An Exploration of Early Spelling in Kindergarten Children With Hearing Loss

Abstract

Purpose

The purpose of this study was to determine if group differences exist in spelling accuracy or spelling errors between kindergarten children with hearing loss and children with normal hearing loss.

Method

Participants included 23 kindergarten children with hearing loss and 21 children with normal hearing. All children used spoken English as their primary language, and the children with hearing loss used amplification. Participants completed three single-word spelling assessments, a language assessment, and an oral reading assessment. Spelling was scored holistically and with two linguistic-based scoring systems.

Results

Children with hearing loss did not differ significantly from children with normal hearing in spelling accuracy or linguistic-based spelling error analyses.

Conclusions

The current study provides evidence that children with hearing loss in kindergarten do not differ significantly in their spelling errors compared to children with normal hearing, aside from a lower proportion of mental graphemic representation errors. With these data, in combination with previous research conducted, speech-language pathologists can further individualize treatment to focus on these specific error patterns. Additionally, this focus of treatment can help better prepare children with hearing loss for spelling and writing tasks in later grades. Future research should be conducted to determine when in elementary school the differences in spelling errors are initially seen.

Spelling is a skill typically taught in early elementary school, but learning to spell English words can be difficult because there is often no transparency between phonemes and graphemes, with English using over 170 graphemes/grapheme sets (–tch, –igh, etc.) to represent 44 phonemes (Moats, 1995). Children with hearing loss (CHL) have documented difficulties in spelling in the later grades (e.g., Apel & Masterson, 2015; Bowers et al., 2014); little is known, however, about early spelling skills in this population. The purpose of this study, therefore, was to evaluate early spelling in CHL who use amplification and spoken language.

Theories of Spelling Acquisition

How children with normal hearing (CNH) acquire the skills to learn how to correctly spell has been researched widely (Apel & Masterson, 2001; Gentry, 2004; Henderson & Templeton, 1986; Masterson & Apel, 2010; Treiman & Bourassa, 2000; Wolter, 2017; Wolter et al., 2009). Two prominent theories on the acquisition of spelling skills for children that have been widely studied are stage theory and repertoire theory. When analyzing how children in kindergarten acquire spelling skills, these theories should not be viewed as mutually exclusive; rather, when taken together, stage theory and repertoire theory provide a robust explanation of spelling acquisition.

Stage Theory

According to the stage theory, spelling is acquired through the mastery of discrete stages, starting from preschool and continuing into middle school and beyond (Gentry, 2004; Henderson & Templeton, 1986; Moats, 1995). These stages give insight to the characteristics of children's strategies as they gain mastery and progress in spelling and rely on three principles of the English spelling system: alphabetic, within-word pattern, and meaning (Henderson & Templeton, 1986). First, children recognize that the English spelling system is alphabetic in that the English letters (graphemes) match up to sounds (phonemes) and English graphemes appear in a left-to-right, sequential manner. Next, children master the within-word pattern principle, which explains that the sound that a letter(s) represents depends on the position of the letter(s) in a syllable and the surrounding letters. That is, not only do sounds match to letters, but also these matches can differ based on other features of a specific word. Finally, children master the principle of meaning, which refers to the fact that words or parts of words can have the same or similar meaning being spelled in the same fashion but is dependent on the context. Additionally, children master spelling of morphology in this stage. The successful progression through these developmental stages results in adultlike spellings of words.

Repertoire Theory

Repertoire theory states spelling development consists of the predominant use of a particular process at different points in time and for particular words. Importantly, these processes do not occur at the exclusion of others, unlike the stage theory (Treiman & Bourassa, 2000; Wolter, 2017). According to the repertoire theory, when children spell words, they rely on one or possible more sources of linguistic knowledge to correctly spell any particular word (Masterson & Apel, 2010; Wolter et al., 2009). The degree to which a child uses the different sources of linguistic knowledge can change over time (Wolter, 2017; Wolter et al., 2009).

The sources of linguistic knowledge that a child has in their repertoire to utilize during spelling include phonological awareness, orthographic pattern knowledge, morphological knowledge, and mental graphemic representations (MGRs; Apel & Masterson, 2001; Wolter, 2017). Phonological awareness, or more specifically phonemic awareness for spelling skills, refers to the understanding that words are composed of phonemes that are both separable and manipulable (Wolter, 2017). Orthographic pattern knowledge refers to the ability to translate phonemes to graphemes, which requires the understanding of the general spelling rules and patterns (Apel & Masterson, 2001; Wolter, 2017). Orthotactic principles, or the positional constraints of phonemes to graphemes, are also a part of orthographic pattern knowledge (Apel & Masterson, 2001). Morphological awareness refers to the knowledge that words are composed of smaller, meaningful word parts, or morphemes, and includes the reflection on and manipulation of those morphemes (Wolter, 2017). Additionally, the knowledge of the use of inflectional and derivational morphemes added to base words and how those words are formed involves the use of morphological awareness (Apel & Masterson, 2001; Wolter, 2017). MGRs are words that a specific child has been repeatedly exposed to and have been stored in the child's long-term memory (Wolter, 2017). For correct spelling of an MGR, the word must be clear and complete in the child's memory (Wolter, 2017).

Spelling Errors in CNH

Several studies have explored the spelling abilities of kindergarten children with typical development. Generally, researchers have concluded that kindergarteners' spelling is largely phonetic; that is, spellings at this point in acquisition tend to represent children's knowledge of the phonological structure of words with a lesser impact of morphological and orthographic knowledge, consistent with the stage theory (e.g., Hannam et al., 2006; Treiman, 1985; Werfel & Schuele, 2012). Other findings have indicated that, consistent with the repertoire theory, kindergarten children's spellings can also represent emerging morphological and orthographic knowledge, such as in the case of differential performance on spelling flaps that do and do not represent a stem ending (Treiman et al., 1994). Therefore, spelling in kindergarten is a skill that is emerging but has been shown to rely on a variety of types of linguistic knowledge.

Bahr et al. (2012) analyzed the types of spelling errors in students in Grades 1–9, specifically looking at phonological, orthographic, and morphological errors. The total number of incorrect words spelled decreased with each grade, with the highest decline in Grades 4 and 5. All of the error types were seen in all of the grades with orthographic pattern awareness errors being the most frequently occurring error throughout all of the grades, accounting for 70% of errors across the grades. Errors of both orthographic pattern awareness and phonological awareness decrease from Grades 1 to 9, with orthographic pattern awareness errors decreasing the most in Grade 5 and phonological awareness errors decreasing after Grade 1. Between Grades 4 and 5, errors of morphological knowledge surpassed the amount of errors of phonological awareness. The increase in morphological awareness errors includes the use of inflectional and derivational morphemes; specifically, students in Grades 1–4 had issues with inflectional morphemes and errors with the misapplication of the appropriate suffix with the base word in Grades 5–9.

Spelling Errors in CHL

Relatively, little research has addressed the spelling errors made by CHL. The relatively little research that has been conducted leaves room to pursue answers to lingering questions about the development of spelling skills for CHL and how to provide appropriate spelling intervention for these children.

Previous research on the spelling of CHL suggests that, like CNH, CHL are more accurate at spelling in written stories than spelling dictated words (e.g., Hayes et al., 2011; Straley et al., 2016). Although this effect of writing context appears to affect spelling accuracy similarly across the two populations, substantial evidence suggests that spelling accuracy is lower for CHL than for CNH (Apel & Masterson, 2015; Hayes et al., 2011); however, the range of performance on spelling tasks for CHL is quite large (e.g., standard scores of 45–128; Straley et al., 2016).

In studies that have examined the linguistic basis of spelling errors for CHL, differences from CNH have also been observed. For example, Bowers et al. (2014) reported that middle school students with hearing loss made more phonological and semantic errors rather than orthographic or morphological errors in spelling dictated words. Likewise, Bowers et al. (2016) reported that phonological errors were the most common error type in writing samples of middle school students with hearing loss. This finding for middle school students has been reported for younger elementary students as well. Apel and Masterson (2015) reported that younger children with cochlear implants (M _age = 8;9 [years;months]) were more likely to produce spellings with errors of phonemic awareness than their peers with normal hearing, and Straley et al. (2016) reported that third to sixth graders with cochlear implants were more likely to produce errors of phonological awareness and orthographic knowledge and less likely to produce errors of mental grapheme representations and morphological knowledge than CNH. For CNH, phonological errors of this magnitude are expected early, but not late, in spelling development (e.g., Masterson & Apel, 2010; Moats, 1995). This body of work indicates that CHL continue to produce errors of phonological awareness in spelling much later than expected for CNH.

Therefore, it is clear that CHL produce more errors in spelling tasks than CNH and that development of spelling skills is protracted (e.g., stage theory) or different (e.g., repertoire theory) for CHL. The research to date that has addressed spelling in this population has been conducted primarily with older CHL in late elementary or middle school grades. Little research was completed on the spelling of younger students with hearing loss. Examining spelling earlier in development may help to answer theoretical questions about spelling acquisition in this population. The focus of this study is on the emergence of phonological awareness errors in spelling for kindergarten CHL.

Methods to Measure Spelling

Traditionally, spelling has been measured as either correct or incorrect, creating a score that reflects percentage of holistic accuracy. This method of measuring a child's ability to spell does not evaluate if there is a pattern of spelling errors across a writing sample or the different linguistic knowledge sources that a child may call upon during spelling attempts (Masterson & Apel, 2010). For example, one child might spell the word “whales” as “wales,” and a second child might spell the word “wail.” Both children have spelled the word incorrectly but have pulled from different linguistic sources for their spelling that would need to be further addressed with explicit instruction. The first child showed an error in mental graphemic representation with the omission of the “h,” while the second child showed a semantic knowledge error with spelling a correct English word along with a morphological knowledge error with the omission of the plural “-s.”

Alternate protocols to analyze the spelling abilities of children have been created to analyze how multiple linguistic knowledge sources interact together to support a child's spelling (Masterson & Apel, 2010). Two of the protocols that have been used with CHL include the Spelling Sensitivity Score (SSS) and the Multilinguistic Coding (MLC) system. Both of these systems allow researchers and clinicians to dive deeper into a child's use of linguistic knowledge to spell words. The SSS provides a quantitative score, and the MLC provides a more detailed qualitative analysis.

SSS

The SSS was created and validated to be sensitive to the increases and uses of the different types of linguistic knowledge a child utilizes in spelling (Apel & Masterson, 2015; Masterson & Apel, 2010). The types of linguistic knowledge evaluated by the SSS are phonological awareness, orthographic awareness, morphological awareness, and mental graphemic representation abilities. The SSS divides target words into individual elements, phonemes, juncture changes, and affixes and evaluates a child's spelling at the level of each element (Masterson & Apel, 2010). For example, the word “butter” would be divided as B-U-TT-ER, consistent with the number of phonemes present in the word “butter.” Multimorphemic words are divided into their phonemic elements as shown above; however, the affix and any type of modifications to the juncture are viewed as separate single elements (Masterson & Apel, 2010). For example, the word “runner” would be divided as R-U-N-N-ER to account for the three phonemes in the root word run, the additional “n” that is added when changing the verb “run” to the noun “runner,” and the morpheme –er (Masterson & Apel, 2010).

The SSS utilizes a scaled scoring system from 0 to 3 for determining the accuracy of the elements spelled (Masterson & Apel, 2010). A score of 3 indicates correct spelling of the element, a score of 2 indicates the spelling is incorrect but represented by a plausible or legal spelling, a score of 1 indicates the incorrect spelling of the element with an implausible or illegal spelling, and a score of 0 indicates the omission of the element (Masterson & Apel, 2010).

Two scores are obtained after analysis of all of the elements from the spelling sample, the SSS-Element (SSS-E) and SSS-Words (SSS-W; Masterson & Apel, 2010). The element score for an individual word is calculated by dividing the total element points awarded by the total number of element points possible. The SSS-E is the average of element scores of each word for a given sample of words. The word score is equivalent to the lowest element score in an individual word. The SSS-W is calculated by dividing the number of word points awarded by the total number of words in the spelling sample (Masterson & Apel, 2010). The SSS-E and SSS-W scores illustrate the amount of linguistic knowledge underlying sources in which the child is applying in their spelling (Masterson & Apel, 2010).

MLC

The MLC system was created to provide a more detailed analysis of spelling errors than provided by the SSS (Bowers et al., 2014). Unlike the SSS, which provides an average score for a given spelling sample, the MLC system does not provide scores to interpret the results. Instead, the MLC system provides a detailed qualitative analysis of different linguistic error patterns (Bowers et al., 2014). The MLC system analyzes the child's use of phonological, orthographic, mental graphemic representation, morphological, and semantic knowledge in spelling (Bowers et al., 2014).

To code within the MLC system, the spelling of all attempted words is placed in an Excel spreadsheet and coded for the specific type of spelling error (Bowers et al., 2014). The words are first examined to conclude if all of the phonemes in the word are represented; if a sound is missing or two letters are reversed in a way that changes the sound order in the word (e.g., jup or jupm for “jump”), a phonological awareness error is coded. The next examination is through the representation of spelling patterns or orthographic errors. This is examined through two error codes; if a phoneme is represented with an illegal or phonologically implausible substitution (e.g., flas ”flash”), an orthographic pattern knowledge error is coded, and if a phoneme is represented with a legal or phonologically plausible substitution (e.g., ake for “ache”), a mental graphemic representation error is coded. Errors in bound morphemes or errors in modification of the root word are coded as morphological awareness errors (e.g., vacatian or vacateion for “vacation”). Words that are spelled incorrectly for the target word but represent a different English word are coded as semantic awareness errors (e.g., night for “nine,” ate for “eight”).

Bowers et al. (2014) examined which coding method, the SSS or the MLC system, provided the information that educators need to develop better spelling interventions based on the types of linguistic errors made by CHL. They concluded that, for middle school students, the MLC provided a more complex and in-depth analysis of a child's spelling abilities than the SSS, thus giving more information about the spelling patterns of CHL (Bowers et al., 2014). It is unclear if the same pattern of results would be true for early spelling measurement (e.g., kindergarteners).

Purpose of the Current Study

Adultlike spelling requires different types of linguistic knowledge sources to interact with each other to correctly spell the word (Apel & Masterson, 2015; Masterson & Apel, 2010). Errors in spelling can give insight into children's linguistic abilities; however, spelling errors of young CHL have not been widely researched. Determining whether and what differences in spelling performance and/or errors occur for CHL compared to CNH could provide direction for teachers and speech-language pathologists in choosing goals and teaching strategies. Additionally, establishing baseline spelling skills and possible differences in kindergarten is a first step in determining when during schooling differences may begin to emerge. In this study, we set out to explore the different types of spelling errors CHL exhibit during kindergarten compared to their peers with normal hearing on a single-word spelling assessment and to compare three types of error analyses: percent correct, the SSS, and the MLC system. Because previous research has not specifically addressed spelling in kindergarten for CHL, we did not have a priori hypotheses. The specific research questions addressed were as follows:

1. Do kindergarten CHL exhibit more spelling errors than CNH?

2. Do CHL produce more unanalyzable spelling attempts or no spelling attempts compared to CNH?

3. Do kindergarten CHL differ on the SSS compared to CNH?

4. Do kindergarten CHL exhibit different proportion of error types using the MLC system than CNH?

Method

This study was approved by the University of South Carolina Institutional Review Board as part of a larger longitudinal study, the Early Language and Literacy Acquisition study (Werfel, 2017).

Participants

Forty-four participants completed testing: 23 kindergarten CHL (M _age = 6;1, SD = 0;3, range: 6;0–6;4), nine boys and 14 girls, and 21 kindergarten CNH (M _age = 6;1, SD = 0;1), 10 boys and 11 girls. Age did not differ between groups (p = .873, d = −0.05). Participants were from states across the South, Northeast, and Midwest regions of the United States.

The CHL were diagnosed with permanent bilateral hearing loss by an audiologist, used amplification and spoken language, and did not have additional diagnoses known to affect language and literacy acquisition (e.g., autism, Down syndrome). Ten children in this group used bilateral cochlear implants, eight used bilateral hearing aids, three children were bimodal (one cochlear implant, one hearing aid), and two children used bone-anchored hearing aids. All CHL previously had received speech and language services; six were no longer in speech and language therapy at the age of 6 years, per parent report. Average age of identification was 10.6 months (SD = 15.86, range: 0–54 months), and average age of amplification was 14.4 months (SD = 15.76, range: 0–60 months). The average time from identification to amplification was 3.3 months (SD = 3.68, range: 0–12 months).

CNH passed a bilateral hearing screening prior to beginning the study and had no diagnoses known to affect language and literacy acquisition. No CNH had received language therapy, though two participants received prior services for speech sound disorder, per parent report.

All participants had nonverbal intelligence within the average range, measured by the Primary Test of Nonverbal Intelligence, during their initial testing session (Ehrler & McGhee, 2008). All participants spoke English at least 70% of the time at home and used spoken English during the testing session. English spoken at home did not differ between groups, p = .45 (CHL: M = 95.33%, SD = 8.34, range: 75–100; CNH: M = 97.45%, SD = 7.85, range: 70–100). Maternal education did not differ between groups and was generally high (e.g., bachelor's degree on average; CHL: M = 15.7, SD = 2.8, range: 12–22 years; CNH: M = 16.9, SD = 2.3, range: 12–22 years; p = .15). Table 1 displays additional demographic information.

