ORIGINAL RESEARCH published: 05 June 2020 doi: 10.3389/fpsyg.2020.01071 Non-phonological Strategies in Spelling Development Naymé Salas* Departament de Didàctica de la Llengua i la Literatura, i de les Ciències Socials, Facultat de Ciències de l’Educació, Universitat Autònoma de Barcelona, Bellaterra, Spain This paper investigated the role that types of knowledge beyond phonology have on spelling development, such as knowledge of morpheme-to-grapheme mappings, of orthographic patterns, and of word-specific orthographic patterns. It is based on the modern view that children do not learn spelling in discrete stages but, rather, they apply different types of strategies from early on. The goals of the paper were threefold: (1) to determine the relative difficulty of different types of non-phonological spelling strategies, (2) to examine the contribution of non-phonological strategies (specifically, morphological, morphophonological, orthographic, and lexical) to conventional spelling scores, and (3) to determine the role of children’s educational level and population type (first- vs. second-language learners) on spelling strategy use. A large sample of 982 children (497 boys), speakers of Catalan (a Romance language similar to Spanish but with a less consistent orthography), participated in the study. They were administered a bespoke dictation task aimed to test their conventional and phonographic accuracy skill, as well as to determine their ability to use different types of non-phonological strategies Edited by: for the spelling of ambiguous phonemes. Data were analyzed with a series of multigroup, Angela Jocelyn Fawcett, Swansea University, United Kingdom multilevel SEMs. Results showed that (1) children across groups found morphological Reviewed by: and lexical strategies harder to apply than orthographic and morphophonological Janina Kahn-Horwitz, strategies and (2) all types of non-phonological strategies contributed greatly to spelling Oranim Academic College, Israel Heikki Juhani Lyytinen, accuracy scores, even after controlling for children’s phonographic skills. Efficient University of Jyväskylä, