This group of dyslexics is referred to in the literature

as the "good-reader poor-speller", or simply the "mixed" group. The leading research

question is whether this mixed group represents a point on a developmental continuum

between good readers/good spellers and poor readers/poor spellers, or whether this group

possesses a distinct learning profile, suggesting a specific disability of developmental

dysgraphia as opposed to developmental dyslexia.

Frith (1980) was probably the first researcher to study the spelling performance of

the good-reader poor speller. In her study, she compared the performance of a group of

adolescent good readers/good spellers (good), good readers/poor spellers (mixed) and

poor readers/poor spellers (poor) on a variety of reading and spelling skills. When

spelling errors were classified as phonetic or nonphonetic, the good and mixed groups

showed a high proportion of phonetic errors, suggesting a basic knowledge of letter-

sound associations. Subjects in the poor group however produced a high proportion of

nonphonetic errors. Frith hypothesized that the good reader/poor speller has mastered the

orthographic strategy in reading but has failed to transfer it to the spelling process. As a







39
result, this group of readers is able to spell regular words but has persistent problems with

irregular words.

Frith's findings were seriously challenged in a subsequent study by Waters, Bruck

and Seidenberg (1985). In this study, 36 third graders were divided into three groups

(good, mixed and poor) based on standardized measures of reading comprehension and

spelling. Contrary to Frith's findings that good readers/poor spellers have adequate

letter-sound association knowledge, the study by Waters and colleagues suggested that

the mixed group does not possess good knowledge of graphophonemic relationships for

either reading or spelling. The mixed and poor groups produced a smaller proportion of

phonetically accurate errors than did the good group in tasks which included nonwords

and five types of words that differed in terms of their spelling regularity. Furthermore,

even though the children in the mixed group had been matched with children in the good

group on both reading comprehension, the number and type of errors made by the mixed

group on both the reading and spelling tasks were more similar to those of the poor group

than those of the good group.

Bruck and Waters (1988) further probed the issue of the good reader/poor speller in

a controlled experiment on third and sixth graders who were identified as good, mixed

and poor based on single-word decoding or reading comprehension scores. Both a

constrained and unconstrained scoring system was used to classify the spelling errors.

When the unconstrained scoring system was used, the mixed group was comparable to

the poor group in that both exhibited proportionately fewer phonetic misspellings than the

children in the good group. Another consistent finding was that the mixed group was

identified with deficits in at least one area of reading, either word recognition or

comprehension. Bruck and Waters concluded that the mixed group represents a point in








the continuum of sound-spelling correspondence knowledge. Distinctions between the

good, mixed and poor groups can be made only at finer levels of sound-letter association

knowledge.

In summary, spelling is a language-based skill with phonological processing

knowledge forming the foundation for efficient spelling performance Poor spelling

skills can be undeniably attributed to deficits in some aspect of phonological processing.

Individual differences in spelling ability are primarily caused by fine distinctions in the

knowledge and use of sound-letter information.

Implications of Socioeconomic Status on Literacy

Socioeconomic differences are conventionally indexed by demographic variables

such as household income and parents' education and occupation, either alone or in some

weighted combination. In educational studies, the socioeconomic status (SES) of a

school or a community may also be estimated by the percentage of the enrolment

qualifying for federal lunch subsidies. Demographic data consistently suggest that

children from poor families are exposed to a broad array of conditions which in

themselves may be highly detrimental to the health, safety and cognitive development of

a young child. When combined, these conditions are likely to serve as strong risk factors

for reading difficulties (National Assessment of Educational Progress, 1995). No known

published work has specifically examined the effects of a low SES on spelling

achievement. It is nevertheless hypothesized that the spelling skills of children coming

from a poor economic background will be depressed, especially if their reading skills are

poor (Snow, Bums, & Griffin, 1998).

Differences in literacy achievement among children as a result of SES have been

documented in several large scale studies (Stubbs, 1980; White, 1982; Baker, Serpell, &








Sonnenschein, 1995; Hart & Risley, 1995). Stubbs (1980) found that 7.5% of children

coming from a high SES were poor readers compared to 26.9% of the children coming

from a low SES. Snow, Bums, and Griffin (1998) suggest that SES differences in

reading achievement are actually a result of differences in the quality of schooling since

lower SES children tend to receive poorer educational opportunities. A recent study by

Alexander and Entwisle (1996) appears to demonstrate that low and high SES children in

elementary grades exhibit a virtually identical rate of progress. However, it is during

non-school time that low SES children fall academically behind their higher SES peers

and get progressively further behind. White (1982) discovered that, at the individual

level, SES is related to achievement only very modestly. However, at the aggregate

level, when achievement was measured as a school or community characteristic, the

effects of SES are much more pronounced. A low SES child attending a generally

moderate SES school is far less at risk than an entire school of low SES children. Baker

and colleagues (1995) compared opportunities for informal literacy learning among

preschoolers in the homes of middle-income and low-income families. They found that

children from middle-income homes had greater opportunities for informal literacy

learning (such as visits to the local library, joint book reading, play activities with print)

than children of low-income homes. Low-income parents however, reported more

reading skills practice, such as the use of flash cards and help with homework, with their

kindergarten-age children than did middle-income families. Baker and colleagues'

findings that low-income homes typically offer some opportunities for literacy practice,

though perhaps of a different nature from middle-class families, have been documented

in several ethnographic studies (Gadsden, 1994; Goldenberg, Reese, & Gallimore, 1992).








Although families from lower SES appear to provide their children with some

language and literacy promoting experiences, it is unclear if the nature and quality of

these experiences are adequate and effective. Hart and Risley (1995) studied 42 families

selected to represent the range of American families by race and SES. They gathered

data from monthly observations during unstructured parent-child interactions in the home

over a period of two years Data collection began when each child was under one year of

age and not yet talking and continued until each child reached 3 years of age. The

parents' SES ranged from uneducated and economically disadvantaged to having

advanced degrees and upper income levels. Findings showed substantial variation across

the families with regard to the amount of time spent parenting, the parents' social

interaction with their children, and the content of the language parents addressed to their

children. The most striking finding was the large variability in the quantity of verbal

interaction across these families, and its strong correlation with SES, rate of vocabulary

growth and IQ. In a follow-up study of a few of the low SES children, the cumulative

impact of poor vocabulary growth was strongly linked to poor school performance

through third grade. Hart and Risley hypothesized that since vocabulary is associated

with reading outcomes, it seems likely that reduced opportunities for verbal interaction

would function as a potential risk factor.

The association of poor reading and spelling outcomes with poverty is not clearly

defined and reflects the accumulated effects of several risk factors, including lack of

access to literacy stimulating preschool experiences and to effective, coherent reading

instruction. Children who are likely to have difficulty with learning to read and spell in

the primary grades are those who begin school with less prior knowledge and skill in

relevant domains, most notably overall verbal abilities, phonological awareness, letter








knowledge and familiarity with reading and spelling scripts. Children from poor

neighborhoods are particularly at risk of entering school with weaknesses in these areas

and hence of falling behind from the onset.

With the advent of modern technology, there has been a growing interest in the

relationship between brain anatomy, reading disability and the contribution of

environmental factors such as SES. A longitudinal study by Eckert (1998) examined

reading readiness as a function of family and environmental variables in conjunction with

brain anatomy. Findings indicated that variation in brain anatomy and SES uniquely

predicted the distribution of verbal and phonological aptitude in a group representative of

the general population. Children who had an atypical anatomical pattern on the phoneme

processing area of the temporal cortex (i.e., they had reversed planar asymmetry) and

came from a low SES background were at greatest risk for reading impairment in fifth

grade.

Measures of Spelling Assessment

There are several commercially available tests for the assessment of English

spelling. The most commonly used ones are the Wide Range Achievement Test-3

(WRAT-3) by Jastak and Wilkinson (1993), the Test of Written Spelling-2 (TWS-2) by

Larsen and Hammill (1986) and the Boder Test of Reading-Spelling Patterns (Boder &

Jarrico, 1982). Subtests of spelling are also included in the Kaufman Test of Educational

Achievement (Kaufman & Kaufman, 1985) and the Woodcock-Johnson Tests of

Achievement-Revised (Woodcock & Johnson, 1989). All these tests are standardized

and meet the minimum psychometric criteria. In addition to these standardized spelling

tests, experimental procedures for the assessment of spelling skills can be found in the

research literature, and informal reading/spelling criterion-referenced inventories are








commercially available. These include the Qualitative Inventory of Word Knowledge

(Henderson, 1990; Schlagal, 1992), the spelling subtest of the Early Reading Screening

Inventory (ERSI) (Morris, 1992) and Mann, Tobin and Wilson's qualitative 5-point

scoring system (1987). This section briefly describes these tests, highlighting their major

strengths and weaknesses.

Standardized tests of spelling assessment

The WRAT-3 and the TWS-2 are the two most widely used standardized measures

of spelling assessment. The WRAT-3 is divided into two levels of difficulty based on the

phonetic transparency and orthographic pattern of the target word. The number of words

tested is determined by a method of ceiling to avoid examinee frustration. A major

criticism against this test is the fact that it does not sample the domain of spelling

adequately. No theoretical rationale is provided for the choice of vocabulary. For

instance, the WRAT-3 contains only predictable and variant-predictable words like abuse

and very few words with consonant blends, digraphs, and diphthongs. Words with

inflected endings, contractions, plurals, or homonyms are not tested. The TWS-2 is

divided into two groups of words, "predictable" words and "unpredictable" words

determined by the computer analysis of the Hanna et al. study (1966). Such a grouping

of spelling patterns is said to add content validity to the test. However, as illustrated in

previous sections, a dichotomous classification of words into "predictable/unpredictable"

spelling patterns is oversimplistic and misrepresents the complexity involved in spelling

acquisition. As pointed out by Moats (1994), the inclusion of words such asyes, much,

she in the "unpredictable" list is puzzling. Furthermore, complex linguistic structures

such as variant inflections are notably absent from both lists. A final but equally

important objection to the use of two distinct word lists is the assumption that each set of








words measures independent spelling abilities. In fact, the high correlation value

between the two lists implies that differences in performance on the two lists characterize

subtypes of spelling disability. No solid evidence exists to show that children with

phonological processing deficits do poorly on one or the other type of spelling words

(Stuart & Masterson, 1992).

The Boder Test of Reading-Spelling Patterns (Boder & Jarricho, 1982) is a novel

attempt to identify three subtypes of reading disability. The "dysphonetic" readers have

deficits in areas of phonological decoding. The "dyseidetic" readers have difficulty with

recall of auditory-visual information. The "mixed dysphonetic-dyseidetic" readers

apparently exhibit a combination of phonological decoding and auditory-visual memory

deficits. A key feature of the test involves scoring the percentage of predictable and

unpredictable unknown words on the reading test spelled as good phonetic equivalents.

Test reliability is seriously affected on two accounts. First, determining good phonetic

equivalence poses a major problem. Second, each child takes a different test as

determined by the results of the reading test.

The Kaufman (1985) and the Woodcock-Johnson (1989) tests cover a relatively

broad domain of spelling in the form of word dictation lists. Although the Kaufman test

provides a format for error analysis of prefixes, suffixes, syllable patterns, and vowel

spellings, neither test has sufficient items to measure fine increments in the spelling skills

of kindergartners or first graders.

Experimental tests of spelling development

The Qualitative Inventory of Word Knowledge (Henderson, 1990; Schlagal, 1992) is

a useful tool for determining a child's instructional level for spelling. The inventory

consists of six word lists drawn from a larger pool of words used to construct basal







46
spelling books for each grade level. The words in each list represent both the underlying

structural order in the orthography and the developmental progression of children's

concepts about print. The beginning levels include high frequency Anglo-Saxon words

(e.g., girl, want) while higher levels contain words of Latin and Greek origin which

embody complex relationships between sound, meaning and spelling patterns (eg.,

television, acknowledge, patient). As a general guideline, 90% accuracy on a list

constitutes independent functioning at that level; 50-89% constitutes instructional level

functioning and scores below 49% constitute frustration level functioning.

Scoring systems designed to be sensitive to developmental changes in early spelling

have been presented by several researchers (Lombardino, Bedford, Fortier, Carter &

Brandi, 1997; Mann, Tobin & Wilson, 1987). Mann and colleagues (1987) selected

words that would elicit preconventional phonetic spellings. The stimuli included the

presence of letter names, short vowels, nasals before stops, liquids, consonant digraphs,

and diphthongs. Scoring was based on quality points assigned according to the

progressive improvements children were expected to make on the basis of developmental

theory. A measure of this type given at the end of kindergarten proves to be a strong

predictor of reading ability at the end of first grade (Tangel & Blachman, 1995).

The spelling subtest of the ERSI (Morris, 1992) is sensitive to changes in the

spelling patterns of kindergarten children and is a strong predictor of reading skills in first

grade (Lombardino, Morris, Mercado, DeFillipo, Sarisky, & Montgomery, 1999). The

spelling subtest consists of twelve words that probe a child's ability to represent

consonant and vowel digraphs (chin, back, road, mail), consonant clusters (dress, step,

lamp), short vowels (back, dress), and inflections (picking, peeked). The words were

written upon dictation, and the subtest received a point-score (maximum points of 42)








based on the accuracy and order of graphemes. Lombardino and colleagues (1997)

modified Morris' scoring system and assigned one point to every sound correctly

represented in spelling. In addition, a qualitative error analysis scheme was proposed. In

this scheme, a spelling pattern was identified for a phonetic unit. In other words, the

scheme examined how a sound, represented in conventional orthography by either one

letter or a group of letters, was spelled by kindergartners. The inflectional endings -ed

and -ing were also treated as one sound. To illustrate, the consonant digraph ck in back

constituted a potential error of consonant digraph reduction; the vowel digraphs in road,

mail, feet, constituted a potential error of vowel digraph reduction. A total often spelling

error patterns were identified, and based on the response types of 100 kindergartners, a

rank ordering of the most frequently occurring spelling errors was established. A child's

performance on the ERSI spelling subtest can be analyzed using this scheme and findings

may be compared (although with caution since the ERSI norms are standardized on a

regional sample) to identify prevalent patterns of spelling errors. A second-generation

study, using the same ERSI spelling data (Ahmed & Lombardino, 2000), classified

spelling errors as types of phonological processes such as omission of prevocalic

consonant cluster and substitution of vowel digraph by letter-name or analogy to another

word. Such a classification scheme aims to see if a relationship exists between a

beginning speller's phonemic awareness and knowledge of conventional orthographic

rules. The scheme also brings to the surface the strategies young spellers use to

compensate for their lack of knowledge for conventional spellings. The scheme has been

applied to three spelling levels within kindergarten and typical patterns of spelling errors

have been identified for each level. The assessment was developed to help clinicians

identify kindergartener's level of spelling skills and characterize that child's strengths








and weaknesses in terms of his abilities to translate a perceived sound into its graphemic

representation.

Statement of the Problem

The Need for this Research

The complexity of English orthography is underscored by the fact that many

children and adults with normal intelligence and above average educational backgrounds

never reach mastery level in formal spelling tasks. Confusions between conventional

sequences of letters, such as or , often worry the most competent of spellers and

are not always resolved by modern Spell Check computer aids. The review of the

literature presented in preceding sections of this chapter emphasizes the fact that spelling

is a language-based skill with phonological processing at its core. In addition, studies on

children's earliest attempts at spelling indicate developmental trends that parallel many

developmental language domains, especially the acquisition of phonology

With the current mandate by many educators to teach children spelling as a logical

and organized language-based skill, there is a growing demand for graded instructional

material. Such instructional material needs to be based on developmental spelling

patterns that integrate phonological and orthographic components of spelling.

Consequently, there is an urgent need to focus spelling research on the normal processes

inherent in the acquisition of spelling during the elementary school grades.

Reading disabilities, known to affect at least one out of every five children in the

U.S. (Livesay, 1995), necessarily imply a deficit in spelling skills The earlier a child is

identified with reading and spelling difficulties, the more promising is a positive

prognosis resulting from intervention. Several spelling assessment tools are available but

none seem to capture the structural and developmental features of English spelling in an







49
integrated fashion. The development of such an assessment tool could facilitate the early

identification of children at risk for spelling deficits and provide a developmental

structure for remedial intervention.

The present study is intended to be a step toward identifying developmental spelling

patterns in second through fifth grade children- The findings of the study will provide a

developmental knowledge base for the recognition of spelling patterns in elementary

grade children. This knowledge base is hoped to provide evidence on the process and

acquisition of spelling in children. The development of spelling patterns will indicate

whether children learn via two distinct routes of knowledge, as proposed by the dual-

route model of spelling acquisition (Frith, 1980a), or children exhibit the gradual

development of connections between the two routes at an early stage, as suggested by the

amalgamation theory of spelling acquisition (Ehri, 1992b). Using this qualitative and

quantitative knowledge base, teachers and clinicians can adopt a procedure validated by

research to identify children with deficits in spelling knowledge.

A second equally important purpose of the present study is to design a tool for the

screening/ assessment of spelling abilities in children. A good tool satisfies at least two

basic conditions of design, homogeneity or representativeness of the standardization

sample and construct validity (McCauley & Swisher, 1984). The standardization process

involves administering the tool to a relatively large sample of subjects who are

representative of the population to be assessed, for example, elementary grade children

who exhibit poor spelling skills. If all subjects in the standardization sample, or subsets

of subjects divided into groups of interest such as grade, behave similarly on the tool, the

sample is considered to be homogeneous. Construct validity ensures that the tool

measures the theoretical construct it was designed to measure (Schiavetti & Metz, 1997).








Assuming that the theoretical construct itself is sound, the validity of a tool can be

assessed by testing for consistency of findings across at least two independent data

samples. The design of the present study accounts for homogeneity of the

standardization sample and construct validity by comparing results when the tool is

administered to two randomly selected and independent samples of subjects.


Experimental Questions

The following six experimental questions were addressed in this study:

1. Do children in grades 2, 3, 4, and 5 respectively and assigned to either a data
exploration set or a hypothesis testing data set exhibit similar patterns of
performance on an experimental spelling recognition test designed to assess
phonemic and orthographic knowledge?


2. Does spelling recognition performance show similar patterns of improvement
from 2"n to 5' grades in both the data exploration set and the hypothesis testing
data set?


3. Are there similar between-grade differences in the performance of the three
spelling patterns tested in the data exploration set and the hypothesis testing data
set respectively:
a. words involving phonemic awareness skills only, e.g., "wam" which tests
the short vowel occurring in a closed monosyllable,
b. words involving orthographic rule knowledge only, e.g., "neff' which
tests the doubling of the final letter when it occurs in a closed
monosyllabic word and is preceded by a lax vowel, and
c. words involving a combination of phonemic and orthographic rule
knowledge, e.g., "franceive" which tests the ability to discriminate and
encode a word initial consonant cluster as well as the orthographic rule of
spelling the long vowel in the sequence rather than when it
follows the letter ?


4. Are there similar within-grade differences in performance on the three spelling
patterns in the data exploration set and the hypothesis testing data set?


5. Are there similar distributions of error types at each grade level for the data
exploration set and the hypothesis testing data set?







51
6. Does SES (socio-economic status, as determined by lunch status) have a similar
effect on children's performance on the experimental spelling recognition test in
both the data exploration set and the hypothesis testing data set?


7. Do developmental spelling patterns and individual profiles of poor spellers,
identified by performance on the ESRT, provide evidence in support of the dual-
route model or the amalgamation theory of spelling acquisition?














CHAPTER 2
METHODOLOGY



The purpose of this study was to investigate patterns of normal spelling acquisition

in American English. Data collected from this study were used to answer questions about

the role of phonemic awareness and orthographic rule knowledge in the acquisition of

spelling. This chapter includes information on the subjects, design of the test,

experimental stimuli, administration of the test, scoring, and analysis procedures.


Subjects

Six schools in Gainesville, Alachua County, Florida, participated in this study. Four

of the participating schools were elementary public schools, one school was a private

Catholic school and one school was an experimental laboratory affiliated with the

University of Florida.

A total of 392 children from grades 2, 3, 4, and 5 participated in this study. The

distribution of subjects by grade and SES is shown below in Table 2-1. A subject's SES

was determined solely by lunch status.

Prior to testing any subject, Human Subjects approval was obtained from the

University of Florida IRB. Informed parental consent forms were sent home with each

student in every class. Small token gifts, such as fancy pencils, erasers, stickers, were

given to each child as a reinforcement for taking the parental consent form home,

showing it to the parents, having it signed, and returning it to the class teacher. Only

52








those students who returned signed permission forms were allowed to participate in the

study. Before administering the test, a Group Child Assent Form was read aloud and

only those children who still wished to participate were included in the study. The

Informed Parental Consent and the Group Child Assent forms are reproduced in

Appendix A.



Table 2-1. Distribution of Subjects by Grade and SES.

No. of Subjects % of Subjects on Free or
Reduced Lunch
Grade 2 82 29%
Grade 3 139 30%
Grade 4 91 35%
Grade 5 80 37%'Y




Design of the Spelling Test

An experimental spelling test, referred to as the Early Spelling Recognition Test

(ESRT), was designed to evaluate the phonemic awareness skills and orthographic rule

knowledge American English-speaking children possess in elementary school grades.

The underlying concepts of ESRT were adopted from two experimental spelling

recognition tasks, Olson and colleagues' Orthographic Choice task (1989) and Stanovich

and Siegel's Word Likeness task (1994). The Orthographic Choice task extends beyond

phonological decoding skills and assesses knowledge of word specific spellings. The

task consists of 25 triads of printed stimuli which are all phonologically similar, but

orthographically dissimilar. The examiner reads aloud the chosen word and uses it in a

sentence, while the child being tested is asked to circle the correct word from a choice of

three written items. For example, the child is presented with the choices "tite, tight, tait"






54
in response to the examiner's question: "tight; these shoes are too tight on my feet." The

Word Likeness task by Stanovich and Siegel (1994) comprises of 24 pairs of nonsense

words presented in written form on a sheet of paper. The child being tested is asked to

circle the item that looks most like a real word. The logic behind this task is that children

have to phonologically decode a novel word then select the one that conforms best to

conventions of the English orthographic spelling system.

One form of the ESRT was designed and the same test was given to children in

grades two through five to enable parallel comparisons of achievement. Nonwords that

conformed to standard rules of English pronunciation were selected as test items to

reduce the effect of reading and spelling familiarity that could result from using real

words.

A preliminary version of the spelling test was given to a group of graduate SLP

students in order to elicit their comments with regard to the appropriateness of the non-

words. Comments were studied and amendments were made to eliminate confusions

with the irregular spelling of common words. For example, the nonword "ness," which

targets the floss rule (doubling of final ), was replaced with "neff' since the former

was associated with the spelling of the very frequently used word "yes" rather than

"dress." The revised version of the spelling test was informally administered to a small

group of children from grades 2, 3, 4, and 5. None of the children exhibited any

frustration over the length and nature of the test. Total raw scores suggested an increase

in the number of correct answers as grade level increased.

Experimental Stimuli

The spelling test consists of 35 nonwords, 19 monosyllabic words such as "chint",

and 16 bisyllabic words such as "bolene". All words occur as single items in isolation.








In cases where a morpho-phonemic orthographic rule was tested, e.g., a variant to the

plural -s, the test item was presented in an appropriate context. For example, item #23

targets the spelling of the plural variant when preceded by the sibilant ; this item

was presented in the context: 'Teaches I have one each, you have two feaches".

The test items represent three types of spelling patterns:

1) Phonemic patterns only (PA-Only). This pattern describes nonword stimuli that
requires phonemic knowledge only, as in the test item "warn" which targets the
spelling of the short vowel la/ in a closed syllable;


2) Orthographic patterns only (Ortho-Only). This pattern describes nonword stimuli
that requires orthographic rule knowledge as in the test item "pindle" which
targets the spelling of vocalic /1/;


3) Combined phonemic and orthographic patterns (PA + Ortho). This pattern
describes a nonword stimuli that requires both phonemic and orthographic rule
knowledge as in the test item "quinch" which targets the spelling of the sound
/kw/ and the versus rule.


Of the 35 test items, 10 stimuli represented the PA-Only spelling pattern, 12

represented the Ortho-Only spelling pattern, and 13 represented the PA + Ortho spelling

pattern. The selection of spelling targets was based on findings reported in published

literature (Moats, 1994) as well as clinical reading/spelling protocols for targeting regular

and irregular sound-letter associations. In Table 2-2 below, the 35 test items are

classified into one of the three spelling patterns. For each test item, the target sound

and/or rule is described.

The spelling test is presented in a multiple-choice format (cf Appendix B for a copy

of the complete test). Four options, consisting of one correct answer and three foils, are

given (cf. Table 2-3 for a sample of the test format). While the first four test items are

relatively simple to spell, that is, they target the short vowels and in








Table 2-2. Classification of Test Items by Spelling Pattern.


Zatch
Quinch

Scrain


Rudge
Garlank
Pindle
Toof
Many hods

They moaded

Two feaches

Two tives

Two lastries


Test Item
Wamr

Ling

Neff

Chint
Yuntrude


Zuleps

Bimolt

Snate

Bolene
Glouch
Prath
Plurish
Swother


Spelling Pattern
PA Only

PA Only

PA + Ortho

PA Only
PA + Ortho


PA Only

PA Only

PA + Ortho

Ortho
PA Only
PA Only
PA Only
PA Only

Ortho Only
PA + Ortho

PA + Ortho

Ortho Only
PA + Ortho
Ortho Only
PA Only
PA + Ortho

PA + Ortho

PA + Ortho

Ortho Only

Ortho Only


Description
Short vowel
in closed
syllable
Short vowel in closed
syllable; -ng# blend
Short vowel in closed
syllable; doubled #
digraph; -nt# blend
Short vowel in closed
syllable; VC-e lengthening
rule
Long vowel in open
syllable; -ps# blend
Long vowel in open
syllable; -lt# blend
#sn- blend; VC-e
lengthening rule
VC-e lengthening rule
/au/ diphthong
MISSING IMAGE

Material Information

Title:
The development of phonemic and orthographic spelling patterns a method for assessing spelling knowledge in children in grades two through five
Physical Description:
vii, 155 leaves : ; 29 cm.
Language:
English
Creator:
Ahmed, Sarah T., 1963-
Publication Date:

Subjects

Subjects / Keywords:
Communication Sciences and Disorders thesis, Ph. D   ( lcsh )
Dissertations, Academic -- Communication Sciences and Disorders -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2000.
Bibliography:
Includes bibliographical references (leaves 141-154).
Statement of Responsibility:
by Sarah T. Ahmed.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 022906650
oclc - 45068160
System ID:
AA00013612:00001


This item is only available as the following downloads:


Full Text









THE DEVELOPMENT OF PHONEMIC AND ORTHOGRAPHIC SPELLING
PATTERNS: A METHOD FOR ASSESSING SPELLING KNOWLEDGE
IN CHILDREN IN GRADES TWO THROUGH FIVE














By

SARAH T. AHMED


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA












ACKNOWLEDGMENTS


Special acknowledgment is given to Dr. Linda Lombardino, who served as my

supervisory committee chair on this dissertation. Dr. Lombardino provided me with

constant support and encouragement throughout my four years as a graduate student at

the University of Florida. An outstanding mentor, she initiated me into the mysterious

realm of learning and research. I hope that, in the years to come, she will remember me

with joy and pride.

I would also like to acknowledge the other members of my supervisory committee--

Dr. Patricia Kricos, Dr. Geralyn Schulz, Dr. Bill Williams, and Dr. Christiana Leonard-

for their support and invaluable guidance at every step of my studies.

Special thanks are extended to the staff and students at St. Patrick's Interparish

School, P. K. Yonge Laboratory School, Glen Springs Elementary School, Metcalfe

Elementary School, Rawlings Elementary School, and Prairie View Elementary School.

This study would never have materialized without their cooperation and willingness to

participate.

I would like to extend thanks to all my colleagues and friends, living near and far,

for their words and deeds of love and kindness throughout my years in Gainesville.

Last, but not least, I wish to express my heartfelt thanks to my parents, my three

siblings and their families for their prayers and unswerving trust in me. May God, The

Merciful, The Compassionate, reward them with everlasting happiness!















TABLE OF CONTENTS






ACKNOW LEDGMENTS .................................. ......................... ................................... ii

ABSTRACT.......................................................................... ...... vi

CHAPTERS

1. INTRODUCTION AND REVIEW OF THE LITERATURE..................................1

Introduction................................... ....................................................................... 1
Review of the Literature................................................................ ...................... 3
The Structure of English Orthography........................... .........................3...
Spelling Acquisition....................................................... .........................9...
Theories on the Process of Spelling............................................................. 17
Phonological Processing as the Foundation for Learning to Spell..............22
Schemes for Spelling Analysis........................... ... ......................... 26
Dyslexia and Spelling................................................... ......................... 33
Implications of Socioeconomic Status on Literacy.................................. 40
M measures of Spelling Assessment................................... ....................... 43
Statement of the Problem................... ...................................................... 48
The Need for this Research............................................... ..................... 48
Experimental Questions................................................. ........................ 50


2 M ETHODOLOGY................................................. ........................................... 52

Subjects............................................................................... .......................52
Design of the Spelling Test.............. ........... .. ........................ 53
Experimental Stimuli............................................................. ............................ 54
Procedure......................................................... ................................................. 57
Scoring................................................................................. .......................... 59
Data Reduction and Statistical Analysis..................................... ........................ 59


iii








Testing for Homogeneity of Samples ..................................................... 60
Testing for Normality of Test Score Distribution....................................61
3 R E SU L T S ..................................................................................... ....................... 68

D ata E xploration................................................... ...................... ....................... 68
Grade Performance on the ESRT...........................................................69
Spelling Performance across Grades.........................................................70
Between-Grade Performance on Three Spelling Error Patterns..............73
Within-Grade Performance on Three Spelling Error Patterns.................76
Most Frequently Occurring Spelling Errors Between and Within
G rades on the ESRT................................................ ............................77
Effect of SES on Spelling Performance....................................................80
Hypothesis Derived from Exploration Data..............................................83
Hypothesis Testing D ata............................................................ ........................ 84
Grade Performance on the Spelling Recognition Test.............................84
Spelling Performance across Grades.........................................................86
Between-Grade Performance on Three Spelling Error Patterns..............88
Within-Grade Performance on Three Spelling Error Patterns .................92
Most Frequently Occurring Spelling Errors Between and Within
Grades................... ............................................93
Effect of SES on Spelling Performance................................................96
Sum m ary of Results.............................................................................................. 99
Com bined Data Sets ...................... .. ....................... ......................... 100
Standardized ESRT Scores....................................................................... 100
Individual Profiles of Spelling Performance across Grades....................101

4 D ISC U SSIO N ..................................................... ............................................. 107

Developmental Spelling Performance across Grades........................................ 108
Between-Grade Performance on Spelling Patterns............................................... 11
Within-Grade Performance on Spelling Patterns................................................115
Quantitative and Qualitative Evidence for the Amalgamation Theory..... 116
Individual Spelling Profiles and their Relationship to the
Amalgamation Theory.................................... 119
D design of the ESRT....................... ............................................................ 120
Clinical Applications of the ESRT....................................... ............................125
Strengths and Lim itations................ ....................................................... 127
Strengths.................. ......................................127
Lim itations................................ ......................................................... 128
Directions for Further Research..................................................................129
Summary and Conclusions...............................................................................130

A PPEN D IC E S........................................................... ............................... ................... 132










A PARENTAL INFORMED CONSENT FORM AND GROUP CHILD
ASSENT FORM .................. ............................................................. 132

B EARLY SPELLING RECOGNITION TEST / EXAMINER'S FORM............... 135

C EARLY SPELLING RECOGNITION TEST / ANSWER FORM.....................138

R E FER EN C E S.......................................................... .................................................. 14 1

BIOGRAPHICAL SKETCH.................................................................................. 55














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

THE DEVELOPMENT OF PHONEMIC AND ORTHOGRAPHIC SPELLING
PATTERNS: A METHOD FOR ASSESSING SPELLING KNOWLEDGE
IN CHILDREN IN GRADES TWO THROUGH FIVE

By

Sarah T. Ahmed

May 2000

Chair: Linda J. Lombardino
Major Department: Communication Sciences and Disorders

American English spelling is a complex process that involves integration of letter-

sound associations with morpho-phonemic and orthographic rule conventions. Current

theories on spelling acquisition propose that children initially rely on salient phonological

cues when writing a word. With exposure to print and formal instruction, children

systematically begin to incorporate rules of phonemic awareness and orthographic

knowledge into the spelling process. The development of spelling skills is of interest to

many professionals. The classroom teacher is interested in providing students with the

most beneficial learning environment. The speech-language pathologist is interested in

early identification of children at risk for spelling difficulties and the delivery of effective

intervention.

The purpose of this study was to examine the development of phonemic and

orthographic spelling patterns in U.S. elementary school grades. The Early Spelling








Recognition Test (ESRT) was designed and administered to 390 children in grades 2

through 5. The ESRT consists of 35 mono- and bisyllabic nonwords and targets three

spelling patterns, phonemic decoding only (PA-Only), orthographic rule knowledge only

(Ortho-Only) and a combination of phonemic decoding and orthographic rule knowledge

in the same word (PA + Ortho).

The results of the study revealed clear developmental trends in all three spelling

patterns across the four grades. Spelling performance was examined in two independent

samples and results indicated similar behavior trends. A statistically significant

difference in spelling performance was noted between grades 2 and 4, grades 2 and 5,

grades 3 and 4, and grades 3 and 5. Within each grade, a significant difference in

performance was noted between the PA-Only and the Ortho-Only spelling patterns and

the PA-Only and the PA + Ortho patterns. In support of the amalgamation theory of

spelling acquisition, children exhibited the ability to integrate phonemic and orthographic

rule knowledge as early as in grade 2, with systematic development of the concepts in the

next three grades. Of the three spelling patterns assessed, words requiring orthographic

rule knowledge were the most difficult, with spelling ofmorpho-phonemic variations

(e.g., the suffixes -s and -ed in different linguistic contexts) representing the highest

percentage of errors in all four grades.














CHAPTER 1
INTRODUCTION AND REVIEW OF THE LITERATURE

Introduction


English spelling is a complex process and continues to pose problems even to the

average literate adult. According to the Wide Range Achievement Test-3 (WRAT-3)

normative tables (Jastak and Wilkinson, 1993), the average adult with a high school

education spells at the eighth grade level. Individuals with language-related learning

disabilities make even less progress with spelling despite the advantages of an

appropriate education, average to high IQ, and high socioeconomic status (Moats, 1995).

Several explanations have been suggested to justify the complexity inherent in the

process of spelling. First, English spelling, which requires mapping a phoneme (sound)

onto a conventional grapheme (letter or group of letters), is complicated by the fact that

there are approximately 70 graphemes relative to only 40 phonemes. This means that the

same sound can be represented by more than one letter combination. Second, the

structure of English spelling is influenced by phonological, morphological, syntactic and

semantic factors. It is one's knowledge of the language that allows the speller to

associate words from the same semantic family (e.g., magic, magical, and magician), or

to unravel the spelling mystery of homophones such as 'their, there, and they're.'

Current theories on spelling acquisition propose that children initially rely on

phonological cues when spelling words. Then, very quickly, they incorporate both

phonological awareness and orthographic knowledge into the spelling process when
I








exposed to print and formal instruction. As a result, familiar words are spelled from

memory and semantic association while new words are encoded by phonological

analysis.

For a long time, research and pedagogic communities viewed spelling as a visual

process of rote memorization. Poor spellers were invariably labeled as lazy or

incompetent, and struggled with spelling failure throughout their lives. Attitudes

changed only recently with the discovery that spelling is a language-based process and

difficulties with spelling result in many instances from a core deficit in phonological

awareness. With this paradigm for studying spelling, research has focused on the process

of spelling acquisition in young children, its relationship to dimensions of language

development, and the oral and written language characteristics of individuals with a

spelling disability. The integrity of spelling instruction techniques and spelling

assessment tasks has been investigated in the light of spelling as a language-based

process. Contrary to expectations, learning to spell is a challenging task for the student.

Equally challenging for the researcher in spelling is the goal to understand the process of

spelling, the reasons some students face difficulties in learning to spell, and the

development of reliable and sensitive assessment and instructional techniques.

The purpose of this study was to examine developmental spelling trends across

elementary grades 2, 3, 4, and 5. A secondary goal of the study was to develop a quick,

yet comprehensive spelling assessment procedure that will help to identify children with

significant problems in spelling and to provide direction for appropriate intervention

goals. The following review examines spelling as a language-based process, how its

structure lends to the complexity of English spelling, how children normally acquire and

master spelling skills, and, finally, what characteristics serve to discriminate good and








poor spellers. A review of existing assessment tools will serve to highlight the need for

clinically reliable spelling assessment and therapy techniques.

Review of the Literature

The Structure of English Orthography

Orthography is the writing, or spelling, system of a language. More specifically, it

describes both the implicit and explicit rules a language employs to establish sound-to-

letter correspondences and their interplay with different syllable patterns. The English

writing system is based on an alphabetic code in which each letter of the alphabet

represents a speech sound. In theory, such an alphabetic writing system models a "bi-

unique" relationship so that a given letter corresponds to one and only one sound

(Carney, 1994). Furthermore, this sound is not symbolized by any other letter.

Unfortunately, the English spelling system fails to comply with this simple theoretical

construct. We are all well aware, from our daily writing experiences, that one letter may

represent different sounds as illustrated by the following set of words: (a) gate -4 /gent/,

(b) gender -4 /d,3ndr&/, and (c) beige -) /beI3/. The letter symbolizes the three

sounds /g/, /dz/ and /3 /, respectively. Conversely, one sound may be represented by

different letters. For example, the sound /fl is written with the letter in "fish,"

in "photo," and in "laughter."

A major issue in any study on the structure of English orthography concerns the

extent to which the system is regular and predictable. Since the era of spelling reforms

which first took impetus in the 1I 7 century, the English orthographic system has been

notoriously criticized for its unpredictable inconsistencies (Follick, 1965; Sampson,

1985; Upward, 1988). Concerns for a unified English orthography eventually reached








such a peak that the eminent British playwright and critic, G. B. Shaw, bequeathed part of

his fortune to the cause of spelling reform (Balmuth, 1982)! However, serious research

into the nature of the English spelling system did not begin until the 1960s with

educators' attempts to revise and modernize curricula for reading and spelling instruction

in the public schools. Research was largely influenced by the foundations of

transformational generative linguistics. English orthography was consequently viewed as

a layered structure with deep and surface representations. The components of spoken

language, that is phonetics, phonology, morphology, syntax, and semantics, were

recruited to explain the irregularities found in spelling forms.

Current literature describes English orthography as a structured and predictable

system in which spellings for phonemes, syllables, and morphemes are determined by the

three major factors of (1) rule-governed sound-to-letter correspondences (Hanna, Hanna,

Hodges, & Rudorf, 1966; Hanna, Hodges, & Hanna, 1971; Venezky, 1970), (2) morpho-

phonemic rules (Carney, 1994; Chomsky & Halle, 1968; Moats & Smith, 1992), and (3)

word semantics (Chomsky, 1970; Henry, 1988). A fourth factor, etymology (also known

as the study of word origins), contributes to some extent towards an understanding of

contemporary American and British English spelling patterns (Henry, 1988).

Sound-letter correspondences

According to a computer analysis on the predictability of English spellings by

Hanna and colleagues (1966), almost one half of the words in English can be spelled

accurately on the basis of sound-to-letter correspondences, while another 37% can be

spelled with one error. The computer analysis also accounted for the fact that many

spelling structures are governed by contextual phonetic constraints. As an example, a

vowel letter such as represents either a long or a short vowel depending on the








syllabic structure in which it occurs. Thus, the letter
corresponds to the long vowel

/el/ in the initial open syllable of the word "table," but the short vowel /1/ in the closed

syllable "crab." Such rules, which predict text-to-sound forms, are termed orthographic

rules. Venezky (1970), and more recently, Carney (1994, 1997) performed an in-depth

analysis of text-to-speech correspondences and organized their findings into sets of rules.

A subset of orthographic rules, referred to as "graphotactic rules" (Carney, 1997), serves

to restrict possible letter sequences. For example, words ending in take a final

as in "have, love, move." Also, the letter is always accompanied by a , as in

"queen, quantity." Another subset of orthographic rules, known as "reference rules",

determines spelling variants within a context. For example, the long vowel liV is spelled

, as in "field," except after the letter where it is spelled , as in "receive."

