Derivational Morphology Bridges Phonology and Orthography: Insights Into the Development of Word-Specific Spellings by Superior, Average, and Poor Spellers

Abstract

Purpose

Morphology, which is a bridge between phonology and orthography, plays an important role in the development of word-specific spellings. This study, which employed longitudinal sampling of typically developing students in Grades 3, 4, and 5, explored how the misspellings of words with derivational suffixes shed light on the interplay of phonological, orthographic, and morphological (POM) linguistic features as students learn to integrate POM features appropriately to generate correct spellings.

Method

Sixty typically developing Grade 3 students were tested using the Spelling subtest from the Wechsler Individual Achievement Test–Second Edition (Wechsler, 2001) and were divided into superior, average, and poor spellers. Students' spelling skill was then assessed using the Wechsler Individual Achievement Test–Second Edition annually for another 2 years. Misspelled derivations from these three testing sessions were analyzed for linguistic feature errors and error complexity/severity. Differences in the integration of POM features across spelling ability levels at Grades 3–5 were analyzed with Kruskal–Wallis analyses of variance.

Results

Longitudinal results demonstrated POM integration for the development of word-specific spellings involving derivational morphology was in its initial stages over Grades 3–5 and was influenced by spelling ability level. Information from a qualitative analysis revealed considerable variability in how students applied their POM knowledge to spell complex derivations.

Conclusions

Word-specific spellings draw on multiple linguistic codes—P, O, and M—and their interconnections. It involves more than an understanding of orthographic rules. Rather, accurate spelling develops through an increased understanding of the phoneme–grapheme relationships as facilitated by the identification of word parts (base + or − affixes) in written language. Educational and clinical implications are discussed.

Spelling may be predictive of reading and writing problems as early as Grade 1 (Ritchey & Coker, 2014; Treiman et al., 2019). Because spelling is predictive, it becomes educationally and clinically important to understand the linguistic role of spelling in language and literacy learning. We begin this article with two myths about English spelling that are relevant for educational and clinical practices. One is the myth that spelling is simply a mechanical skill and handwriting is just a motor ability (James et al., 2016; Straub & Alias, 2013). In reality, spelling is a complex linguistic behavior that involves transcription of linguistic information into word forms. Moreover, spelling accuracy competes for the cognitive resources associated with writing (Hayes & Berninger, 2014) and can affect the length and quality of written expression (Berninger, 2009).

A second myth is that learning to spell in English is difficult because of the apparent lack of consistency in letter–sound correspondences in written words. However, the reality is that English spelling is based on a morphophonemic orthography (Nunes & Bryant, 2006; Venezky, 1970, 1999) consisting of three interrelated linguistic codes as described in triple word form theory (Bahr et al., 2012). These are phonology (P), the sounds in heard/spoken words; orthography (O), the letters in read/written words; and morphology (M), the relationships between base words and their affixed forms in both heard/spoken and read/written words (hereafter referred to as POM). ¹ Learning to integrate these codes and their cross-code connections renders English spelling predictable, and the transparency of English spelling becomes more apparent when M is viewed as the bridge between P and O in both heard/spoken and read/written words. In other words, an understanding of spelling in English relies on knowledge of how affixes can influence the pronunciation and orthographic regularity of words. According to this perspective, M functions as the bridge by which permissible P–O correspondences are applied for each grapheme in a word-specific spelling. In this way, POM interrelates with the corresponding word meaning (see Ehri, 1997; Venezky, 1970, 1999). As such, both decoding (word recognition) and encoding (spelling) rely on the increasing developmental integration of the three linguistic codes (Ehri, 1997; Treiman, 2017), demonstrating that spelling is a word formation process, not simply memorization of spelling rules and forms.

The purpose of the current study is to explore how the misspellings of words with derivational suffixes shed light on cross-code POM relationships in three groups of typically developing spellers with varying ability—superior, average, and poor—followed from Grades 3 to 5. Derivational suffixes were of interest because they contribute to learning to read and spell the complex words increasingly encountered in Grade 4 and above in read texts and in composing. Findings will be relevant to designing instructional strategies to assist upper elementary–age students in applying POM integration to create word-specific spellings.

Key Terms for Understanding Spelling Development

We briefly review four concepts relevant to this study. These are POM coding, derivational suffixes, word-specific spelling, and morphological complexity.

POM coding refers to the storing and processing or “essentializing” of each POM code. In other words, writers predict general regularities in P, O, and M, likely using statistical learning mechanisms (Treiman & Kessler, 2013), which result in the word-general knowledge used in spelling. Regularities might include (a) likely phoneme–grapheme connections for a phoneme (P) and one or two letters (O); (b) the relative frequency of legal letter groups (O; e.g., –ic in music is less frequent than –k in meek); and (c) the probability that a letter sequence, such as –er is a suffix (M) in writer but not ladder. Derivational suffixes mark the part of speech, modify the semantic codes of base words, and are of special interest in Grade 4 and above because they contribute to how syntax is constructed in written affixed words. If each code—P, O, and M—only functioned independently, there might be chaos in learning how to efficiently manage these multiple levels of language, that is, the subword (POM), word, and syntax levels. Therefore, learning to interconnect these codes two at a time (P–O, O–M, or P–M) or three at a time (P–O–M) is essential so that the codes work together collaboratively for generating word-specific spellings (Ehri, 1997; Treiman, 2017). Word-specific implies spellings are linked to meaning, stored in memory, and can be retrieved with minimal effort (Carlisle, 2000), that is, an accurate spelling. For a more complete review of the research on POM and its influence on multiple language levels, see Silliman et al. (2018).

As just indicated, the word-specific spelling of derivations represents an important linguistic bridge for regulating the preservation of base words in the context of surface irregularities (i.e., morphological complexity). These complexities can involve shifts in pronunciation (i.e., phonology) and/or orthography (e.g., strong–strength, circle–circular) along a continuum from more (content–contentment) to less (magic–magician) transparent semantic relationships. The notion of continuum is important because the ease of access to derivational meanings, including their familiarity, is facilitated by the degree of overlap among the POM components (Carlisle & Stone, 2005), which can then affect the depth of morphological analysis. For example, transparent derivations, such as excite–excitement, in theory, are easier lexical items for children to spell because both meaning and orthography are accessible once spellings are fully linked to semantic relationships and stored in memory (Carlisle, 2000). It appears, therefore, that the morphological complexity of particular derivations can influence the stability of word-specific knowledge (Bourassa & Treiman, 2014). These findings further suggest that an understanding of POM relationships can compensate in learning to spell derivations when the semantic relationships are less transparent, as in strong–strength.

Identifying Misspellings

It then follows that, in the process of learning these word-specific spellings during the elementary and middle-school grades, children's errors on dictated spelling tests may reflect gaps in the developmental orchestration of the three linguistic codes (Carlisle & Goodwin, 2014; Carlisle et al., 2001). As illustrated in Figure 1, these gaps, which can involve P–O, O–M, and/or P–M discontinuities, may variably result from a student's familiarity with the overall writing theme or topic and the genre (narrative, expository) being constructed, the quality of spelling/writing instruction from the early grades onward, and the depth of general and word-specific knowledge possessed by individual students. The deeper the word knowledge, “the more likely he or she is to recognize it, spell it, define it, and use it appropriately in speech and writing” (Joshi et al., 2008–2009, p. 9).

Figure 1.

Examples of incomplete cross-code integration.^a

However, investigations of spelling ability often rely on a simple percent correct to reflect spelling skill (e.g., Kohnen et al., 2015; Wanzek et al., 2017), which provides minimal information about the individual contributions of P, O, and M to an integrated linguistic code. Other approaches depict how close the error is to the target word by quantifying students' phonological and orthographic knowledge as displayed in their misspellings (e.g., Bruck & Waters, 1988; Caravolas et al., 2001). Thus, the manner in which errors are described and explained influences the outcomes of any study and affects the design of spelling instruction as well.

In the current study, a standardized normed measure based on number of correct spellings was used to define spelling ability groups, but analysis of spelling errors of words with one or more derivational suffixes was used to investigate a specific kind of morphological coding (i.e., derivational suffixes) in developing POM interrelationships. In this type of analysis, the integrity of the “phonological skeleton” (Bourassa & Treiman, 2003) is paramount, that is, whether the consonant–vowel sequencing (CV, CVC, CCVC, etc.) is maintained. When the phonological skeleton of the target is intact, the spelling is plausible and the representations for each phoneme can be analyzed based on their relationship to the target. The phonological elements of a spelled word at this point can be described using either constrained or unconstrained systems of analysis, depending on whether or not the spelling is driven by phoneme–grapheme correspondence rules that include positional constraints (Bruck & Waters, 1988).

Constrained Scoring Systems

Constrained scoring systems consider misspelled words as phonetically accurate if legal orthographic patterns create a match for the target. For example, if the target rain is spelled rane, a constrained analysis would consider this type of production as an appropriate spelling because it is an orthographically legal way to represent the phonological structure of the target (i.e., the vowel-silent e pattern is a legal orthographic pattern resulting in a long vowel sound). Contemporary constrained approaches include the Spelling Sensitivity Score (SSS; Masterson & Apel, 2010a, 2010b; Werfel & Krimm, 2015), which is related to the Spelling Performance Evaluation for Language and Literacy, Second Edition (Masterson et al., 2006). The SSS results in two scores: SSS-Word, which reflects the score of the most severe error among the graphemic elements in the spelled word, and the SSS-Element, which displays an average of the scores for all word elements, both those that are correct and those in error. It is unresolved whether the Spelling Performance Evaluation for Language and Literacy, Second Edition/SSS relates to the integrated POM resources that students recruit to produce word-specific spellings (Werfel et al., 2019), such as derivations.

More recently, Daffern et al. (2015) developed the Components of Spelling Test to systematically assess POM aspects of Standard English spelling. This test was carefully crafted to assess common spelling patterns representing the P, O, and M linguistic features necessary for spelling skill. This assessment was validated using a large school-age sample and has been shown to be reliable in identifying spelling errors within a particular linguistic category. However, it is not clear how the integration of POM knowledge is assessed on this test and how multiple linguistic feature errors within a target word, such as exitmant for excitement, would be handled.

Unconstrained Scoring Systems

In an unconstrained analysis, a misspelled production does not need to include orthographically legal patterns to be considered phonologically plausible. As long as a plausible grapheme represents each sound in the word, an unconstrained scoring system assumes the phonological structure is represented. For example, in the misspelling of the word charge as crg, an unconstrained analysis acknowledges the phonetic plausibility of the spelling because the c is marking the digraph representing /tʃ/, the r represents the rhotic vowel /ɑɚ/, and the g represents the /dʒ/; hence, the phonological structure of the target /tʃɑɚdʒ/ is maintained. An unconstrained analysis allows for the evaluation of underlying linguistic features employed in each spelling attempt, while also permitting an examination of how POM impacts spelling strategies over time. To this end, two scoring systems were developed to evaluate the contributions of POM to the development of word-specific spellings.

The Phonological–Orthographic–Morphological Analysis of Spelling (POMAS; Bahr et al., 2012) is an unconstrained, qualitative scoring system grounded in the notion that POM relationships must be coordinated to result in accurate spelling (i.e., triple word form theory). This type of analysis does not rely on predetermined rules for identification, but instead the scorer analyzes the misspelling in terms of the linguistic strategies that are used.

POMAS. Briefly, spelling errors first are coded as belonging to one of the three linguistic error categories: P, O, and M. The emphasis then is on what linguistic resources the student uses when spelling a word rather than focusing on the identification of specific P, O, or M rules. On the POMAS, phonological errors include phoneme omissions or substitutions that affect the phonological skeleton (i.e., the phonetic elements) of the target word. An example would be carls for careless, where the second vowel is missing. Orthographic errors account for grapheme substitutions and omissions that occur when the phonological skeleton is intact. An example would be carelless for careless, where unnecessary letter doubling occurred. Errors within the morphological category involve misidentification of the word form (i.e., the spelling of a homonym for the target word and errors involving the base word or affix in a derivation). An example would be absents for absence, where application of the wrong suffix resulted in a homonym. More detail about this scoring system is available in Bahr et al. (2012). Given the number of linguistic features that can be in error, the POMAS is useful for a qualitative analysis of the linguistic knowledge that the student possesses, but it does not provide a score that is useful in tracking the development of the cross-code integration that is necessary for word-specific spellings.

POMplexity. The POMplexity metric (Bahr et al., 2020) was devised as a measure that quantifies the severity of the POMAS error(s) produced in misspelled words. This scoring system also permitted evaluation of the individual effects of P, O, and M on misspelled words. Inspired by Stoel-Gammon's (2010) Word Complexity Measure, POMplexity is designed to measure how different the misspelling is from the target spelling (i.e., jumpt and gomt for jumped, and excitment and exitment for excitement). In other words, a POMplexity score quantifies error complexity or severity by indicating how far a misspelling is from the target word.

A POMplexity score is generated by selecting a numerical value (on a scale of 0–3) for each linguistic feature error identified by the POMAS. This value is a reflection of the severity of the linguistic feature error and is based on the traditional notion that an omission or addition of a grapheme disrupts the word from more than a substitution of a grapheme (Bourassa & Treiman, 2003). Each element in the misspelled word is analyzed for linguistic feature errors. Spelled words can have errors in all three linguistic categories. A value of 0 indicates that a word part was spelled correctly. If a misspelled word contained multiple errors within a particular P, O, or M category, then these scores were added to get a total P, O, or M POMplexity score for each word. Lower POMplexity scores reflect stronger integration of linguistic features during word-specific spelling. When P, O, and M are zero, then cross-code integration is complete. Table 1 illustrates how POMAS and POMplexity work together to quantify the severity of a misspelling.

Table 1.

Illustration of how Phonological–Orthographic–Morphological Analysis of Spelling (POMAS) and POMplexity provide different information about spelling errors.

Spelling	POMAS code	POMplexity score
		P	O	M	Total
excitement		0	0	0	0
exitement	O–consonant deletion M–base misspelled	0	2	1	3
excitment	O–letter name M–base misspelled	0	1	1	2
excement	P–phoneme deletion O–long vowel error M–base misspelled	2	1	1	4
exitment	O–consonant deletion O–letter name M–base misspelled	0	3	1	4
ecxitment	P–letters reversed O–letter name M–base misspelled	0.5	1	1	2.5
exsitment	P–ambiguous letter O–letter name M–base misspelled	1	1	1	3
exightment	O–consonant deletion O–graphemic pattern error M–base misspelled	0	3	1	4
exitmant	O–consonant deletion O–letter name O–schwa vowel M–base and suffix misspelled	0	4	2	6

Note. The unique contributions of P, O, and M to the misspelling are indicated, and the total scores illustrate how the total score can be the same for different combinations of P, O, and M errors. P = phonology; O = orthography; M = morphology.

Study Overview

The current research was informed by two insights. First, error severity alone does not provide necessary information about developmental patterns, individual differences, or tailoring instruction to develop POM integration. Second, capturing the separate and combined contributions of POM across time requires moving beyond cross-sectional studies to follow how spelling develops over time. Deeper understanding of longitudinal variations in the derivational misspellings of typically developing students can also serve as a benchmark for the misspelling patterns of students with a variety of learning problems as might be found, for example, in oral language impairment, dyslexia, and dysgraphia (Bahr et al., 2020).

Few longitudinal studies have examined POM growth (Pacton & Deacon, 2008) in general or in varying spelling ability groups. Thus, the current study's goal was to sample the relationships among linguistic feature knowledge and spelling patterns longitudinally on a dictated spelling task in Grades 3, 4, and 5 and to gain insight into the process of learning to integrate the POM codes in 10 derivational spellings. The research questions examined whether the POM results varied by spelling ability groups and whether the POM findings for spelling ability groups changed across Grades 3, 4, and 5.

Method

Participants

Under a data sharing plan, another research group provided the misspelling data and standardized test means and standard deviations for verbal comprehension and dictated spelling for 20 superior spellers (10 girls, 10 boys), 20 average spellers (10 girls, 10 boys), and 20 poor spellers (10 girls, 10 boys) in Grade 3. The previous research group carefully identified typically developing students and divided them into superior, average, and poor spelling groups, recognizing that there is variability in spelling skill. Consistent with previous research, poor spellers are not always associated with a language learning disability (Fayol et al., 2009; Garcia et al., 2010; Holmes et al., 2008; Lennox & Siegel, 1996; Martin-Chang et al., 2014).

For this study, the Spelling subtest of the Wechsler Individual Achievement Test–Second Edition (WIAT-II; Wechsler, 2001) had been used to identify the superior spellers with scores of 122 or greater, the average spellers with scores between 105 and 112, and the poor spellers with scores at 95 or below. The lowest score of the superior spelling group differed from the highest score of the average spelling group by 10 points; the lowest score of the average spelling group differed by 10 points from the highest score of the poor spelling group. On this spelling dictation measure, the examiner pronounced the target word, then used it in a sentence, and pronounced the word again. These children were also given the same spelling test in Grades 4 and 5. Spelling ability remained constant in that no student's score ever resulted in a change of placement within the spelling ability groups over Grades 3–5. Table 2 lists the WIAT-II means and standard deviations for the spelling ability groups in Grades 3, 4, and 5 and provides the means and standard deviations associated with the Wechsler Intelligence Scale for Children–Third Edition Verbal Comprehension factor (Wechsler, 1991) for each spelling ability group when they entered the study. Students attended public schools in a large urban school district. Sample ethnicity was Caucasian (70%) and Asian (25%) with others of mixed or not reported ethnicity. Parental level of education ranged from high school to more than high school but less than college to at least a college degree (40%). A parental questionnaire and interviews ruled out developmental disabilities and co-occurring medical conditions associated with learning problems.

Table 2.

Wechsler Individual Achievement Test–Second Edition Spelling subtest means and standard deviations (in parentheses) for the spelling ability groups in Grades 3, 4, and 5.

Grade	Superior spelling	Average spelling	Poor spelling
3	133.75 (8.24)	106.80 (3.99)	87.50 (4.45)
4	132.20 (5.69)	105.95 (6.40)	87.70 (6.08)
5	129.80 (5.90)	108.00 (3.53)	84.80 (10.46)
WISC-III Verbal Comprehension	122.00 (10.17)	112.80 (15.5)	104.10 (17.40)

Note. The means and standard deviations of the Wechsler Intelligence Scale for Children–Third Edition (WISC-III) Verbal Comprehension factor (Wechsler, 1991) for each spelling ability group when they entered the study is also presented.

Analyzing POM Features in Misspelled Words With Derivational Suffixes

Spelling Words

The current study focused on the spelling of 10 words taken from the WIAT-II that contained derivational morphology: careless, absence, prestigious, pharmaceutical, conscientious, excitement, subsidize, edition, assistants, and a transformation of a base word, the adjective strong to a noun, strength. It should be noted that the word assistants has both a derivational suffix (–ant) and an inflectional suffix (–s). It is recognized that students in Grades 3, 4, and 5 may not be familiar with all of these words, but their attempts at spelling these complex derivations should provide insight into their understanding of word-general knowledge as they learn to integrate POM features.

Procedure

The 10 derivational words were extracted from the WIAT-II Spelling subtest for 60 students with varying spelling ability in Grade 3 and who were followed through Grade 5. Both POMAS, which identifies each linguistic feature error (P, O, or M) in a misspelled word, and POMplexity, which quantifies the severity of each linguistic feature error (P, O, or M) in a misspelled word, were applied to code each misspelled word (see Table 1 for a scoring of the word excitement by several different students). Using the POMAS scores as a guide, a graduate student in speech-language pathology quantified each linguistic feature error using the POMplexity scoring system. For example, carelles for careless would receive two scores in the O category, 2 points for unnecessary doubling of the l and 2 points for the lack of doubling for the s. That results in a total of 4 points in O, a score of 0 for P because the phonological skeleton was present, and then 1 point in the M category for the misspelling of –less. In this case, the misspelling was limited to the second word part in this complex word and the errors involved orthographic difficulties.

To ensure reliability of coding, both the POMAS and POMplexity scores were evaluated. All spelling words were initially coded with the POMAS by the first author and a research assistant. Their linguistic feature codes were compared for each spelling word. Any disagreements were resolved by a third rater (i.e., the second author). Once POMAS coding was verified, the research assistant assigned a POMplexity rating to each POMAS code identified in a misspelled word. Then, the first author randomly selected 20% of the words from each participant and each year (N = 157 items) for recoding, and the POMplexity codes from each examiner were compared for each misspelling. Coding was identical for 142 of the 157 items, yielding a 90% agreement rate.

Results

Individual Participant Data

Every student did not spell each of the dictated words with derivational suffixes on the WIAT-II Spelling test because there were test ceilings (i.e., testing ceased after a specified number of errors). Table 3 lists the number of attempts for each word by grade. Words not tested due to basals were considered to be correctly spelled, while words not spelled because of test ceilings were omitted. The result was 845 observations for the target 10 spelling words with derivational suffixes (or derivational transformations) across Grades 3–5. Table 4 provides descriptive data for the POMplexity scores generated by grade. The means demonstrate a general decrease in POMplexity scores over time, which indicates increasing integration of POM features as the student learns the word-specific spellings. This finding supports the strength of the POMplexity scale in identifying the degree of integration of linguistic features as students attempt new words.

Table 3.

Number of students who attempted to spell each derivation by grade.

Word	Grade 3 (n = 60)	Grade 4 (n = 60)	Grade 5 (n = 60)
Careless	55	56	48
Strength	39	44	44
Absence	30	35	36
Excitement	30	35	34
Subsidize	17	27	29
Edition	16	25	28
Assistants	14	24	27
Prestigious	12	22	26
Pharmaceutical	10	19	20
Conscientious	8	17	18

Note. If the student had reached ceiling with a specified number of spelling errors for grade level, then no more items were administered on the Wechsler Individual Achievement Test–Second Edition Spelling subtest.

Table 4.

Descriptive statistics: mean POMplexity scores and standard deviations by grade and spelling ability level: superior (Sup), average (Avg), and poor.

Score	Grade 3			Grade 4			Grade 5
	Sup	Avg	Poor	Sup	Avg	Poor	Sup	Avg	Poor
Phonological scores
M	0.32	0.542	0.75	0.28	0.332	0.513	0.157	0.256	0.631
SD	0.256	0.498	0.862	0.28	0.319	0.558	0.137	0.182	0.49
Orthographic scores
M	1.114	1.877	3.356	1.321	1.614	1.962	0.99	1.516	1.755
SD	0.76	1.458	2.07	0.635	1.397	1.386	0.725	0.679	1.423
Morphological scores
M	1.083	1.565	2.098	1.106	1.364	1.55	0.949	1.733	1.757
SD	0.484	0.63	1.102	0.292	0.665	1.085	0.395	0.576	1.026

Analysis of POMplexity Scores by Spelling Ability and Grade

The POMplexity scores for each linguistic feature category (P, O, and M) served as the dependent measures in three separate Kruskal–Wallis analyses of variance (ANOVAs; one for each linguistic category) because POMAS and POMplexity scoring procedures do not result in independence among the linguistic categories. Nonparametric statistics were utilized because the number of students who spelled each derivation varied by spelling ability group and grade (i.e., an unequal n across groups), and there was the possibility of nonnormal distributions of data. Each of these analyses will be described below.