Table 1.

Demographic information.

Group	Race	Ethnicity	Other languages spoken at home (< 30% of the time)
CNH	White – 18 African American – 0 Asian – 1 Asian/White – 1 Native Hawaiian/Pacific Islander/White – 1	Latinx or Hispanic – 0	Mandarin (Chinese) Ukranian Farsi
CHL	White – 17 African American – 4 Asian – 1 Asian/White – 0 Native Hawaiian/Pacific Islander/White – 0	Latinx or Hispanic – 6	Spanish Albanian American Sign Language Filipino Russian

Note. One CHL participant's parents did not report race or ethnicity. CNH = children with normal hearing; CHL = children with hearing loss.

Procedure

As part of a larger study, participants' language was assessed using the Comprehensive Assessment of Spoken Language–Second Edition General Language Ability Index score (CASL-2 GLAI; Carrow-Woolfolk, 2017). Participants' reading was assessed using the Woodcock Reading Mastery Test–Third Edition (WRMT-III; Woodcock, 2011). Three single-word spelling assessments were completed by the children in the study, the Test of Written Spelling–Fifth Edition (TWS-5; Larsen et al., 2013), the Phonological Awareness Literacy Screening–Kindergarten (PALS-K; Invernizzi et al., 2004), and SSS-Kindergarten/Grade 1 Spelling List (SSS-Spelling List; Apel et al., 2017). After completion of the testing, the assessments were initial- and double-scored by trained graduate research assistants who worked on the Early Language and Literacy Acquisition study. Measures were administered in a randomized order for each participant, and administration followed published administration protocols.

Language, Reading, and Spelling Measures

CASL-2

The CASL-2 is an evaluation of an individual's oral language skills based on the Integrative Language Theory for ages 3;0–21;11 (Carrow-Woolfolk, 2017). Participants completed eight subtests—Receptive Language, Synonyms, Expressive Vocabulary, Sentence Expression, Grammatic Morphemes, Sentence Comprehension, Inference, and Pragmatic Language—to contribute to the General Language Ability Index (GLAI) score, which serves as a general measure of an individual's spoken language ability (Carrow-Woolfolk, 2017). Forty-two of the participants completed the CASL-2; two participants completed the CASL. ¹ Interrater reliability ranges from .86 to .97, with a median agreement of .92 (Carrow-Woolfolk, 2017).

For Receptive Vocabulary, participants were presented with four pictures and instructed to choose the picture that represented what the examiner stated. For Synonyms, the examiner read a word plus four choices and instructed the participants to verbally choose the best answer. For Expressive Vocabulary, participants were presented with a sentence with or without picture support and were instructed to verbally complete the sentence with one word. For Sentence Expression, the examiner presented a picture to the participant and instructed the participant to finish a sentence with a word or phrase or explain a picture. For Grammatic Morphemes, the examiner stated a sentence and instructed the participant to decide if the sentence needed to be fixed and then fix the sentence or if the sentence was grammatically correct with or without picture support. For Sentence Comprehension, the participant was presented with four pictures and was instructed to point to the picture that matched what the examiner stated. For Inferences, the participant was instructed to deduce information from a sentence or phrase the examiner stated with or without picture support. For Pragmatic Language, the examiner read a scenario in which the participant was instructed to respond to a question about the scenario with or without picture support.

WRMT-III

The WRMT-III is a comprehensive battery of nine assessments to measure reading readiness and reading achievement (Woodcock, 2011). Participants were given five subtests—Word Identification, Word Attack, Word Comprehension, Passage Comprehension, and Oral Reading Fluency—to contribute to the Total Reading Score, which serves as a broad measure for a participant's reading comprehension skills. Forty-two participants completed the WRMT-III; two participants did not complete the WRMT-III due to time constraints. Reliability is .87–.95.

For Word Identification, participants were instructed to read aloud English words of increasing difficulty. For Word Attack, participants were instructed to read aloud nonsense words of increasing difficulty. Word Comprehension consists of three subsections: Antonyms, Synonyms, and Analogies. For each, single English words were presented to the participant. Participants were instructed to read the word aloud and give a single word response per subsection, for example, “hot” as the antonym for “cold.” For Passage Comprehension, participants were instructed to read a short passage and fill in the blank with one word. For the Oral Reading Fluency task, participants were asked to read Passage A, which contains 80 words, in a natural, reading voice that was not rushed while timed.

TWS-5

The TWS-5 is a norm-referenced test of spelling with administration using a dictated word format (Larsen et al., 2013). All participants started at Item 1, and administration was ended when a child incorrectly spelled five words in a row, per TWS-5 administration protocol (Larsen et al., 2013). Reliability is .93.

PALS-K

The PALS-K is a criterion-referenced screening tool that can be used by teachers to determine if a child is above or below benchmark for skills learned in kindergarten (Invernizzi et al., 2004). For administration, the examiner read five consonant–vowel–consonant words to the child one at a time and the child was instructed to write their response in the space provided. Responses were scored per the scoring procedures. Each grapheme was given a score of 1 if the grapheme was written regardless of kinetic reversals, error of order; for example, “ten” would receive 3 points, and “net” would receive 3 points. A bonus point was awarded for correct spelling of the word. Static reversals, writing the mirror image of a single letter, are not considered to be an error per scoring protocol (Invernizzi et al., 2004). All participants completed the PALS-K spelling measure.

SSS-Kindergarten/First Grade Spelling List

The SSS-Spelling List is a list of 25 words varying in elemental units and morphological units appropriate for kindergarten/first grade (K/1) level (Apel et al., 2017). The SSS-K/1 Spelling List can be found in Appendix A. For administration, the examiner read each word aloud to the child and used the word in a sentence, and then the word was read once more. The child was instructed to write their response in the space provided on the record form. The responses were scored in two ways: (a) as correct or incorrect and (b) analyzed using the SSS. All participants completed the SSS-K/1 Spelling List.

Analysis

All measures were first scored according to published instructions. Standard scores were calculated when available. After scoring the standardized assessments, the following procedures were used for the SSS-K/1 Spelling List.

Percent Correct

First, percent correct was computed for the SSS-K/1 Spelling List (Apel et al., 2017). Each correct spelling received a score of 1, and each incorrect spelling received a score of 0. The number of correct spellings was divided by 25 and multiplied by 100 to determine the percentage of correct spellings of the words for each child.

SSS

Next, the researcher created an Excel spreadsheet for each participant for the SSS-K/1 Spelling List (Apel et al., 2017). The Excel spreadsheet included the target spelling of each word in Column A and the participant's spelling of the word in column B. A code created by the researcher was used to denote a response as nonanalyzable (XXX) and for no attempt.

The Excel spreadsheet for each participant then was individually uploaded and run through the Computerized Spelling Sensitivity System (Masterson & Hrebec, 2011). The Computerized Spelling Sensitivity System (C-SSS; Masterson & Hrebec, 2011) populated the SSS-E and the SSS-W scores based on the SSS procedures outlined above (Masterson & Apel, 2010). A dictionary internal to the C-SSS was utilized, which included legal and illegal substitutions for each of the word elements. If the word was not included in the dictionary, the researcher broke the word into elements per the SSS procedures (Masterson & Apel, 2010). Words that were deemed nonanalyzable or no attempt were made to spell the word were not included in the SSS analysis. As previously described, the element score for an individual word is calculated by dividing the total element points awarded by the total number of element points possible. The SSS-E is the average of element scores of each word for a given sample of words. The word score is equivalent to the lowest element score in an individual word. The SSS-W is calculated by dividing the number of word points awarded by the total number of words in the spelling sample (Masterson & Apel, 2010). The SSS-E and SSS-W were averaged for each participant. The word score of each attempted word was used to further analyze the specific nature of the group differences. The word score was used over the element score because each word was given a whole number (0, 1, 2, 3) while the element score gave partial numbers (0.15, 2.73, etc.). Each participant was given a participant ID, and each word was given a word ID (1–25) to be used in a mixed-effects model analysis, described below.

MLC

The researcher created a worksheet from an MLC template (Werfel, 2016) for each participant for the SSS-K/1 Spelling List. The worksheet included the target spelling of each word in the first column and the participant's spelling of the word in the second column. The worksheet included six columns for possible types of errors: phonological awareness, orthographic pattern knowledge, mental graphemic representation, morphological knowledge, semantic knowledge, and “other.”

The coding of the spelling words followed a coding manual adapted from Bowers et al. (2014). To be deemed a phonological awareness (PA) error, there must be an omission and/or addition of a phoneme to a word (brandsh for brandish) or letter transpositions that change the sound order in a word (engotiate for negotiate). For an error to be one of orthographic pattern knowledge (OPK), there must be an illegal phoneme-to-grapheme correspondence (salut for salute) or an illegal positional constraint of a grapheme (ckollar for collar). A mental graphemic representation (MGR) error includes a legal but incorrect phoneme-to-grapheme representation (opake for opaque) or letter transpositions that do not change the sound order in a word (abel for able). A morphological knowledge (MK) error includes the omission, addition, or incorrect spelling of an affix (zealus for zealous) or the omission, nondropping, or incorrect spelling of a juncture (continueity for continuity). Semantic knowledge (SK) errors include a correct spelling of a homophone for the target word (night for knight), correct spelling of a real word that is not the target word (hug for rug or sun for fountain), or the correct spelling of an alternate verb tense for the target word (know for knew). The type of errors given the “other” code includes the use of nongrapheme elements in the spelling (nine-teen for nineteen) or no spelling attempt made of the word. Each word receives an error code for each error in the word; words can have multiple codes.

The first author coded each worksheet the coding manual. The sum of the errors was calculated and recorded at the bottom of each sheet. The total number of errors produced in each category were added together to calculate the total number of errors produced overall. To compute the proportion of each category, the sum of the individual category errors was divided by the total number of errors recorded. For example, to calculate the proportion of PA errors, the total number of PA errors was divided by the total number of errors the child produced overall.

Statistical Methods

To address Research Question 1 to examine if CHL exhibit more spelling errors compared to CNH, an independent-samples t test was completed using the calculated percent correct from the SSS-K/1 Spelling List. To address Research Question 2 to determine if CHL produce more unanalyzable spelling attempts or no spelling attempts compared to CHL, an independent-samples t test was completed using the total number of unanalyzable or not attempted spellings was completed using the SSS-K/1 Spelling List responses. To address Research Question 3 to examine the differences between CHL and CNH using SSS, an independent-samples t test was completed to compare the overall SSS-E and SSS-W mean scores. Additionally, a mixed-effects model comparison was completed using the individual SSS word scores to examine deeper into the specific nature of group differences in the language skills used with word spelling. To address Research Question 4 to determine if CHL exhibit different proportion of error types using the MLC system, a Mann–Whitney U was completed. Additionally, correlational analyses were completed to determine how correlated the SSS-Spelling List (Apel et al., 2017) scores were compared to the TWS-5 standard scores and the correlation between the TWS-5, WRMT-III Oral Reading Fluency, and CASL-2 GLAI standard scores.