These reference rules, when applied to underlying forms, can predict the spellings of both

vowel alternations found in word pairs like "divine divinity" and consonant variations

as in "sign signature."

Morpho-phonemic rules

A large part of the English vocabulary is derived from morphological rules applied

to root or stem words. The spelling of these derived words is determined largely by

morphological transformations that are applied to a common root or stem. In this way, the

words "please, pleases, pleased, pleasant, pleasurable, displeasing" all belong to the same

family of the stem word "to please"; they all exhibit a common underlying spelling

structure despite variant phonetic representations.

Morphological rules allow the derivation of related words by combining suffixes and

prefixes to roots and stems. Spelling structure and pronunciation are in synchrony when

neutral suffixes such as "-ible" or "-y" are affixed to a stem (e.g., "visible, wisely") since








the stress pattern and the vowel quality of the derived word remains unchanged (Tyler

and Nagy, 1987). Many suffixes, however, are non-neutral because they entail a shift in

the pronunciation of the base word even though the spelling of both base and derived

words remains the same. Such an influence of morphology on spelling is well illustrated

by the phonetic variations of the suffix "-s" which marks either the plural or the 3"

person singular present tense. Depending on the phonetic context of the stem, the suffix

is pronounced as /s/ (e.g., "books, likes"), or /z/ (e.g., "pens, buys") although its spelling

remains consistent. Other examples include variations in vowel quality, as in the word

pair "anxious anxiety." The morpho-phonemic structure of English orthography

consequently often presents a dilemna for spelling (Chomsky, 1970). Although a

consistent spelling of the roots of the words such as "medical, medicine, medicinal" may

be an advantage for reading, it is clearly a disadvantage for spelling. If the speller had no

knowledge of the root, the sounds in the derived words would be ambiguous for spelling

at a surface phonetic level. Similarly, in order to spell the ambiguous second vowel in

the word "confidential," a speller should know that it is derived from "confide" and that

both words follow a regular vowel alternation pattern.

Word semantics

Knowing the underlying morpho-phonemic transformations is not always sufficient

to predict the correct spelling of a word. In English spelling, meaning often takes

precedence over sound because lexical roots and other morphemes are usually spelled

consistently despite pronunciation shifts between the base and the derived forms. In the

following pairs of words, "social society" and "electrical electricity," the shifts in

consonant pronunciation provide misleading spelling cues. However, if the words are

associated with a common semantic and lexical form, the spelling of the consonants








becomes clear. Many derivational patterns involve predictable vowel shifts that are not

represented in spelling, as in the word pairs "serene serenity" and "imagine -

imagination". The chances of spelling such words correctly are higher if judgment is

based on the meaning of the root form and the derivational rules applied to it, rather than

phoneme-grapheme correspondences.

It is believed that both spoken and written words are remembered in relation to other

related words. Also, word meanings are thought to be coded within a network of a family

of words (Nagy et al., 1989). The links between these words consist of inflectional and

derivational relationships. When one word in the family is accessed, the other related

words are also activated for possible retrieval. Furthermore, Nagy and colleagues

demonstrated that connections between words in memory are not only dependent on letter

matching but more on meaning and the sharing of structural elements at the morphemic

level. This theoretical model suggests some important implications for the spelling of

homonyms such as "flour flower," "aloud allowed," and "kernel colonel."

Selection of the appropriate spelling can only be determined by contextual meaning

combined with previous knowledge of word meaning and spelling.

Word origins

Finally, word origins play a fairly significant role in determining the spelling

patterns of many words in English. Although the English writing system is largely based

on an alphabetic code, and the spelling of a high percentage of its lexicon can be

predicted by morpho-phonemic and orthographic rules, there are still many everyday

words with unpredictable spelling patterns. For instance, why is the orthographic rule of

violated in the words "rich, much, such, which"? Or, why is the final

doubled in "dress, lass, moss, watercress" but not in an overused word like "yes"? There








is no scientific explanation to these questions, and some researchers (Hanna et al., 1966;

Hillerich, 1978) have classified such words as "exception" spelling words. Exception

words can only be learned by rote memorization of letter sequences. Henry (1988) on the

other hand, strongly believes that knowledge of word origins facilitates the learning of

exception spelling forms. Numerous historical forces have shaped the development of

written English, with Anglo-Saxon, Romance, and Greek languages playing a role in

establishing the present English lexicon, its pronunciation, and spelling. As described by

Venezky (1970), then categorized by Calfee and Drum (1986), words of Anglo-Saxon

origin do abide by basic letter-sound correspondences. These correspondences, however,

do not always parallel similar sound correspondences found in words of a different

origin. This phenomenon explains in part the occurrence of different letters to represent

the same sound. In this context, the If I sound in the Anglo-Saxon word "ship" is written

with a while the same sound is written with in "mission," a word of Latin

origin. In the same light, words of Greek origin, which include mainly scientific terms,

are characterized by typical letter digraphs to represent a familiar sound. Consequently,

the /f/ sound in the word "finger" is spelled with the letter but the same sound is

spelled with a in words such as "physics, photosynthesis, hemisphere." The same

pattern is noted for the blend /kl/ which is spelled in Anglo-Saxon words like

"clean, class, clown," but in words of Greek origin like "chlorophyll, chloroform,

chloroplasts."

The examples cited above underline the complexity of English spelling which is

influenced by at least four linguistic factors. A word of caution, however, is necessary at

this point. English spelling is not an illogical system that transcends the limits of our

learning abilities. Proficiency in English spelling is a gradual process of learning and








requires an integrated knowledge of all four factors described in earlier sections of this

chapter. With the exception of basic, monosyllabic words, the spelling of most English

lexical items incorporates at least two concepts, that is, the application of sound-letter

correspondences, partially determined by word origins and phonetic context, and

orthographic or morphological rules. In addition, automaticity in spelling develops with

increasing exposure to the printed text, and confusions between spelling choices for

suffixes (e.g., "-tion or -sion"?) and r-controlled vowels (e.g., "-ir, er, ur"?) are

eventually eliminated. The spelling of familiar and overused words becomes established

in memory whereas the spelling of novel words can be analyzed by analogy to words of

similar sound structure.

Spelling Acquisition

English spelling is a complex developmental process that has been studied by

analysis of children's naturalistic writing attempts (Bissex, 1980; Gentry, 1982; Read,

1975; 1986; Treiman, 1993) as well as experimental data (Treiman, 1985a; 1985b; 1985c;

1991; 1994; Treiman, Berch, Tincoff, & Weatherston, 1993; Treiman, Cassar, &

Zukowski, 1994; Treiman, Zukowski, & Richmond-Welty, 1995). According to these

studies, spelling is a continuous process of learning and consequently, it develops in a

broadly predictable sequence of stages between kindergarten and second grade. These

stages of spelling development have been described in several ways. The following

discussion will trace the development of Ehri's model of spelling acquisition (1993)

which emphasizes the link between speech processing and writing.

In a pioneering study, Read (1975) analyzed the spellings of precocious preschoolers

and suggested that the acquisition of spelling is a creative process similar to the learning

of speech. In his analysis, he found that children spell primarily by trying to symbolize








the sounds in words rather than by trying to reproduce memorized strings of letters.

Henderson and Beers (1980) applied Read's analytical framework to first graders'

invented spellings and concluded that first graders use the same spelling strategies

observed in Read's preschoolers. Based on a categorization of spelling strategies, the

researchers proposed three general stages of spelling development: (1) a pre-reading

stage, in which children spell by combining letters and numbers randomly, without

making any sound-symbol correspondences; (2) a letter-name stage, in which children

spell by segmenting the word into sounds and then selecting alphabet letters that contain

those sounds in their names; and (3) a transitional stage, in which children exhibit an

emerging awareness that English orthography is not a fixed one-to-one, sound-symbol

writing code. Based on Bissex's (1980) case study of spelling development "GNYS AT

WRK" and Henderson and Beers' (1980) general scheme for stages of spelling

acquisition, Gentry (1982) posited five stages of spelling development pre-

communicative, semi-phonetic, phonetic, transitional, and correct. These stages reflect

the growth of spelling skills exhibited by a child who moves from the earliest attempts at

writing to the use of conventional spelling forms.

Most recently, Ehri (1991, 1992a, 1993) incorporated components of all previous

spelling stage models and proposed a revised four-stage model of spelling acquisition that

reflects relationships between orthographic, phonological and morphological units

observed in children's invented spellings. Ehri's model stresses the fact that English

orthography has strong phonological underpinnings and children handle the complexities

of the writing system by devising strategies. These strategies, often creative in nature,

define the different stages since they reflect the writer's knowledge of phoneme-

grapheme relationships, orthographic constraints, and grammatical rules such as








affixation and stress patterns. Furthermore, as knowledge of conventional spelling

patterns increases through refinement of phonemic awareness, knowledge of the syntactic

and semantic structure of the language and general exposure to print, early spelling

strategies become suppressed and new strategies emerge to deal with more complex

structures. Although some studies have criticized the validity of stage models as

representative of the normal process of learning to spell (Varnhagen, McCallum, &

Burstow, 1997), Ehri's model of spelling acquisition provides a sound longitudinal

guideline to children's expected performance in spelling tasks (Henderson, 1990).

In Ehri's four-stage model of spelling acquisition, each stage is characterized, and

named, according to the predominant strategies used during that period of acquisition.

During the first stage, called the pre-communicative, or pre-phonetic stage, children

produce scribbles, consisting of randomly selected letters and numbers, to represent

words or sentences. Letters do not correspond to sounds and are not differentiated from

numbers. Thus, tack is spelled as

and muffin as (Ehri, 1993). In many

instances, the written forms of words are meant to represent their meanings such that

whale is spelled with more letters than the names of small objects like mosquito

(Lundberg and Torneus, 1978). The pre-phonetic stage begins very early, as soon as a

child realizes that his speech can be transcribed into a written form.

In the next stage, called semi-phonetic, children learn the names and sounds of the

alphabet letters, and attempt to use this knowledge when selecting letters for their

spellings. Initially, children record only one or two letters corresponding to sounds,

usually occurring in word-initial and final positions. Examples of semi-phonetic

spellings take the form of for pickle, for road, and for feet

(Lombardino, Bedford, Fortier, Carter, & Brandi, 1998). Letter name strategies play a








prominent role in the selection of letters representing a speech sound. As shown by the

above examples, this strategy often yields fairly accurate phonetic representations of a

word. However, the names of a few English letters are misleading guides to sound and

may suggest the wrong spellings. For example, the phoneme /w/ occurs in the name of

the letter but /w/ is never spelled as in English. Kindergarteners and

preschoolers studied by Treiman, Weatherston and Berch (1994) spelled /w/ with a ,

producing spellings such as for went and for wife. Similarly, the affricate

/If/ is heard in the letter name (pronounced as /eitf), thus explaining the spelling

for the word chin.

The semi-phonetic stage is characterized by syllabic representations of salient

consonants and vowels, with strong reliance on letter name strategies. An example of

children's attempts to represent units of syllables rather than individual phonemes is aptly

illustrated in the spelling of final consonant clusters. At this stage, children often fail to

represent the first consonants of clusters. These omissions occur for a variety of

consonants, but especially for nasals like the I/n/ of drink (spelled ) and the

liquids /r/ and /I/ as in warm (spelled ) and cold (spelled ) (Read, 1986;

Treiman, 1993; Treiman, Zukowski, & Richmond-Welty, 1995). These errors arise

because children consider the consonant cluster as one single sound unit and therefore,

represent it with one letter only (Treiman, 1997a).

The reduction of initial consonant clusters forms another case in which children

group together separate phonemes. Children often omit the second and third consonants

in clusters, spelling play and street (Bruck & Treiman, 1990; Treiman,

1985). Children consider the spoken word play to contain the initial consonant unit /pl/

followed by a vowel. Consequently, they symbolize the onset of the syllable with a








single letter rather than analyzing it into two separate phonemes and representing each

phoneme with a letter symbol.

Finally, children's tendency to use units larger than single phonemes when relating

speech into print is reflected in their use of consonant letters whose names (e.g., says

/el/) are similar to the sounds in the word, or part of the word. Examples include

for farmer, for elephant, and for bumblebees

(Treiman, 1997a; Treiman & Tincoff, 1997). The first of apparently

stands for both the vowel and the liquid sounds, which, when combined, constitute the

name of the letter . A similar explanation holds for the spellings of the other two

words. Such errors seem to be more common for the consonants and because

of the close similarity between the pronunciations of their alphabet names and the vowel-

liquid syllable. In particular, children substitute a single letter for the /el/ and /ar/

syllables, as opposed to the /al/ and /Er/ syllables. In other words, it is more likely that a

child will spell the word bell as than ball as because in the first example,

there is a stronger phonological association between the vowel //! and the following

liquid I/l. Given this fact, and given the close link that children in the semi-phonetic

stage form between the syllable /El/ and the name of the letter , the word bell is

spelled . In contrast, the /a/ and /1/ in ball do not form such a strong unit and

children tend to divide the syllable rime into a separate vowel followed by a consonant

(Treiman, 1993, 1994).

The third stage, called phonetic, appears when children learn to segment words into

constituent sounds and to select a letter for every sound. Examples of phonetic spellings

are for quick and for back. In order to produce phonetic spellings,








beginners need to know that a syllable in English necessarily consists of a vowel.