The first analysis considered differences in P-POMplexity scores across spelling ability groups. The Kruskal–Wallis ANOVA revealed significant differences attributable to spelling ability level at Grades 4 and 5 (see Table 4). Mann–Whitney U tests with a Bonferroni correction (p = .02) were used for post hoc testing. At Grade 4, there was a significant difference between the poor and both the average and superior groups. The average and superior groups were significantly different from one another in Grade 5 (see Figure 2). This finding suggests that the poor group produced misspellings that contained more severe phonological errors than the other two groups in Grade 4. However, in Grade 5, only a difference between the average and the superior group was noted. The poor group's P-POMplexity score raised slightly with a large standard deviation (SD = 0.49) in Grade 5. This is most likely related to their attempts to spell more difficult derivations and, in doing so, using a phonological strategy to spell these words, that is, stranth for strength and axsitement for excitement.

Figure 2.

Changes in P-POMplexity means over time by spelling ability group (superior, average, and poor). Significant group differences are marked with an asterisk (*).

The second Kruskal–Wallis ANOVA considered O-POMplexity scores. Again, there was a significant difference attributable to spelling ability level at Grade 4 (see Table 4). Post hoc testing with the Mann–Whitney U using a Bonferroni correction (p = .02) indicated that poor spellers performed below the superior and average spellers (see Figure 3). The average and superior spellers produced less severe orthographic errors than the poor spelling group, indicating stronger integration of orthographic knowledge in the spelling of complex derivations in Grade 4. Once again, variability in O-POMplexity scores increased at Grade 5, obscuring the differences between spelling ability levels, because the average and superior spellers attempted to spell the more complex derivations.

Figure 3.

Changes in O-POMplexity means over time by spelling ability group (superior, average, and poor). Significant group differences are marked with an asterisk (*).

The third Kruskal–Wallis ANOVA involved M-POMplexity scores. In this case, there was a significant difference attributable to spelling ability level in Grades 3–5 (see Table 5). Post hoc testing with the Mann–Whitney U test with a Bonferroni correction (p = .02) revealed that those with poor spelling ability had higher M-POMplexity scores than the average spelling ability group in Grade 3 (see Figure 4). The superior spelling ability group had more accurate spelling of words with derivational suffixes than the poor spelling group at Grades 3, 4, and 5 and from the average spelling group in Grades 4 and 5. These results suggest that the superior spelling group was more adept at word formation than the other spelling groups. The average spelling group evidenced better integration of morphological features in Grade 3 but achieved M-POMplexity scores that were closer to the poor spelling group in Grades 4 and 5.

Table 5.

Results of the Kruskal–Wallis analyses of variance by grade.

Score	N	H	df	p
P-POMplexity	56
Grade 3		2.057	2	.357
Grade 4		10.521	2	.005
Grade 5		8.460	2	.015
O-POMplexity	56
Grade 3		5.202	2	.074
Grade 4		10.044	2	.007
Grade 5		2.424	2	.298
M-POMplexity	53
Grade 3		30.183	2	< .001
Grade 4		14.470	2	.001
Grade 5		15.642	2	< .001

Figure 4.

Changes in M-POMplexity means over time by spelling ability group (superior, average, and poor). Significant group differences are marked with an asterisk (*).

In summary, the results, based on longitudinal sampling of the same students in Grades 3, 4, and 5, indicate that cross-code integration for words with derivational suffixes develops across these grades as students increasingly encounter words with derivational morphology. Grade 4 was a critical year in that the poor spelling group produced higher P-POMplexity and O-POMplexity scores, which correspond to less well-developed word-specific spellings, than the average or superior spelling ability groups. The average group had significantly higher P-POMplexity scores (less well developed) in Grade 5, suggesting that they were not developing POM integration for phonology at the same rate as the superior spelling group. Differences among spelling ability groups were most evident in the M-POMplexity scores. The superior spelling group had lower scores (better development) than the poor spelling group for Grades 3, 4, and 5 and the average group in Grades 4 and 5. The average group had lower M-POMplexity scores (better developed) than the poor spelling group at Grade 3. These findings imply that poor spellers continued to struggle throughout Grades 3, 4, and 5 compared to their grade peers with integration of POM features when spelling words with derivational suffixes. However, no spelling ability group mastered the spelling of words with derivational suffixes by Grade 5, suggesting the continued need for instruction on POM integration and word-specific spelling into middle school.

Exemplars of Word Formation Patterns

During the POMAS scoring, several different spelling patterns emerged. According to the Word-Specific Continuum of Derivational Complexity (Silliman et al., 2018), students gradually distill their word formation process by refining word-general knowledge and integrating P, O, and M information into new word forms. However, the manner in which this refinement and consolidation takes place may vary depending on item and participant characteristics.

Table 6 displays examples of derivational spelling patterns for students in Grades 3, 4, and 5. From a descriptive perspective, three developmental patterns emerged in written word formation. The first prototype appears to represent more “linear development.” One speller's attempts to transcribe prestigious, as shown in Table 6, demonstrates this trend: from an ambiguous attempt in Grade 3 (prejudace) to increasing refinements involving the matching of P to O in Grades 4–5. However, in Grade 5, the student is finally able to spell the base (prestige), which suggests some familiarity with the word meaning but continuing difficulty with applying specific suffixation rules.

Table 6.

Examples of derivational misspelling patterns from two different students as they attempted target word spellings over three test dates.

Target word	Grade 3	Grade 4	Grade 5
absence	absince	abcece	abcence
	absence	absence	abscence
excitement	excitement	exitment	excitment
	exitment	ecxitment	ecxitment
subsidize	subsadice	subsidise	subsadise
	subsadise	subsidize	subsodize
prestigious	CE	prestijust	prestigons
	prejudace	prestejuse	prestijus
pharmaceutical	CE	CE	CE
	parmisuticle	farmicudical	pharmiscudical
conscientious	CE	CE	CE
	contiengense	conchientious	conchienchous

Note. Target words are from the Wechsler Individual Achievement Test–Second Edition (Wechsler, 2001). CE = ceiling effects that precluded the student from attempting the specific target word.

In contrast to the linear configuration, the second pattern entails “recursion.” Borrowing from Christiansen and Chater (1999), recursion involves dynamic processing of a language structure that requires ongoing linguistic shifts among its components until a meaningful end state emerges; hence, recursive strategies frequently display intermittent patterns of discontinuity and stability. These “back and forth” shifts occurred in another speller who demonstrated recursiveness in lexicalizing word-specific knowledge of the target word, excitement. In this instance, the student traversed over time from an initial accurate spelling in Grade 3 to an e-drop misspelling (excitment) in Grade 5 that plausibly combined the appropriate suffix with a word that ended in silent –e. The recursive aspects were evident in Grades 4–5 where the student struggled with letter sequencing when spelling the base, which points to an unstable integration of P, O, and M. Thus, the achievement of lexicalization in word-specific knowledge may be refined over a relatively long period for certain individuals.

The final strategy involved “inadequate morphological decomposition,” which may be another recursive variation. Some students relied on phonological analysis strategies to spell the target word. These less mature strategies, found in the spelling of homophones (assistance for assistants and addition for edition) and low-frequency derivations (e.g., farmersuitical for pharmaceutical or exsightment for excitement), may have resulted from a combination of factors. These variables could include limited semantic and syntactic analysis of the sentential context in which the target word was embedded and incomplete word familiarity. For example, as in the previous examples of pharmaceutical and excitement, students parsed at the syllable level instead of decomposing at the morphemic level. The result was the selection of real words to represent syllables in the target words. In this case, the student substituted an appropriate orthographic representation for the target phonology. This type of misspelling suggests increased word-general knowledge in terms of possible orthographic spellings, but incomplete integration of this information with morphology. As such, these examples of misspelling patterns again illustrate that lexicalization is a gradual process of integrating P, O, and M knowledge as students learn to form meaningful words.

Discussion

These longitudinal results demonstrate that the interconnections for the cross-code POM integration in spelling words with derivational suffixes develop across the upper elementary grades. The current investigation offers new insight into the cross-code relationships influencing the students' progression from word-general to word-specific knowledge. Specifically, the POMplexity scoring rubric provided a unique window into the individual and integrated contributions of P, O, and M in the development of spelling accuracy for words with derivational suffixes for superior, average, and poor spellers in Grades 3, 4, and 5.

Cross-Code Integration in Spelling Ability Groups

Students must discover how to leverage their knowledge of POM features as they learn to spell complex words with derivational suffixes. In the early grades, students rely on word-general knowledge in their attempts to spell new words. With time (and greater exposure to vocabulary and spelling/writing instruction), students begin to understand that word-general knowledge is not sufficient for the development of a word-specific spelling. The development of word-specific spellings is not complete by Grade 5, at least for this sample, as evidenced by the current data. It would appear that students need to go through a sequence of experimenting with word-general knowledge when they attempt to spell new words, which early on results in word-specific spellings for simple words and later on for increasingly complex word-specific spellings.

A decline in POMplexity scores indicates that the speller is closer to achieving POM integration for a word-specific spelling. Cross-code POM integration becomes more complex as students have to learn word-specific spellings for increasingly more complex morphological linguistic features. Grade 4 was a pivotal year for differentiating the poor spellers from the average and superior spellers. Both the P- and O-POMplexity scores were higher for the poor ability group in this grade, indicating that these students were struggling with cross-code integration. Moreover, it was apparent that poor spellers may have relied more on phonologically based strategies for spelling more complex derivations. In contrast, beginning in Grade 4, average spellers generally did not consistently differ from the superior and poor groups. This finding may be related to the observation that average spellers expressed notable variability in their knowledge of P, O, and M features. Given that students continued to attempt new derivations each year they were tested (i.e., performance ceilings increased from the previous year of testing; see Table 3), they would demonstrate poorer integration of POM as they attempted to use their word-general spelling knowledge to spell these less familiar words. Therefore, it is not surprising that P-, O-, and M-POMplexity scores slowly declined over Grades 3–5, but that the variability among students within groups increased.

Closer inspection of the misspellings produced for these words revealed a reliance on phonological strategies to spell these complex derivations. With time, students were able to parse the base word from the suffix. This decomposition became apparent as the suffixes tended to be spelled correctly more often than the base word. Support for this finding comes from Nunes and Bryant (2009), who linked word-specific knowledge to the development of morphemic spelling rules. As exposure to written word and vocabulary knowledge increases, students infer derivational patterns, which assist in the decomposition of the base from the suffix. Therefore, the word-specific knowledge acquired from parsing bases from affixes assists students in the spelling of complex derivations, even when the definition of the word (and its spelling) is less well developed. The result is the accurate spelling of suffixes while the base word continues to be spelled phonologically until its meaning is well consolidated with P, O, and M information. Base word familiarity influences the student's ability to access word meaning. However, the recognition and understanding of suffixes is also critical for the understanding of words in their syntactic and semantic contexts as well (Tyler & Nagy, 1989).

This phonological to morphological parsing strategy became especially evident when students used a real, but unrelated, word to represent features of derived words (e.g., FARMaSUITicle, subsiDIES, abSCENTS). Interestingly, these real words often bridged morpheme units in order to replace syllables, as in FARMERsuiticle, where the real word farmer replaced both the morpheme pharm– and the syllables phar/ma. In this case, the meaning of the morphological unit farmer is ignored for the sake of representing the syllabic unit. Here, a phonological strategy takes precedence (with epenthesis of the /r/ in farmer), where a student sounds out a word and uses a letter–sound correspondence to represent the phonological skeleton of the word.

This real word strategy parallels the overgeneralization pattern for word-specific knowledge in homophonous forms that Nunes and Bryant (2009) describe. In this account, students activate phoneme–grapheme relationships and then apply these patterns in situations where the phonological structure of an unknown word is similar. A related pattern emerged from this study wherein students frequently substituted homophones for the WIAT-II words as in addition for edition and assistance for assistants. Despite the fact that these homophones were presented to the students within a sentence, they did not utilize the sentential context or engage in a semantic analysis of the appropriate word form. Instead they focused on the syllable level. In other words, as Nunes and Bryant suggest, children link specific spellings with particular sounds instead of their meanings. It may be that the same strategy was applied here to the spelling of homophones as these students selected plausible orthographic patterns to encode phonological structures. Even when they were unable to encode word meaning, these students did not produce implausible orthographic or phonological forms.

Another source of evidence on the greater metalinguistic load generated by the complex derivations in this study were the recursive errors found in the spellings of students in Grades 4 and 5, even when the suffix was spelled correctly as with farmasuitical. A supposition is that students understand the need to appropriately order the base and the suffix (i.e., apply general morphotactic knowledge). The base word issue, however, suggests an existing gap among concepts of pharmacy, the specificity of the semantic representation, and the coordination of these knowledge sources with the syntactic functions of the transformed meaning, pharmaceutical (Nagy et al., 2014; Silliman et al., 2018). Hence, the stabilization of cross-code relationships depends on the still unresolved dynamic interplay between word formation processes and vocabulary knowledge (Bahr et al., 2009; Berninger et al., 2010; Silliman et al., 2018). A deeper understanding of morphological word forms is critical in this process. Results further underscore the relevance of multiple language levels, from the subword to syntax levels, for attaining accuracy in derivational spellings.

Conclusions

Accurate spelling is multidimensional. It involves more than an understanding of orthographic rules. Rather, spelling accuracy develops through an increased understanding of the phoneme–grapheme relationships in a particular language as mediated by the identification of word parts (base + or − affixes) in written language, which often involves the recognition of spellings borrowed from other languages (Nunes & Bryant, 2009). As such, phonology and morphology play a significant role in the morphophonemic development of English spelling (Boulware-Gooden et al., 2015).

The availability of longitudinal data across a critical period of spelling development can provide important developmental information. However, as in the case of this study, missing spelling data due to test ceilings and the collapsing of POMplexity scores across individuals in a spelling ability group may not provide a true picture of spelling strategy use for a particular individual. In addition, misspellings from a wider variety of complex derivations, varying in transparency, would provide a clearer picture of the process of spelling development. Nevertheless, this investigation provides deeper insight into the development of cross-code relationships as applied to the spellings of complex derivations over time by students who varied in spelling ability. It is recognized that similar research methods applied to students from more diverse socioeconomic status levels, diagnosed with specific language disabilities (see Bahr et al., 2020) or spelling analyses from different morphophonemic orthographies, may result in dissimilar findings (Abbott et al., 2016).

The current results indicate how this group of students relied on phonology to form written words when linguistic knowledge, acquired from oral and written language, was inadequately consolidated. At the same time, use of familiar orthographic patterns was common, like –ough or silent –e, or real words, as with exSIGHTment, to represent phonological patterns. Moreover, some accurate derivational suffix spellings occurred early in students in Grade 3. In other words, students recognized the utility of these morphological/syntactic devices and appeared able to transfer this knowledge to new words, even when the spelling of the base word was unstable. More studies in this area are needed using different orthographies and variations in the nature of spelling/writing instruction that students receive.

Educational and Clinical Implications

Derivational misspellings, whether more or less disrupted, occur when word-general knowledge and/or word-specific knowledge is incomplete. Inadequate word-general knowledge occurs in such misspellings as exitment, which has a substituted base form, exit for excite. By comparison, insufficient word-specific knowledge is reflected in the misspelling, conchientious where gaps exist in the connections among the conceptual, semantic, and syntactic domains (Silliman et al., 2018).

The key aim for planning instruction seeks to build metalinguistic awareness of language structure at multiple language levels. As specialists in language development, speech-language pathologists (SLPs) can work with teachers to help students understand the relationships among spelling, morphology, vocabulary, syntax, reading, and writing. When spelling is viewed as more than memorization or letter–sound correspondences, students can begin to see spelling as a word formation process. It takes more than just letters and sounds to make a word. The integration of POM is essential. For instance, there are multiple ways to depict a particular sound. The [k] sound can be represented with a c, k, ck, and even ch (as in ache) or the long a vowel can be represented in the following ways: bake, weigh, day, eight, and aim. But what about a word like again? The –ai in this word does not make the long a sound and should be taught as an exception or what has been referred to as an alternation (Venezky, 1999). It is essential to teach students these alternations so that they can develop an understanding of word patterns that will assist them in reading and spelling new words. The goal is to develop word-general knowledge, which then can be perfected into word specific spellings.

Beginning in the early grades, “phonological, orthographic, and morphological awareness of word forms, their parts, and their interrelationships” (Berninger et al., 2008, p. 17) are essential for promoting word-specific spelling. There are a number of evidence-based approaches that SLPs and teachers can adapt to provide explicit, yet appealing, instruction that is crafted to developmental and educational needs. Sources include, but are not restricted to, the following: Berninger and Wolf (2016), Goodwin and Perkins (2015), Goodwin et al. (2020), Greene and Coxhead (2015), and McKeown et al. (2017). These researchers encourage SLPs and teachers to strengthen the quality of the student's lexical representation (Harris & Perfetti, 2017), if not the multidimensional nature of language, as opposed to focusing on the memorization of orthographic patterns. SLPs and teachers might consider two overlapping instructional avenues. Both invite the strengthening of word (lexical) knowledge and word formation processes but approach this goal in somewhat different ways: POM integration through pattern instruction and specific instruction in morphological processing and derivational morphology.

POM Integration Through Pattern Instruction

As mentioned earlier, English is a morphophonemic orthography, meaning that identification of word parts (M) can signal how sounds (P) of a word should be represented in the orthography (O). It is suggested that weekly spelling lists be derived from the academic vocabulary utilized in the week's lessons. This vocabulary could be taught within activities that stress word meaning and focus on teaching word parts. Meaningful writing activities would highlight the words and orthographic patterns that students misspell. Given the cognitive resources needed to produce written text, students often misspell familiar words when encouraged to communicate in an academic or disciplinary language register (Bahr et al., 2020). A reason is that writing tasks strain the available attentional and cognitive resources because writing involves all levels of language (conceptual, semantic, syntactic, morphological, orthographic, and phonological). Identification of spelling patterns within a writing task therefore provides good insight into POM integration and the quality of lexical representations.

In the upper elementary grades and even into middle school (Bahr et al., 2009), the emphasis should be on spelling within new vocabulary learning. As the data in this study showed, students will misspell new (or less lexicalized) words in different ways. Instruction then should focus on the integration of POM features instead of selecting error patterns that are unique to a specific student.

For instance, the spelling of short vowels and vowel digraphs continue to be difficult for most students (Bahr et al., 2012). To heighten metalinguistic awareness, the suggestion is to develop activities to practice POM integration by identifying spelling alternations and their exceptions. When a student misspells throw/through, this error illustrates a lack of POM integration, especially within P and O. But why might a student produce this error? One solution is that it looked right since throw is a correct spelling, even though that word does not meet the syntactic/semantic demands of the sentence that the student wrote. However, this error could also indicate a lack of integration between P and O. During word formation, the student could tap into the –ough pattern and associate it with the long [o] sound, as in though. Recognizing that though was not quite right, the student could retrieve throw, which is a more familiar word. While the explanation provided is speculative, it does illustrate the integration of POM features necessary to move from word-general to word-specific spellings. Spelling instruction that focuses on strengthening word knowledge and word formation processes helps students decode academic vocabulary that they will encounter in the upper grades.

To teach POM integration, teachers and SLPs could focus on multiple spellings of particular phonological and/or orthographic patterns. Sorting tasks using words that contain the same P or O patterns could be used to assist students in pattern identification while simultaneously strengthening metalinguistic awareness. An example of word types that stress either P or O patterns is provided below using the –ough pattern (see Table 7).

Table 7.

Examples of word lists that target a particular P (phonological) or O (orthographic) spelling pattern.

/u/ pattern (focus on P)	–ough pattern (focus on O)
food [fud]	bough [baʊ]
two [tu]	cough [kaf]
dude [dud]	dough [do]
you [ju]	ought [ɔt]
through [θru]	rough [rʌf]
brew [bru]
blue [blu]

Teachers/SLPs would provide lists of words, easily found on the Internet, that follow these patterns. Students would be asked to sort the words into patterns, focusing individually on the P or O target. As a result, students become familiar with alternations, as well as the exceptions to the patterns, as the –ou in could is different from you. This type of activity is useful to all students within the class and can identify students who need more structured assistance. Similar activities could be developed at the M level, using derivational words that alter P (as in electric–electricity) and P–O (as in strong–strength) relationships. For additional guidance on word selection, see Nagy and Hiebert (2011).

Instruction in Morphological Processing and Derivational Morphology

Although students can learn spelling patterns from weekly spelling lists, instruction that is more effective involves engaging them in identifying letter patterns and word parts, including affixes. A focus on derivational morphology can be helpful in teaching students how to coordinate various levels of language more explicitly. Goodwin et al. (2020) propose the following activities to help students develop enhanced metalinguistic knowledge by becoming “word detectives,” which is a form of morphological problem solving connected to the curriculum (see Appendix).

For effective use of morphological problem solving and the word detective concept, the presumption is that SLPs can work with teachers to assist them in understanding the role of morphological processing and the word formation processes. For example, SLPs can help teachers by identifying, with morphological problem solving, how poor spellers may be refining their strategies in productive ways to spell derivations over time.

Regardless of the approaches selected, the instructional aim is to build into the existing curriculum the kind of spelling and writing instruction that facilitates high-quality POM representations across multiple language levels. POM representations are the basis of more literate derivational knowledge in academic content areas.

Appendix

Word Detective Activities (Goodwin et al., 2020)

1. Identify units of meaning (i.e., morphemes) when encountering a new word to fortify morphological knowledge. This involves locating affixes and bases but could also extend to the recognition of phonemes and orthographic patterns associated with particular sounds.

2. Distinguish base words to strengthen lexical knowledge (see McKeown et al., 2017, for engaging activities). For instance, recognizing that probable is related to probably but not problem draws attention to the need for students to look for word parts and not just spell out individual syllables.

3. Use suffixes as devices to vitalize syntactic knowledge. Sentence frames and base words can be provided for students to complete sentences with the appropriate base + suffix construction. For instance, the student could be asked to complete the following sentence using a target base word + suffix: The workers requested a ____________ of the new computer program (demonstrate). For a variety of sentence construction and combining activities that can be adapted to derivational morphology activities, see Saddler (2018).

Funding Statement

The authors acknowledge Grant HD25858 from the National Institute of Child Health and Human Development, which supported the data collection that identified the three spelling ability groups and spelling errors for analysis in the current study. They also appreciate the work that Ellen St. John did in scoring all the spelling words and preparing the data for statistical analysis.