Reliability

All spelling, language, and reading assessments were initial- and double-scored by trained graduate research assistants. The interrater reliability was 100% for each spelling, language, and reading assessment. Interrater agreement was conducted for both the SSS and MLC scoring. A second-year graduate student in speech-language pathology double-scored 30% of the SSS data for reliability. A doctoral student who is a certified speech-language pathologist double-scored 30% of the MLC data for reliability. Intraclass correlation was used to calculate agreement. Agreement for SSS scoring was excellent, .992 overall: .986 for SSS-W and .998 for SSS-E. Agreement for the MLC coding was also excellent, .890 overall: range of .839–.919 across categories.

Results

Table 2 displays means and standard deviations of the standardized language, reading, and spelling measures for each group. During the 6-year-old year, we observed significant group differences for spoken language scores, and Cohen's d effect sizes indicated a large effect of hearing loss status on overall spoken language scores. Significant group differences were not observed for reading or spelling measures in kindergarten; however, effect sizes indicated small to medium effects of hearing loss status on literacy skills.

Table 2.

Descriptive statistics of standardized study measures.

Measures	CHL M (SD)	CNH M (SD)	t	p	d
CASL-2 GLAI	91.65 (16.98)	110.81 (11.48)	4.341	< .001	1.31
WRMT-III total score	105.71 (17.44)	114.14 (16.67)	1.601	.117	0.49
TWS-5	94.09 (15.07)	101.10 (17.49)	1.409	.166	0.43
PALS-K Developmental Spelling a	14.48 (7.02)	17.38 (3.72)	1.734	.092	0.51

Note. CHL = children with hearing loss; CNH = children with normal hearing; CASL-2 = Comprehensive Assessment of Spoken Language–Second Edition (Carrow-Woolfolk, 2017); GLAI = General Language Ability Index; WRMT-III = Woodcock Reading Mastery Test–Third Edition; TWS-5 = Test of Written Spelling–Fifth Edition; PALS-K = Phonological Awareness Literacy Screening–Kindergarten.

^a Raw score; max score = 20. All others are standard scores.

Correlational analyses were completed to determine how spelling was related to language and reading skills across groups. Using Pearson correlations, the TWS-5 standard scores of each group were correlated with the SSS-W and SSS-E overall scores, the CASL-2 GLAI standard scores, and the WRMT-III Total Reading standard scores. The TWS-5 standard scores were found to be highly correlated with the SSS-W overall scores, SSS-E overall scores, CASL-2 GLAI standard scores, and WRMT-III Total Reading standard scores. Table 3 displays the significance values by assessment and group.

Table 3.

Test of Written Spelling–Fifth Edition correlations with other study assessments.

Variable	SSS-W	SSS-E	CASL-2 GLAI	WRMT-III total reading
CNH	.94**	.83**	.62**	.92**
CHL	.85**	.80**	.49*	.77**

Note. SSS-W = Spelling Sensitivity Score-Words; SSS-E = Spelling Sensitivity Score-Element; CASL-2 = Comprehensive Assessment of Spoken Language–Second Edition; GLAI = General Language Ability Index; WRMT-III = Woodcock Reading Mastery Test–Third Edition; CNH = children with normal hearing; CHL = children with hearing loss.

^* p < .05.

^** p < .01.

Research Question 1: Difference in Number of Spelling Errors

When analyzing the accuracy of spelling using percent correct, CNH and CHL did not differ statistically on the SSS-K/1 Spelling List (see Figure 1). CNH spelled on average 24.95% of words correct (SD = 27.68, range: 0%–88%), whereas CHL spelled on average 15.83% of words correct (SD = 18.27, range: 0%–56%). This difference was not statistically significant (p = .200), and Cohen's d effect size was 0.39, indicating a small effect of group on spelling accuracy.

Figure 1.

Group comparison of percent correct on SSS-K/1 Spelling List. Solid line denotes the median. SSS = Spelling Sensitivity Score; K/1 = kindergarten/first grade; CNH = children with normal hearing; CHL = children with hearing loss.

Research Question 2: Difference in Unanalyzable Attempts/No Spelling Attempts

CNH and CHL did not differ significantly on production of the number of unanalyzable attempts or no spelling attempts on the SSS-K/1 Spelling List. CNH had a mean of 0.24 words that were unanalyzable or not attempted (SD = 1.09), and CHL had a mean of 1.57 words (SD = 5.32). This difference was not statistically significant (p = .296), and Cohen's d effect size was 0.34, indicating a small effect of group on unanalyzable or not attempted words.

Research Question 3: Difference with SSS

SSS–Overall Word Score

When comparing the means from the SSS-W between CNH and CHL, scores between the groups did not significantly differ (see Figure 2). CNH had a mean SSS-W of 1.36 (SD = 0.75, range: 0–2.80), while CHL had a mean SSS-W of 1.12 (SD = 0.62, range: 0–2.12). This difference was not statistically significant (p = .259), and Cohen's d effect size was 0.35, indicating a small effect of group on spelling accuracy.

Figure 2.

SSS-Words scores by group. Solid line denotes the median. SSS = Spelling Sensitivity Score; CNH = children with normal hearing; CHL = children with hearing loss.

SSS–Overall Element Score

When comparing the means from the SSS-E between CNH and CHL, scores between both the groups did not significantly differ (see Figure 3). CNH had a mean SSS-E of 2.12 (SD = 0.58, range: 0.59–2.93), while CHL had a mean SSS-E of 1.81 (SD = 0.72, range: 0.33–2.64). This difference was not statistically significant (p = .125), and Cohen's d effect size was 0.47, indicating a small effect of group on spelling accuracy.

Figure 3.

SSS-Elements scores by group. Solid line denotes the median. SSS = Spelling Sensitivity Score; CNH = children with normal hearing; CHL = children with hearing loss.

Mixed-Effects Model

To further examine the specific language skills that were utilized while spelling that were observed with SSS-W scoring, a mixed-effects model comparison was completed. Each participant ID and word ID were entered as random effects in the null models, and group was entered as the fixed effect. See Appendix B for the R syntax used in this analysis. Group was not a significant predictor of word score compared with the null model for any of the word scores; Word Score 0 (change in Akaike information criterion [ΔAIC] = −1.78, χ² = 0.82, p = .365), Word Score 1 (ΔAIC = −1.9, χ² = 0.08, p = .783), Word Score 2 (ΔAIC = −1.38, χ² = 0.62, p = .431), and Word Score 3 (ΔAIC = −0.2, χ² = 1.8, p = .180). Table 4 displays the model comparison statistics for each word score.

Table 4.

Model comparisons for mixed-effects models for each word score.

Model	β group	ΔAIC	X2	p
1. Word score 0, null model
2. Word score 0, including group	.09
		−1.78	0.82	.365
3. Word score 1, null model
4. Word score 1, including group	.02
		−1.9	0.08	.783
5. Word score 2, null model
6. Word score 2, including group	−.01
		−1.38	0.62	.431
7. Word score 3, null model
8. Word score 3, including group	−.09
		−0.2	1.8	.180

Note. ΔAIC = change in Akaike information criterion when group was added as a predictor.

Research Question 4: Proportion of Errors With MLC System

A Mann–Whitney U test was used for this analysis. Significance was set at p < .008 to account for multiple comparisons. In MLC coding, it is possible for each word to contain more than one error and, thus, receive more than one code. The groups did not differ on the percentage of misspelled words that contained more than one error (p = .57, d = 0.17). When analyzing the differences in proportion of errors on the MLC, CNH and CHL did not differ in proportion of errors for any categories: PA errors (p = .589, d = 0.09), OPA errors (p = .366, d = 0.16), MGR errors (p = .041, d = 0.63), MK errors (p = .690, d = 0.11), SK errors (p = .612, d = 0.05), and “other” errors (p = .334, d = 0.33). Figure 4 displays the proportion of errors with the use of MLC. See Table 5 for means and standard deviations for each group.

Figure 4.

Multilinguistic coding errors: proportion by linguistic type. CNH = children with normal hearing; CHL = children with hearing loss; CI = confidence interval.

Table 5.

Means and standard deviations of multilinguistic coding variables by group.

Variables	CHL M (SD)	CNH M (SD)
Percentage of words with > 1 misspelling	55.00 (23.18)	51.05 (22.43)
Phonological awareness errors a	0.20 (0.20)	0.19 (0.18)
Orthographic pattern knowledge errors a	0.46 (0.14)	0.44 (0.13)
Mental grapheme representations errors a	0.18 (0.13)	0.27 (0.15)
Morphological knowledge errors a	0.02 (0.02)	0.02 (0.02)
Semantic knowledge errors a	0.08 (0.06)	0.08 (0.06)
Other errors a	0.05 (0.19)	0.01 (0.02)

Note. CHL = children with hearing loss; CNH = children with normal hearing.

^a Reported in proportion.

Discussion

The purpose of this article was to determine whether differences in spelling errors occur between CNH and CHL in kindergarten. Previous research with older students suggests that spelling accuracy and linguistic knowledge use during spelling differs for CHL and CNH. Potential differences in spelling as early as kindergarten, however, had not been well explored.

No Differences in Spelling Scores Between Groups

In this study, children's single-word spellings were analyzed via three methods, namely, percent correct, SSS, and MLC, to compare spelling between the two groups. Percent correct was used to compare spelling accuracy, and SSS and MLC were used to further analyze types of linguistic knowledge children in kindergarten are utilizing in spelling. Results from this study revealed that, in kindergarten, CNH and CHL did not show significant differences in the number of spelling errors overall or in the number of not-attempted spellings and unanalyzable spelling attempts. Additionally, kindergarten CHL did not significantly differ in their SSS-E and SSS-W scores when compared to CNH. Although the SSS-E scores were not significantly different, it should be noted that the CHL mean score fell in a different category, illegal spelling of the element, than the CNH mean score, legal but incorrect spelling of the element. This categorical difference may be an early indicator of group differences in spelling that will emerge over elementary school.

Likewise, CHL did not make significantly different proportions of errors in phonological awareness, orthographic pattern knowledge, mental graphemic representations, morphological knowledge, semantic knowledge, or “other” errors when compared to CNH utilizing the MLC system. Children in both groups had a higher proportion of orthography-based errors (i.e., orthographic pattern knowledge and mental graphemic representations) than any other category. Finally, the correlational analyses completed indicated that spelling is strongly related to reading and spoken language for both CHL and CNH during the kindergarten year.

In contrast to our present findings, previous evidence from older students with hearing loss suggests that spelling accuracy is indeed lower for CHL than for CNH (Apel & Masterson, 2015; Hayes et al., 2011). Additionally, children with cochlear implants between Grades 3 and 6 do demonstrate proportionally more errors in phonological awareness and orthographic pattern knowledge compared to CNH (Werfel, 2016). Similar differences in error patterns have been observed in middle school samples of students with hearing loss (Bowers et al., 2014, 2016).

The findings of this study indicate that these errors that are seen in late elementary and middle school grades have not yet emerged in kindergarten. This finding is particularly interesting when interpreted in light of the group differences in spoken language between CHL and CNH. Although both groups had mean language scores within the average range, CHL had scores that were on average over a full standard deviation below CNH. Both groups had relatively low spelling accuracy in kindergarten (< 25% accurate). Perhaps this low overall accuracy explains why group differences are observed in later grades but had not yet emerged for these kindergarteners; everyone's spelling skills were just starting to emerge. Developmental spelling scores, however, were relatively high and did not differ across groups. Finally, our linguistic analyses revealed no difference in proportion of error types across the groups. We conclude, therefore, that our study provides converging evidence that CHL have similar spelling skills in kindergarten to CNH.