Beginning writers therefore need to know how to symbolize vowels. At the phonetic

stage, children tend to use their knowledge of alphabet letter names creatively to derive

short vowel spellings from the letter name articulated in a similar position. For example,

//l is derived from
, /1/ is derived from , and /U/ is derived from <0>. Long

vowels are represented with one letter with the same name. Diphthongs and back vowels

are often spelled creatively, such as for pile, for boy, and for

boat.

The phonetic spelling stage is driven by an awareness of fine-grained articulatory

phonetic detail that is not symbolized in conventional spelling. Such is the explanation

for the affrication on the initial blends

/0/ digraph
#pl- blend; digraph
#sw- blend;
-4 /6/
digraph
rule
digraph;
rule
cluster;
diphthong rule
rule
Hard ; -nk# blend
Cons-le rule
/u:/ digraph
Short vowel in closed
syllable; plural -s -> /z/
digraph; past tense -
ed -) /id/
digraph; plural -es
- /iz/
-4 with plural
suffix
-) with plural
suffix





Table 2-2. continued

Test Item
S/he vrays

S/he snies

They ploiked

They glayed

They zocked

I was frazing

I was vapping

I am gomping


Two nashes
Franceive


Table 2-3. Sample of the Test Format

1. wem warn worn wim
5. yontrude yuntrude yentrude yuntrud
14. zach zash zah zatch



monosyllabic words, the remaining items are arranged without consideration of the level

of spelling difficulty. Thus, the three spelling patterns occur in a random order.

Procedure

The spelling test was administered as a one-time group activity to children from the

same grade. No more than 30 children were included in a group. The children were

seated either in rows or side by side around a large table in a quiet room within the


Spelling Pattern
Ortho Only

Ortho Only

PA + Ortho

PA + Ortho

Ortho Only

Ortho Only

Ortho Only

PA + Ortho


Ortho Only
PA + Ortho


Description
Diphthong with
suffix -s
Diphthong + suffix -s
-)
diphthong; past tense
-ed /t/
diphthong; past tense
-ed -> /d/
rule; past tense -ed
-) /t/
Drop silent -e with suffix
-ing
Double consonant with
suffix-ing
-mp# blend; -ing suffix
without consonant
doubling
Plural -es 4/&z/
#fr- blend; digraph
before








premises of the school. Individual test forms (cf. Appendix C) and pencils were

distributed to each child. Clear instructions were given, explaining the rules of the

activity. These included:

1) maintain silence;

2) don't look at neighbors' responses since all words are make-believe words with
no fixed spelling forms;

3) watch the examiner's lips.

If there were any questions, the children were instructed to raise their hands and the

examiner or her supervisor would come to help them. Furthermore, the children were

reassured that their performance on this spelling activity would in no way interfere with

their class grades.

The test was administered by the principal investigator and her supervisor. When

needed, a loudspeaker and microphone system was used to amplify the investigator's

voice, gain the children's attention and reduce external distractions. Before beginning the

actual test, two sample test items were illustrated as examples. In the actual test, each

item was read aloud twice with a 5-second interval between readings. Children were

instructed to listen to each word carefully, read all four choices on their paper, then circle

the most appropriate spelling form. A maximum of 5 10 seconds was allowed to answer

each item. The investigator and her supervisor watched the children individually to

ensure that all items were answered and that one choice only was selected for each item.

Complete administration of the test required 30 to 45 minutes, depending on the grade

and attention level of the group. Immediately following the test administration, all test

forms were checked; incomplete forms or those that exhibited a pattern of random answer







59
choice were discarded. From the entire pool of 400 data test forms, a total of eight were

discarded.

Scoring

Raw scores were computed for each test individually after the child's name and

grade had been replaced by a code. Four scores were calculated: (1) a total raw score

consisting of the number of items correctly identified on the test; (2) a subscore

consisting of the number of PA-Only items correctly identified on the test, (3) a subscore

consisting of the number of Ortho-Only items correctly identified on the test; and (4) a

subscore consisting of the number of PA + Ortho items correctly identified on the test.

Table 2-4 summarizes the breakdown of the raw scores for the three subdivisions and

the total items. One point was awarded for each item. The three subscores, which were

of unequal weight, were converted to a possible total of 10 points to allow relative

comparison between the three spelling patterns.



Table 2-4. Distribution of Scores.

Spelling Test Stimuli Number of Items Tested and
Highest Possible score
PA-Only 10

Ortho-Only 12 (converted to a scale of 10)

PA + Ortho 13 (converted to a scale of 10)

Total Test Score 35



Data Reduction and Statistical Analysis

The data were divided into two sample sets, each set consisting of the spelling scores

obtained by subjects from three randomly selected schools. One set of data was used for








data exploration, a statistical method that employs graphical and arithmetical procedures

to observe trends in data behavior, propose hypotheses based on these trends and support

these hypotheses by numerical testing (Rice, 1988). The second set of data was used to

retest the hypotheses proposed at the completion of data exploration and, in case of

agreement in findings, confirm the assumption that children from different schools within

the county exhibit similar patterns of spelling acquisition. Since there was no evidence

that the data were normally distributed in either of the two sets, data analysis was

performed using nonparametric statistics, a method of analysis that does not assume

normal distribution of the sample. Data in both sets were also analyzed using parametric

statistics to compare the power of the two methods of analysis as well as to test if sample

size influences analysis results when the assumption of normality is violated.

Testing for Homogeneity of Samples

The data in each sample set were first tested for homogeneity of variances, a

statistical procedure that allows the researcher to determine whether or not several

samples represent a subset of the same population. In this study, the grade performance

on the Early Spelling Recognition Test was tested for homogeneity across schools by

comparing data observations from one grade and one school with data observations of the

same grade at other participating schools. For example, the sample of scores obtained by

grade 2 in school I was compared to the sample of scores obtained by grade 2 in school 3

and school 6. The same procedure was repeated for grades 3, 4, and 5 respectively for

the two sample sets of data observations. Results of a one-way analysis of variance

(ANOVA) indicated no significant difference (p > 0.05) when performance in the same

grades at different schools, was compared (Tables 2-5 and 2-6). Similar findings of no

statistically significant difference were found when the Kruskal-Wallis test, a non-








parametric measure parallel to the one-way ANOVA, was performed on the same two

data sets (Tables 2-7 and 2-8). Based on these findings for positive homogeneity of

variances, all subsequent data analyses were performed on samples in which data

observations were pooled from the three schools randomly selected for the data

exploration and the hypothesis testing sample sets respectively.



Table 2-5. Homogeneity of Samples by Grade (2-5) and School (1, 3, and 6) Data
Exploration Set

Sum of Df Mean F Sig.
..... ...... Squares __ __ Square
GRADE 2
Between Groups 154.490 2 77.245 2.846 .072
Within Groups 949.852 35 27.139
Total 1104.342 37

GRADE 3
Between Groups 8.297 2 4.149 .131 .878
Within Groups 1362.137 43 31.678
Total 1370.435 45

GRADE 4
Between Groups 150.870 2 75.435 3.733 .085
Within Groups 646.673 32 20.209
Total 797.534 34

GRADE 5
Between Groups 5.859 2 2.930 .142 .868
Within Groups 702.573 34 20.664
Total 708.432 36
Statistical significance at p < .05



Testing for Normality of Test Score Distribution

As a general rule, statistical tests are based on the assumption that the data under

study are normally distributed. If this basic assumption is violated, the statistical test

loses its power, and test results are neither valid nor reliable. Nevertheless, when the









normality assumption is not met, data can be analyzed using nonparametric tests (Sprent,

1989).



Table 2-6. Homogeneity of Samples by Grade (2-5) and School (2, 4 and 5) -
Hypothesis Testing Set


GRADE 2
Between Groups
Within Groups
Total


Sum of Df Mean F
Squares Square

6.345 2 3.172 .253
514.564 41 12.550
520.909 43


GRADE 3
Between Groups 37.967 2 18.983 .956
Within Groups 1330.619 90 19.860
Total 1368.586 92

GRADE 4
Between Groups 42.008 2 42.008 2.348
Within Groups 679.767 55 17.889
Total 721.775 57


GRADE 5
Between Groups 6.017 2 6.017 .521 .477
Within Groups 288.650 40 11.546
Total 294.667 42
Statistical significance at p < .05



Table 2-7. Homogeneity of Samples by Grade (2-5) and School (1, 3 and 6) (Kruskal-
Wallis Test) Data Exploration Set

Chi-Square Df Asymp. Sig
GRADE 2 5.763 2 .056

GRADE 3 .102 2 .950

GRADE 4 4.505 2 .105

GRADE 5 .722 2 .697
Statistical significance at p < .05








Table 2-8. Homogeneity of Samples by Grade (2-5) and School (2, 4 and 5) (Kruskal-
Wallis Test) Hypothesis Testing Set

Chi-Square Df Asymp. Sig
GRADE 2 .556 2 .757

GRADE 3 1.967 2 .374

GRADE 4 .977 2 .323

GRADE 5 .267 2 .606
Statistical significance at p < .05



The data for this study were first tested for normality by graphical and numerical

methods. Box-n-dot plots were drawn to represent the distribution of the pooled total test

scores obtained by each grade (Figures 2-1 and 2-2). A box-n-dot plot (John, 1988, pp.

312-346) is a graphical display that gives an indication of the center of the data

(represented by the middle line in the box and referred to as the median or 50th

percentile), the spread of the data (represented by the horizontal lines above and below

the box, these lines are approximately 1.5 standard deviations above and below the mean

respectively) and the presence of outliers (represented by open circles). The box-n-dot

plot also shows the symmetry or skewness of the distribution of data values by the

positioning of the median within the box. If the median is situated around the center of

the box, the data are assumed to be normally distributed.

Observation of these plots indicated that in most cases, the median was not

symmetrically situated, thus suggesting that total test scores in a grade did not necessarily

have a normal distribution. The Lilliefors Test for Normality (Conover, 1980, p. 357-61)

was performed, and results confirmed the graphical observation that the data of this study

were not normally distributed in all cases (Tables 2-9 and 2-10). The data were










40




30




20



I-
0 10


0

Z o
N= 3 46 35 37
2.00 3.00 4.00 5.00

GRADE


Figure 2-1. Box-n-Dot Plot for Data Exploration Set (Schools 1, 3, and 6).






40




30















N 44 93 50 43
200 300 4.00 500


GRADE


Figure 2-2. Box-n-Dot Plot for Hypothesis Testing Set (Schools 2, 4, and 5).








consequently analyzed using the Kruskal-Wallis test and Wilcoxon's Signed Ranks test,

two nonparametric statistical methods (Conover, 1980, p. 229-37). The analysis was

performed using SPSS Base 7.5 for Windows.



Table 2-9. Tests of Normality Data Exploration Set

Skewness Statistic Df Sig.
Grade 2 .194 .121 38 .172
Grade 3 -.189 .112 46 .132
Grade 4 .398 .156 35 .017*
Grade 5 -.972 .180 37 .010*
statistically significant at p < .05



Table 2-10. Tests of Normality Hypothesis Testing Set

Skewness Statistic Df Sig.
Grade 2 .065 .120 44 .117
Grade 3 -.689 .072 93 .200
Grade 4 -.751 .101 58 .200
Grade 5 -1.129 .174 43 .002*
* statistically significant at p < .05



The Kruskal-Wallis test, which is based on the ranked ordering of observations

across data sets under study, determines whether or not any number of independent

samples of unequal size belong to the same population. In case the independent samples

do not belong to the same population, Fischer's Test of Least Significant Differences, a

follow-up procedure of multiple comparisons, identifies pairs of data sets that are

significantly different (Sprent, 1989). This study involved a number of independent

samples of unequal size that were related in some aspects but were significantly different

in behavior in other aspects. In other words, the data were divided into independent

samples by grade (4 grades), and spelling patterns (3 patterns). The model used








combinations of samples tested for membership in the same population as outlined in

Table 2-11.


Table 2-11. Combinations of Independent Samples in Study

COMBINATION DATA EXPLORATION
EXAMPLE
1. Grade x 3 Schools (Grade 2 x 3 Schools)
(Grade 3 x 3 Schools)
(Grade 4 x 3 Schools)
(Grade 5 x 3 Schools)

2. Spelling Pattern x 4 (PA-Only x Grade 2)
Grades (PA-Only x Grade 3)
(PA-Only x Grade 4)
(PA-Only x Grade 5)

3. Spelling Pattern x 4 (Ortho-Only x Grade 2)
Grades (Ortho-Only x Grade 3)
(Ortho-Only x Grade 4)
(Ortho-Only x Grade 5)

4. Spelling Pattern x 4 (PA + Ortho x Grade 2)
Grades (PA + Ortho x Grade 3)
(PA + Ortho x Grade 4)
(PA + Ortho x Grade 5)


HYPOTHESIS TESTING
EXAMPLE
(Grade 2 x 3 Schools)
(Grade 3 x 3 Schools)
(Grade 4 x 3 Schools)
(Grade 5 x 3 Schools)

(PA-Only x Grade 2)
(PA-Only x Grade 3)
(PA-Only x Grade 4)
(PA-Only x Grade 5)

(Ortho-Only x Grade 2)
(Ortho-Only x Grade 3)
(Ortho-Only x Grade 4)
(Ortho-Only x Grade 5)

(PA + Ortho x Grade 2)
(PA + Ortho x Grade 3)
(PA + Ortho x Grade 4)
(PA + Ortho x Grade 5)


Wilcoxon's Signed Ranks test, a statistical procedure used to explore the

relationship between several related samples, was administered to determine whether or

not a difference exists in the performance on the three spelling patterns within the same

grade. All non-parametric procedures were repeated using parallel parametric methods of

analysis in order to compare results and note whether or not the selection of a statistical

method of analysis contingent on the assumption of normality interferes with








interpretation of results. The Kruskal-Wallis test was matched and compared with the

One-Way-Analysis of Variance while Wilcoxon's Signed Ranks test was matched with

the Paired Samples T-test.

In summary, parametric and nonparametric methods of statistical analysis indicated

that grade performance on the Early Spelling Recognition Test was similar in all

participating schools, thus satisfying the homogeneity of samples condition when

designing and standardizing a tool. Based on these findings of homogeneity of samples

by grade and school, data observations from three schools were pooled to form a data

exploration set and a data testing set respectively. All subsequent statistical analyses

were performed on the two sets of pooled data. Tests of normality indicated that data in

both the data exploration and the hypothesis testing sets were not normally distributed for

all grades. Consequently, all data analyses, reported in the next chapter, were performed

using nonparametric and parallel parametric statistical methods.













CHAPTER 3
RESULTS


This chapter reports the results of the statistical analysis done on the sample of

spelling scores obtained by children in grades 2, 3, 4, and 5 respectively on the Early

Spelling Recognition Test (ESRT). Findings are presented in two sections, (1) data

exploration and (2) hypothesis testing. Within each section, findings are presented

relative to the six experimental questions posed in Chapter 1. First, how do children from

grades 2, 3, 4 and 5 respectively perform on the Early Spelling Recognition Test?

Second, is there a difference in spelling performance across the four grades? Third, are

there between-grade differences in the performance of the three spelling patterns (PA-

Only, Ortho-Only, and PA + Ortho)? Fourth, are there within-grade differences in the

performance of the three spelling patterns? Fifth, what are the most frequent error types

at each grade level? And sixth, does SES influence spelling performance? For each

question, the results of the parametric and non-parametric statistical tests will be

presented side-by-side, followed by a description of the means and medians.

Data Exploration

Data from three schools, selected at random from a group of six schools, were

combined and used to examine patterns of spelling development in grades 2, 3, 4, and 5

respectively. The number of subjects in each grade was unequal: there were 38 subjects

in grade 2, 46 subjects in grade 3, 35 subjects in grade 4, and 37 subjects in grade 5,

resulting in a total number of 156 subjects in the sample set.








Grade Performance on the ESRT

Total scores, defined as the total number of correct responses on the ESRT, were

examined for each of the four grades independently using both parametric and non-

parametric methods of statistical analysis. The distribution of scores in each grade was

studied relative to the mean, median, standard deviation and interquartile range (IQR)

(Table 3-1.) The IQR, defined as half the range of scores that fall between the 25th and

the 75h quartiles, is a measure of the variability of scores from the median value. The

IQR is also approximately equivalent to 1.5 standard deviations from the mean.

Approximate equivalencies between the IQR, standard deviations and percentiles are

summarized in Table 3-2.


Table 3-1. Distribution of Total ESRT Scores by Grade.

Grade 2 Grade 3 Grade 4 Grade 5
N 38 46 35 37
Mean 19.9 20.6 26.3 26.6
Std. Deviation 5.5 5.5 4.8 4.4
Median 19.5 21.5 28.0 28.0
IQR 7.0 9.0 6.5 5.0



Table 3-2. Approximate equivalencies between measures.

Measures on Box-n-Dot plot Measures in Standard Deviations
75 percentile 0.9 SD above the mean
Upper end of IQR (93d percentile) 1.5 SD above the mean
25 percentile 0.9 SD below the mean
Lower end of IQR (7th percentile) 1.5 SD below the mean
Extremes Scores between 1.25 and 1.5 SD above and
below the mean respectively
Outliers Scores 1.5 SD above and below the mean
respectively








In each of the four grades, the mean and the median of the total test scores were

dissimilar yet relatively close in value In grade 2, the mean (19.9) is slightly higher than

the median (19.5), suggesting that the distribution of total test scores is very slightly

positively skewed In each of the grades 3, 4, and 5 respectively, the mean total test

score is consistently lower than the median (x = 20.6, M = 21.5 for grade 3; x = 26.3, M =

28 for grade 4; x = 26.4, M = 28 for grade 5), suggesting that the distribution of total test

scores in each grade is negatively skewed.

Box-n-dot plots were drawn for each grade independently to identify scores that fell

below the 250 quartile and above the 75th quartile (Fig. 3-1). In grade 2, eleven scores

fell below the 25h quartile while eight scores fell above the 75* quartile. In grade 3,

twelve scores fell below the 25th quartile and eight scores fell above the 75"' quartile. In

grade 4, seven scores fell below the 25h quartile while seven scores fell above the 75*'

quartile. In grade 5, ten scores fell below the 25th quartile with two of the scores falling

below the 7' percentile, that is, these two scores were outliers and were greater than 1.5

SD below the mean. In grade 5, nine scores fell above the 75th quartile.

Spelling Performance across Grades

Mean and median total test scores showed that spelling performance on the ESRT

improved consistently across grades 2 through 5 (Table 3-1). This improvement across

grades was supported by a statistically significant Kruskal-Wallis test statistic, x'(3, 156)

= 43.7, p < .05 as well as a one way analysis of variance (ANOVA), F (3, 156) = 19.1, p

< .05 (Table 3-3 and Table 3-4).

Fisher's least significant difference (LSD) post hoc comparisons indicated a

statistically significant improvement in spelling performance between grades 2 and 4,

grades 2 and 5, grades 3 and 4 and grades 3 and 5. No statistically significant













4U







0




o 01200
S(11) (12
1-


) 10
..J

0
N= 38 46 35 37
2.00 3.00 4.00 5.00

GRADE




Figure 3-1. Box-n-Dot Plots for Total Test Scores in each Grade.
Note: Box-n-dot plots show the distribution of scores relative to the median score The
bold line inside the box represents the median score for each of the grades and (x) inside
the box represents the mean score for each of the grades. The upper border of the box
represents the 75h quartile while the lower border represents the 25h quartile. The
horizontal lines extending above and below the box respectively show the IQR, that is,
half the range of scores that fall between the two quartiles. The horizontal line below the
25t' quartile is equivalent to the 7- percentile. The horizontal line above the 75th quartile
is equivalent to the 93' percentile. Both boundaries represent 1.5 standard deviations
from the mean. Numbers in parentheses represent the number of scores that fell below
the 25t1 quartile or above the 75t quartile. Empty circles represent outliers, which are
scores greater than 1. 5 SD above and below the mean respectively.








Table 3-3. Kruskal-Wallis Test Statistics for Spelling Performance across Grades.

Grade N Mean Rank Chi-Square Df Asymp. Sig.
2.00 38 53.4 43.7 3 .000*
3.00 46 59.0
4.00 35 102.6
5.00 37 105.7
* Difference significant at p < .05



Table 3-4. ANOVA for Spelling Performance across Grades.

Sum of Squares Df Mean Square F Sig.
Between Groups 1499.5 3 499.8 19.1 ,000*
Within Groups 3980.7 152 26.2
Total 5480.3 155
* Difference significant at p < .05



improvement in spelling performance was noted between grades 2 and 3 and between

grades 4 and 5 (Table 3-5).


Table 3-5. Multiple Comparisons and LSD between Grades on Total Test Scores.

GRADE GRADE Mean Difference Std. Error Sig.
2.00 3.00 -.78 1.12 .50
4.00 -6.44* 1.19 .00
5.00 -6.78* 1.18 .00

3.00 2.00 .78 1.12 .49
4.00 -5.66* 1.15 .00
5.00 -5.99* 1.13 .00

4.00 2.00 6.44* 1,19 .00
300 5.66* 1.15 .00
5.00 -.33 1.20 .78

5.00 2.00 6.78* 1.18 .00
3.00 5.99* 1.13 .00
4.00 .33 1.20 .78
* Mean difference significant at p < .05








Between-Grade Performance on Three Spelling Error Patterns

Performance on each of the three error patterns, PA-Only, Ortho-Only, and PA +

Ortho, improved from grades 2 to 5, as depicted by mean and median scores (Table 3-6).

This improvement was supported by a statistically significant difference on each of the

three spelling patterns across the four grades (Tables 3-7 and 3-8).



Table 3-6. Descriptive Measures on Three Spelling Patterns by Grade

PA-Only Ortho-Only PA + Ortho
Grade 2 N 38 38 38
Median (IQR) 7.5(1.5) 4.9 (2.5) 5.4(3.1)
Mean (std. Dev.) 7.4(1.5) 4.7(1.9) 5.2(1.9)

Grade 3 N 46 46 46
Median (IQR) 8.0 (2.2) 5.0(3.4) 5.4(3.1)
Mean (std. Dev.) 7.7(1.7) 5.1 (1.9) 5.4(1.9)

Grade 4 N 35 35 35
Median (IQR) 9.0 (2.0) 6.7 (2.5) 7.7 (1.5)
Mean (std. Dev.) 8.7 (1.5) 6.9 (1.8) 7.1 (1.5)

Grade 5 N 37 37 37
Median (IQR) 9.0 (2.0) 7.5 (2.9) 7.7 (2.3)
Mean (std. Dev.) 9.0(1.1) 7.1(2.2) 7.1(1-5)




Table 3-7. Relationship of Performance on Each Spelling Pattern between Grades -
Kruskal-Wallis Test.

PA-Only Ortho-Only PA + Ortho
Chi-Square 29.78 38.11 31.74
Df 3 3 3
Asymp. Sig. .000* .000* .000*
* Difference significant at p < .05









Table 3-8. Relationship of Performance on Each Spelling Pattern between Grades -
ANOVA

Sum of Df Mean Square F Sig.
Squares
PA-Only
Between Groups 69.48 3 23.162 10.664 .000*
Within Groups 330.12 152 2.172
Total 399.61 155

Ortho-Only
Between Groups 176.255 3 58.752 15.823 .000*
Within Groups 564.375 152 3.713
Total 740.629 155

PA + Ortho
Between Groups 123.117 3 41.039 13.204 .000*
Within Groups 472.427 152 3.108
Total 595.544 155
Difference in performance significant at p < .05



Fisher's LSD post-hoc analyses showed that in spelling words of the PA-Only type,

there was a significant difference between grades 2 and 4, grades 2 and 5, grades 3 and 4

and grades 3 and 5. No significant difference for spelling word type occurred between

grades 2 and 3 and grades 4 and 5 (Table 3-9).

Fisher's LSD post-hoc analyses showed that in spelling words of the Ortho-Only

type, there was a significant difference between grades 2 and 4, grades 2 and 5, grades 3

and 4, and grades 3 and 5. No significant difference for spelling word type occurred

between grades 2 and 3 and grades 4 and 5 (Table 3-10).

Fisher's LSD post-hoc analyses showed that in spelling words of the PA + Ortho

type, there was a significant difference between grades 2 and 4, grades 2 and 5, grades 3

and 4 and grades 3 and 5. No significant difference for spelling word type occurred

between grades 2 and 3 and grades 4 and 5 (Table 3-11).









Table 3-9. Between-Grade Multiple Comparisons of Performance on PA only Words.


Grade
2.00



3.00


Grade
3.00
4.00
5.00

2.00
4.00
5.00


5-00 2.00
3.00
4.00
* Difference significant at p < .05


Mean Difference
-.3272
-1.3173*
-1.6316*

.3272
-.9901*
-1.3043*

1.3173*
.9901*
-.3143

1.6316*
1.3043*
.3143


Std. Error Sig.
.323 .313
.345 .000
.340 .000


Table 3-10. Between-Grade Multiple Comparisons of Performance on Ortho-Only
Words


Grade
2.00



3.00


Grade
3.00
4.00
5.00

2.00
4.00
5.00


.00 2.00
3.00
4.00
Difference significant at p < .05


Mean Difference
-.3030
-2.2278*
-2.3394*

.3030
-1.9248*
-2.0364*

2.2278*
1.9248*
-.1116

2.3394*
2.0364*
1116


Std. Error
.422
.451
.445

.422
.432
.426

.451
.432
.454

.445
.426
.454











Table 3-11. Between-Grade Multiple Comparisons of Performance on PA + Ortho
Words.


2.
2.


4.


rade Grade
00 3.00
4.00
5.00

00 2.00
400
5.00

00 2.00
3.00
5.00

00 2.00
3.00
4.00
Difference significant at p < .05


Mean Difference
-.1470
-1.8579*
-1.8606*

.1470
-1.7109*
-1.7136*

1.8579*
1.7109*
-2.7027

1.8606*
1.7136*
2.703


Std. Error Sig.
.386 .704
.413 .000
.407 .000

.386 .704
.395 .000
.389 .000

.413 .000
.395 .000
.416 .995

.407 .000
.389 .000
.416 .995


Within-Grade Performance on Three Spelling Error Patterns

Mean and median scores obtained on each of the three spelling error patterns, PA-

Only, Ortho-Only and PA + Ortho, showed a pattern of difficulty in spelling performance

across all four grades (Table 3-6). All grades performed best on the PA-Only type of

words and most poorly on the Ortho-Only type of words. Scores on the PA + Ortho type

of words were consistently lower than on the PA-Only type of words but slightly higher

than on the Ortho-Only type of words.

Within each grade, performance on the three spelling error patterns showed a

statistically significant difference (for grade 2, X2(2, 113) = 32.9, p < .05; for grade 3, X2

(2, 137) = 50.4, p < .05; for grade 4, X2(2, 104) = 19.4, p < .05; for grade 5, x2(2, 109)=

62.9, p < .05). Both the Wilcoxon Signed Ranks Test and the Paired Samples T-test

showed that for each of the four grades, a significant difference occurred between the








spelling performance of a) PA-Only and Ortho-Only type of words and b) PA-Only and

PA + Ortho type of words. In grade 2, a statistically significant difference also occurred

between the spelling performance of Ortho-Only and PA + Ortho type of spelling words.