These findings align with both the stage theory and repertoire theory when analyzing kindergarten children's spelling acquisition. In this study, the spellings of CHL in kindergarten displayed a greater proportion of orthographic than phonological errors and showed evidence of transitioning between the stages of alphabetic and within-word spelling. The participants relied on their repertoire of linguistic skills, by representing a variety of linguistic features in their spellings.

Linguistic-Based Scoring of Spelling

This study further supports the position of Masterson and Apel (2010) that spelling should not be viewed as only correct or incorrect. This holistic way of quantifying spelling skills does not allow professionals to determine the linguistic skills an individual child may be utilizing and/or not utilizing to attempt spellings, which is important for intervention planning by both speech-language pathologists and teachers. Although there were no significant group differences in the study, the SSS-E scores revealed potential categorical differences between the groups and the MLC coding revealed patterns of types of linguistic errors across groups that were more common than others. This information could help guide spelling instruction and intervention in kindergarten. With the SSS and the MLC system, professionals can determine the underlying linguistic skills that a child needs for further intervention to be successful, and we suggest that even in kindergarten a combination of holistic and linguistic spelling measures can provide a robust picture of a child's spelling abilities. Specific to this study, these analyses allowed us to see that not only was spelling accuracy similar across the groups, but so were the patterns of errors. We conclude that the SSS and MLC can be used in a complementary manner for young children and both are valuable additions to spelling assessment.

Implications for Intervention

Results from this study indicated that kindergarten CNH and CHL do not significantly differ in their spelling abilities when measured holistically or linguistically. This finding leads us to consider that kindergarten spelling instruction and intervention for CHL may not necessarily need to differ in content from that for CNH. Previous research has shown, however, that CHL in middle school begin to show spelling-based deficits attributable to phonological awareness and semantic knowledge (Apel & Masterson, 2015; Bowers et al., 2014). When we consider our findings alongside these previous findings with older students with hearing loss, we propose that spelling instruction in kindergarten may need to focus more specifically on phonological awareness and semantic knowledge than is necessary for CNH in order to prevent these later group differences. This is, of course, a data-based hypothesis that should be tested in future research.

Limitations and Future Directions

The participants with hearing loss in this study utilized amplification and spoken language as their primary communication mode. It would be interesting to compare spelling abilities in the same study of CHL who use different types of communication modalities (i.e., amplification and spoken language, total communication, manually coded English, sign language). Amplification devices used by the participants varied; thus, the results cannot be generalized to one specific hearing device. The sample size of this study is relatively small, and our sample has relatively high maternal education. These findings should be replicated in a larger sample of young CHL. We did not log amplification use or fit, and future studies should consider including these audiologic variables. Additionally, the spelling assessments administered were single-word spelling assessments. This type of spelling assessment is typical for how spelling is taught and tested in school; however, this spelling task does not generalize well to how spelling is completed outside of school. Outside of school, spelling is rarely completed one word at a time and utilizes contextual clues to determine the correct spelling of words. It would be interesting to determine if the same types of errors are seen when completing a story-type writing assessment to allow for contextual clues. The findings from this study should be interpreted with those limitations in mind.

Research on spelling for CHL has not been completed across subsequent grades leading up to middle school. In order to determine the grade in which the shift from similar spelling performance to proportionally more errors in phonological awareness and semantic knowledge occurs, longitudinal designs are needed. We are currently completing such a design in our lab. The information from the grade-specific research will help to further influence the intervention received by CHL regarding their spelling skills, which in turn influences the reading and language skills of children.

Conclusions

In conclusion, kindergarten CHL do not appear to differ significantly in the overall spelling errors compared to CNH when viewing spelling in the conventional correct/incorrect scoring procedures or in proportion of linguistic-based errors across categories. Linguistic-based spelling scoring, such as the MLC system and SSS, should be utilized in addition to the conventional scoring method to determine the appropriate spelling intervention for CHL and CNH. More research needs to be completed in order to generalize these results, gain more understanding of the spelling abilities of children with and without hearing loss throughout elementary and middle school, and determine the best course of intervention for individual CHL to increase their spelling abilities, which in turn will positively influence their reading and language skills.

Author Contributions

Carson Aho: Conceptualization (Equal), Data curation (Supporting), Formal analysis (Lead), Writing - Original Draft (Lead). Krystal L. Werfel: Conceptualization (Equal), Data curation (Lead), Formal analysis (Supporting), Funding acquisition (Lead), Supervision (Lead), Writing - Original Draft (Supporting), Writing - Review & Editing (Lead).

Appendix A

SSS-Spelling List

1. Rake

2. Nine

3. They

4. Ship

5. Pool

6. Reach

7. Whales

8. Unite

9. Jet

10. Head

11. Hang

12. Putting

13. Bowl

14. Fire

15. Cute

16. Leaf

17. Why

18. Shut

19. Bake

20. Then

21. Cure

22. Pie

23. White

24. Key

25. Jog

Appendix B

R Syntax for Mixed-Effects Model

Word Score 0

> m0 = Imer(WordScore_0~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

> m0.null = Imer(WordScore_0~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

>anova(m0, m0.null)

Word Score 1

> m1 = Imer(WordScore_1~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

> m1.null = Imer(WordScore_1~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

>anova(m1, m1.null)

Word Score 2

> m2 = Imer(WordScore_2~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

> m2.null = Imer(WordScore_2~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

>anova(m2, m2.null)

Word Score 3

> m3 = Imer(WordScore_3~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

> m3.null = Imer(WordScore_3~(1 I PID)+(1 I TargetID)+Group,data=R_Database,REML=FALSE)

>anova(m3, m3.null)

Funding Statement

Research reported in this article was supported by National Institute on Deafness and Other Communication Disorders Award R03DC014535 (PI: Werfel). This article reports the findings of the first author's master's thesis. We appreciate the feedback on design from Kenn Apel and Gina Crosby-Quinatoa. Study data were collected and managed using REDCap electronic data capture tools hosted at the University of South Carolina. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.