In grades 3, 4 and 5, there was no statistically significant difference in scores between the

Ortho-Only and PA + Ortho error types (Table 3-12).



Table 3-12. Differences between Spelling Patterns in each Grade: Wilcoxon Signed
Ranks Test and Paired Samples Test

Wilcoxon Signed Rank Test Paired Samples Test
Grade Pairs Mean Z Asymp. Mean (std. T Df Sig
Rank Sig. Dev.)
2 1 19.89 -5.19 .000* 2.6(1.7) 9.62 37 .000*
2 21.28 -5.12 .000* 2.1 (1.6) 7.96 37 .000*
3 15.86 -2.16 .031* -.5(1.3) -2.31 37 .026*

3 1 23.48 -5.82 .000* 2.6(1.7) 10.52 45 .000*
2 24.95 -5.82 .000* 2.3 (1.7) 9. 15 45 .000*
3 23.08 -.86 .388 -.3 (1.8) -1.21 45 .232

4 1 20.45 -4.61 .000* 1.7(1.7) 5.85 34 .000*
2 18.82 -4.40 .000* 1.6(1.5) 6.13 34 .000*
3 17.66 -.26 .797 -.1 (1.7) -.39 34 .698

5 1 20.05 -4.69 .000* 1.9(1.8) 6.34 36 .000*
2 19.68 -4.79 .000* 1.9(1.6) 7.05 36 .000*
3 22.90 -.12 .904 -5.4(1.6) -.02 36 .984

Pair I = PA-Only Ortho-Only
Pair 2 = PA-Only PA + Ortho
Pair 3 = Ortho-Only PA + Ortho
Paired difference significant at p < .05



Most Frequently Occurring Spelling Errors Between and Within Grades on the ESRT

A list of specific phonemic and orthographic spelling errors was created from the

targets included in the Experimental Spelling Recognition Test (Table 3-13). The








frequency of occurrence of these spelling errors was counted for grades 2, 3, 4 and 5

respectively and the numbers of errors was calculated based on the total number of

possible errors in each grade. The number of errors was then converted into a percentage

of the total number of possible errors for each error type by grade (Tables 3-14 and 3-15).

In all four grades, the highest number of phonemic errors occurred on spelling the long

and short vowel represented by the letter . For example, a common error occurred on

the word "zuleps" where the long vowel sound in the first open syllable was represented

with the letter sequences . Similarly, the short vowel sound in the first closed

syllable of the word "yuntrude" was often represented with the letters or .



Table 3-13. Frequent Spelling Error Targets across Grades.

Errors of Phonemic Awareness Orthographic Errors
1. in open and closed syllables 1. Adding suffix -s to a root
2 vowel digraphs, e.g., 2. Adding suffix -ed to a root
3. consonant digraphs, e.g., 3. -tch/-ch? -dge/-ge? -ck/-k?
4. consonant blends, e.g., <-nt, fr> 4. Syllable type: V-Ce
5. Syllable type: Con-I e
6. Doubling rule
7. /



Table 3-14. Percentage of Phonemic Awareness Error Types by Grade

Error Types Grade 2 Grade 3 Grade 4 Grade 5
in open/closed syllables 47.4% 52.2% 45.7% 40.5%
Vowel digraphs 17.4% 19.1% 9.7% 8.6%
Consonant digraphs 21.7% 13.6% 7.8% 7.4%
Consonant blends 14.0% 21.7% 9.5% 2.7%



In grade 2, a high percent of errors occurred on selection of the appropriate

consonant and vowel digraphs. Fewer errors were made on consonant blends. A similar








pattern of frequency of error types was noted for grade 3. Both digraphs and consonant

blends were problematic spelling targets. In grades 4 and 5, the total percent of errors by

type was relatively reduced. However, in grade 4, an equal percent of errors was made

on digraphs and consonant blends. By grade 5, errors on consonant blends were

markedly reduced, but persisted on vowel and consonant digraphs.


Table 3-15. Percentage of Orthographic Error Types by Grade.

Error Types Grade 2 Grade 3 Grade 4 Grade 5
-y + s -) ies 45% 53% 29% 19%
-Vy + s Vys 15% 20% 17% 26%
-sh/ch + s -) sh/ches 28% 17% 19%/ 15%
-fe + s yves 17% 14% 15% 12%
suffix s -/z/ 12% 17% 08% 12%

Suffix -ed- /t/ 37% 39% 14% 19%0/
Suffix -ed /id/ 18% 12% 7% 7%
-ay + ed -ayed 14% 15% 9% 6%
consonant doubling 33% 41% 25% 28%
drop silent -e 14% 22% 19% 11%

-tch/-ch? -dge/ge? -ck/k? 59% 53% 29% 24%

V Con-e syllable 23% 23% 9% 12%
Con-I e syllable 28% 24% 11% 18%

Doubling s rule 05% 06% 2% 4%

-ie -) ei / 19% 12% 14% 190/%



In all four grades, the most frequent orthographic error occurred on the rule-based

selection of the consonant combinations , and . Other

prominent orthographic errors involved knowledge of spelling rules for adding the suffix

to a root word that ends in the vowel as well as the three phonetic








representations of the suffix . In other words, children from all four grades had

some degree of difficulty understanding that the form of the suffix remains the same

despite differences in pronunciation. As a result, they tended to select the spelling form

that was closest to the phonetic representation of the word, for example, when they heard

"they ploiked yesterday", they selected the spelling form instead of .

This trend in spelling orthographic rule-based patterns from a phonetic standpoint

decreased as grade level increased but did not disappear by grade 5.

Effect of SES on Spelling Performance

One third of the subjects in this data set were on free or reduced lunch status, the

criterion used to divide children into one of two socioeconomic levels. Within each

grade, a similar proportion of subjects were on free or reduced lunch status (Table 3-16).

Descriptive measures showed that subjects on free or reduced lunch status consistently

obtained a lower total test score on the experimental spelling recognition test relative to

their peers who were not on free or reduced lunch (Table 3-16). Both the Kruskal-Wallis

test and an ANOVA supported the observation that the overall spelling performance of

the two groups of subjects was significantly different, X2 (1, 154) = 9.77, p < .05; F (1,

154) = 12.23, p< .05.

However, this difference failed to reach statistical significance for every grade.

Only in grades 3 and 4 did lunch status of a child result in a statistically significant

difference on total ESRT scores, X2 (1, 45) = 5.73, p < .05; F (1,45) = 6.86, p < .05 for

grade 3, X2 (1, 34)= 3.70, p <.05; F (1, 45)= 5.33, p < .05 for grade 4. In grade 2, where

35.8% of the children were on free or reduced lunch, there was no statistically significant

difference in the spelling performance between the two groups, X2 (1, 37) = 0.59, p > .05;

F (1,37) = 0.46, p > .05. Similarly, in grade 5, where 29.7% of the children were on free








Table 3-16. Proportion of Subjects on Free/ Reduced Lunch by Grade and Descriptive
Measures of their Spelling Performance.

Frequency Percent Mean Std. Median IQR
Dev.
Grade 2 .00 24 63.2 20.3 4.8 20.0 7.0
1.00 14 35.8 19.1 6.6 18.5 9.0

Grade 3 .00 27 58.7 22.3 4.8 23.0 5.0
1.00 19 41.3 18.3 5.7 17.0 8.0

Grade 4 .00 25 71.4 27.4 4.2 28.0 5.5
1.00 10 28.6 23.5 5.2 23.5 9.7

Grade 5 .00 26 70.3 27.1 4.1 28.0 3.0
1.00 11 29.7 25.6 5.3 25.0 8.0

All .00 102 65.4 24.3 5.3 25.5 7.2
Grades 1.00 54 34.6 20.9 6.4 21.0 9.2
.00 = paid lunch status; 1.00 = free/ reduced lunch status




Table 3-17. Effect of Lunch Status on Spelling Performance by Grade (Kruskal-Wallis
Test.)

Lunch N Mean Rank Chi-Square Df Asymp. Sig.
Grade 2 .00 24 20.56 .599 1 .439
1.00 14 17.68

Grade 3 .00 27 27.46 5.73 1 .017*
1.00 19 17.87

Grade 4 .00 25 20.10 3.70 1 .045*
1.00 10 12.75

Grade 5 .00 26 19.56 .23 1 .628
1.00 11 17.68

* Difference significant at p < .05








or reduced lunch, there was no statistically significant difference in spelling performance

between the two groups, X2(l, 36) = 0.23, p > .05; F (1, 36) = 0.81, p > .05 (Tables 3-17

and 3-18).



Table 3-18. Effect of Lunch Status on Spelling Performance by Grade (ANOVA.)

Sum of Df Mean Square F Sig.
Squares
Grade 2 Bet Groups 14.08 1 14.08 .46 .500
Within Groups 1090.26 36 30.28
Total 1104.34 37

Grade 3 Bet Groups 184.75 1 184.75 6.86 .012*
Within Groups 1185.68 44 26.95
Total 1370.43 45

Grade 4 Bet Groups 110.88 1 110.88 5.33 .027*
Within Groups 686.66 33 20.81
Total 797.54 34

Grade 5 Bet Groups 16.04 1 16.04 .81 .374
Within Groups 692.39 35 19.78
Total 708.43 36
* Difference significant at p < .05



Descriptive measures also showed that subjects on free or reduced lunch status

consistently obtained lower scores on the three subtests of the ESRT, the PA-Only, the

Ortho-Only, and the PA + Ortho, relative to their peers who were not on free or reduced

lunch (Table 3-19). Regardless of SES, scores on the Ortho-Only spelling pattern were

generally lower than on the PA-Only and the PA + Ortho spelling patterns in all four

grades.








Table 3-19. Descriptive Measures of Performance on three Spelling Patterns by Grade
and SES.

PA-Only Ortho-Only PA + Ortho
Mean Median Mean Median Mean Median
(SD) (IQR) (SD) (IQR) (SD) (IQR)

Grade 2 .00 7.46 7.5 4.86 5.0 5.41 5.4
(1.3) (2.5) (1.7) (1.6) (1.73) (3.1)
1.00 7.21 7.5 4.58 3.75 4.95 5.0
(1.8) (1.75) (2.3) (3.5) (2.15) (4.2)

Grade 3 .00 8.03 8.00 5.63 5.80 5.87 6.20
(1.6) (2.0) (1.50) (1.70) (1.79) (1.50)
1.00 7.21 7.00 4.24 4.20 4.69 4.60
(1.68) (2.00) (2.20) (2.50) (1.95) (3.10)

Grade 4 .00 9.00 9.00 7.24 7.50 7.47 7.70
(1.38) (1.00) (1.73) (2.50) (1.55) (2.30)
1.00 7.90 8.00 6.35 6.70 6.16 5.80
(1.59) (3.25) (1.97) (2.90) (1.63) (3.10)

Grade 5 .00 9.11 9.00 7.44 7.50 7.21 7.70
(1.07) (1.25) (1.68) (1.82) (1.36) (2.30)
1.00 8.72 9.00 6.28 5.80 7.05 7.30
(1.01) (2.00) (2.59) (3.30) (1.78) (2.30)

00 = paid lunch status; 1.00 = free or reduced lunch status



Hypothesis Derived from Exploration Data

Based on the results of the data exploration, the following hypothesis were

formulated:

1. A significant improvement in spelling performance occurs between a) grades 2
and 4, b) grades 2 and 5, c) grades 3 and 4, and d) grades 3 and 5.


2. Within each grade, children obtain the highest score on words of the PA-Only
spelling pattern and the lowest score on words of the Ortho-Only spelling pattern.








3. Within each grade, there is a significant difference in the spelling performance
between a) words of the PA-Only pattern and Ortho-Only pattern, and b) words of
the PA-Only pattern and PA + Ortho pattern.


4. In spelling words of the PA-Only, the Ortho-Only and the PA + Ortho patterns
respectively, there is a significant improvement between a) grades 2 and 4, b)
grades 2 and 5, c) grades 3 and 4, and d) grades 3 and 5.


5. The most frequent phonemic spelling errors in grades 2, 3, 4, and 5 include (a)
representation of the long or short vowel by the letter and (b) spelling vowel
digraphs.


6. The most frequent orthographic spelling errors in grades 2, 3, 4, and 5 include (a)
the rule-based selection of consonant combinations like -tch/-ch, -dge/ge and -
ck/k, and (b) the understanding that a suffix may have several phonetic forms
which, based on rules, may or may not require changes in spelling forms.


7. SES, determined by lunch status, has a significant effect on overall spelling
performance. However, this effect is not consistent across all four grades.



Hypothesis Testing Data

Data from the three schools not included in the Data Exploration were combined and

analyzed to verify the hypotheses formulated from the findings of the Data Exploration.

Patterns of spelling acquisition in grades 2, 3, 4, and 5 were examined using the same

statistical methods applied in the Data Exploration section. The number of subjects in

each grade was unequal: there were 44 subjects in grade 2, 93 subjects in grade 3, 58

subjects in grade 4, and 43 subjects in grade 5.

Grade Performance on the Spelling Recognition Test

The distribution of scores in each grade for the mean, median, standard deviation

and interquartile range (IQR) is shown below in Table 3-20. In each of the four grades,

the mean and the median of the total test scores were dissimilar yet relatively close in








value. In each of the grades 2, 4, and 5 respectively, the mean total test score was

consistently lower than the median (x = 22.5, M = 23.0 for grade 2; x = 23.8, M = 24.5

for grade 4; x = 26.9, M = 28 for grade 5), suggesting that the distribution of total test

scores in each grade is negatively skewed. In grade 3, the mean (22.6) was higher than

the median (22.0), suggesting that the distribution of total test scores is positively

skewed.

Box-n-dot plots were also drawn for each grade independently (Figure 3-2) in order

to identify scores that fell below the 25f" quartile and above the 75th quartile. In grade 2,

six scores fell below the 25b quartile while eleven scores fell above the 75"h quartile. In

grade 3, twenty-two scores fell below the 25t" quartile and twenty-two scores fell above

the 75" quartile. Among the twenty-two scores that fell below the 25d" quartile, four were

outliers. These outliers fell below the lower boundary of the IQR and were at least 1.5

standard deviations below the mean score obtained in grade 3. In grade 4, ten scores fell

below the 25" quartile and eighteen scores fell above the 75"' quartile. Among the ten

scores that fell below the 25"' quartile, there was one outlier with a score that fell below

the lower boundary of the IQR. In grade 5, there were nine scores that fell below the 25d'

quartile and eight scores that fell above the 75"' quartile. Among the nine scores that fell

below the 25"' quartile, two were outliers with scores that fell below the lower boundary

of the IQR


Table 3-20. Distribution of Total ESRT Scores by Grade.

Grade 2 Grade 3 Grade 4 Grade 5
N 44 93 58 43
Mean 22.5 22.6 23.8 26.9
Std. Deviation 3.5 5.2 5.4 4.4
Median 23.0 22.0 24.5 28.0
IQR 5.75 6.0 7.25 5.0








Spelling Performance across Grades

Mean and median total test scores showed that spelling performance on the ESRT

improved from grade 2 through grade 5 (Table 3-20). This improvement across grades

was supported by a statistically significant Kruskal-Wallis test statistic, X (3, 237) = 32.2,

p < .05 as well as a one way analysis of variance (ANOVA), F (3, 237) = 11.2, p <.05

(Tables 3-21 and 3-22).



Table 3-21. Kruskal-Wallis Test Statistics for Spelling Performance across Grades.

Grade N Mean Rank Chi-Square Df Asymp. Sig.
2.00 44 63.1 32.2 3 .000*
3.00 93 78.3
4.00 48 98.1
5.00 43 129.8
* Difference significant at p < .05



Table 3-22. ANOVA for Spelling Performance across Grades.

Sum of Squares Df Mean Square F Sig.
Between Groups 567.5 3 189.2 11.2 .000*
Within Groups 2862.5 234 16.8
Total 3429.9 237
* Difference significant at p < .05



Fisher's least significant difference (LSD) post hoc comparisons indicated a

statistically significant improvement in spelling performance between grades 2 and 4,

grades 2 and 5, grades 3 and 4, grades 3 and 5 and grades 4 and 5. No statistically

significant improvement in spelling performance occurred between grades 2 and 3 (Table

3-23). This finding differs slightly from the results of the data exploration where no



















30( (18) (8)

(11)


W 20 r (7)
r (6) (18) 3
0 0139-00
C (9)
(0 019300


I-- 051o00
0178.00
<0169.00
0
l-- 0,-4 1
N= 44 93 56 43
200 300 400 500


GRADE

Figure 3-2. Box-n-Dot Plots for Total ESRT Scores by Grade.
Note: Box-n-dot plots show the distribution of scores relative to the median score. The
bold line inside the box represents the median score for each of the grades and (x) inside
the box represents the mean score for each of the grades. The upper border of the box
represents the 75th quartile while the lower border represents the 25L' quartile. The
horizontal lines extending above and below the box respectively show the IQR, that is,
half the range of scores that fall between the two quartiles. The horizontal line below the
25th quartile is equivalent to the 7' percentile. The horizontal line above the 75th quartile
is equivalent to the 93"' percentile. Both boundaries represent 1-5 standard deviations
from the mean. Numbers in parentheses represent the number of scores that fell below
the 25'h quartile or above the 75h quartile. Empty circles represent outliers, which are
scores greater than 1.5 SD above and below the mean respectively.








significant difference in spelling performance was noted between grades 4 and 5, a

discrepancy that could result from the nature of the data samples in each set.


Table 3-23. Multiple Comparisons and LSD between Grades on Total Test Scores.

GRADE GRADE Mean Difference Std. Error Sig.
2.00 3.00 -1.18 .16 .155
4.00 -2.89* .93 .002
5.00 -5.54* 1.03 .000

3.00 2.00 1.18 .83 .155
4.00 -1.70* .81 .038
5.00 -4.35* .93 .000

4.00 2.00 2.89* .93 .002
3.00 1.70* .81 .038
5.00 -2.65* 1.02 .010

5.00 2.00 5.54* 1.03 .000
3.00 4.35* .93 .000
4.00 2.65* 1.02 .010
Mean difference significant at p < .05


Between-Grade Performance on Three Spelling Error Patterns

Performance on each of the three error patterns, PA-Only, Ortho-Only, and PA +

Ortho, improved from grades 2 to 5, as depicted by mean and median scores (Table 3-

24). This improvement was supported by a statistically significant difference on each of

the three spelling patterns across the four grades (Tables 3-25 and 3-26).

Fisher's LSD post-hoc analyses showed that in spelling words of the PA only type,

there was a significant difference between grades 2 and 4, grades 2 and 5, grades 3 and 4

and grades 3 and 5. No significant difference for spelling word types occurred between

grades 2 and 3 and grades 4 and 5. These findings are similar to the findings of the

exploration data (Table 3-37).








Table 3-24. Descriptive Measures on Three Spelling Patterns by Grade.

PA-Only Ortho-Only PA + Ortho
Grade 2 N 44 44 44
Median (1QR) 9.0(2.0) 5.4(1.6) 5.8 (2.3)
Mean (std. Dev.) 8.4(1.6) 5.5 (1.5) 5.8(1.6)

Grade 3 N 93 93 93
Median (IQR) 8.0(2.0) 5.8(2.5) 6.2(3.1)
Mean (std. Dev.) 8.1 (1.6) 6.0(1.8) 6.4(1 6)

Grade 4 N 58 58 58
Median (IQR) 9.0(3.0) 5.8 (2.5) 6.2 (2.3)
Mean (std. Dev.) 8.9(1.4) 6.6(1.4) 6.8 (1.7)

Grade 5 N 43 43 43
Median (IQR) 9.0(1.0) 7.5 (2.7) 7.7(2.3)
Mean (std. Dev.) 9.3 (.9) 7.3 (1.6) 7.7 (1.3)



Table 3-25. Relationship of Performance on Each Spelling Pattern between Grades -
Kruskal-Wallis Test.

PA only Ortho only PA + Ortho
Chi-Square 16.04 14.18 27.05
Df 3 3 3
Asymp. Sig. .001* .003* .000*
* Difference significant at p < .05


Fisher's LSD post-hoc analyses showed that in spelling words of the Ortho-Only

type, there was a significant difference between grades 2 and 4, grades 2 and 5, grades 3

and 4 and grades 3 and 5. No significant difference for spelling word type occurred

between grades 2 and 3 and grades 4 and 5 (Table 3-28).

Fisher's LSD post-hoc analyses showed that in spelling words of the PA + Ortho

type, there was a significant difference between grades 2 and 4, grades 2 and 5, grades 3

and 4 and grades 4 and 5 (Table 3-29).








Table 3-26. Relationship of Performance on Each Spelling Pattern between Grades -
ANOVA.

Sum of Df Mean Square F Sig.
Squares
PA-Only
Between Groups 34.29 3 11.431 5.186 .002*
Within Groups 374.74 234 2.204
Total 409.03 237

Ortho-Only
Between Groups 42.09 3 14.029 5.257 .002*
Within Groups 282.89 234 2.669
Total 324.98 237

PA + Ortho
Between Groups 79.09 3 26.362 10.472 .000*
Within Groups 427.94 234 2.517
Total 507.02 237
Difference in performance significant at p < .05



Table 3-27. Between-Grade Multiple Comparisons of Performance on PA-Only Words

Grade Grade Mean Difference Std. Error Sig.
2.00 3.00 -.3051 .300 .310
4.00 -.5190* .336 .041
5,00 -.8752* .374 .020

3.00 2.00 -.3051 .300 .310
4.00 -.8091* .295 .007
5.00 -1.1804* .337 .001

4.00 2.00 .5190* .336 .041
3.00 .8091* .295 .007
5.00 -.3713 .370 .317

5.00 2.00 .8752* .374 .020
3.00 1.1804* .337 .001
4.00 .3713 .370 .317
Significant difference at p < .05








Table 3-28. Between-Grade Multiple Comparisons of Performance on Ortho-Only
Words.

Grade Grade Mean Difference Std. Error Sig.
2.00 3.00 -5.5163 .411 893
4.00 -.9240* .442 .039
5.00 -1.8540* .574 .002

3.00 2.00 5.5160 .411 .893
4.00 -.8688* .389 .027
5.00 -1.7988* .533 .001

4.00 2.00 .9240* .442 .039
3.00 .8688* .389 .027
5.00 -.9300 .558 .099

5.00 2.00 1.8540* .574 .002
3.00 1.7988* .533 .001
4.00 .9300 .558 .099
Difference significant at p < .05



Table 3-29. Between-Grade Multiple Comparisons of Performance on PA + Ortho
Words.

Grade Grade Mean Difference Std. Error Sig.
2.00 3.00 -.3990 .321 .103
4.00 -1.2624* .359 .001
5.00 -2.1807* .399 .000

3.00 2.00 .3990 .321 .103
4.00 -.8842* .315 .000
5.00 -1.2841* .360 .000

4.00 2.00 1.2624* .359 .001
3.00 .8842* .315 .000
5.00 -.4272 .395 .111

5.00 2.00 2.1807* .399 .000
3.00 1.2841* .360 .000
4.00 .4272 .395 .111
* Difference significant at p < .05








Within-Grade Performance on Three Spelling Error Patterns

Mean and median scores obtained on each of the three spelling error patterns, PA-

Only, Ortho-Only and PA + Ortho, showed a pattern of difficulty in spelling performance

across all four grades (Table 3-24). All grades performed best on the PA-Only type of

words and most poorly on the Ortho-Only type of words. Scores on the PA + Ortho type

of words were consistently lower than on the PA-Only type of words but slightly higher

than on the Ortho-Only type of words.

Within each grade, performance on the three spelling error patterns showed a

statistically significant difference for grade 2 (X'(2, 130) = 47.9, p < .05), grade 3 (X2(2,

277) = 50.4, p < .05), grade 4 (X2 (2, 172) = 51.4, p < .05), and grade 5 (2 (2, 127) = 35.2,

p < .05). Both the Wilcoxon Signed Ranks Test and the Paired Samples T-test showed

that for each of the four grades, a significant difference occurred between the spelling

performance of a) PA-Only and Ortho-Only type of words and b) PA-Only and PA +

Ortho type of words. In grades 4 and 5, no statistically significant difference in scores

occurred between the two error types Ortho-Only and PA + Ortho. In grade 2, there was

a statistically significant difference between the error types Ortho-Only and PA + Ortho.

Contrary to the findings of the data exploration, in grade 3 also, a significant difference

occurred between the error types Ortho-Only and PA + Ortho when data were analyzed

using the parametric paired samples test. On the other hand, the nonparametric Wilcoxon

signed ranks test failed to identify a significant difference between the Ortho-Only and

PA + Ortho error types in grade 3 (Table 3-30).








Table 3-30. Differences between Spelling Patterns in each Grade: Wilcoxon Signed
Ranks Test and Paired Samples Test.

Wilcoxon Signed Rank Test Paired Samples Test
Grade Pairs Mean Z Asymp. Mean (std. T Df Sig.
Rank Sig. Dev.)
2 1 20.07 -4.82 .000* 2.6 (2.1) 7.45 43 .000*
2 21.58 -4.96 .000* 2.9 (2.0) 8.60 43 .000*
3 19.83 -1.24 .214 .3(2.0) .90 43 .374

3 1 36.25 -6.00 .000* 2.1 (2.2) 7.98 92 .000*
2 38.09 -5.53 .000* 1.7 (2.0) 6.80 92 .000*
3 36.40 -1.56 .118 -.4 (1.7) -2.10 92 .039*

4 1 22.94 -5.29 .000* 2.4 (1.6) 9.45 57 .000*
2 21.85 -5.06 .000* 2.1 (1.6) 8.05 57 .000*
3 23.17 -.84 .400 -.25(1.6) -1.02 57 .312

5 1 15.61 -4.09 .000* 1.9(1.6) 6.38 42 .000*
2 13.50 -4.36 .000* 1.6(1.3) 6.42 42 .000*
3 12.23 -1.04 .297 -4.0(1.9) -1.09 42 .284

Pair 1 = PA-Only Ortho-Only
Pair 2 = PA-Only PA + Ortho
Pair 3 = Ortho-Only PA + Ortho
Paired difference significant at p < .05


Most Frequently Occurring Spelling Errors Between and Within Grades

A list of specific phonemic and orthographic spelling errors was created from the

targets included in the ESRT (Table 3-13). The frequency of occurrence of these spelling

errors was counted for grades 2, 3, 4, and 5. The number of errors was then converted

into a percentage of the total number of possible errors for each error type by grade

(Tables 3-31 and 3-32). In all four grades, the highest number of phonemic errors

occurred on spelling the long and short vowel represented by the letter . For

example, a common error occurred on the word "zuleps" where the long vowel sound in

the first open syllable was represented with the letter sequence . Similarly, the




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EPQDX4ZQI_JUQAKS INGEST_TIME 2013-03-25T13:48:43Z PACKAGE AA00013612_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES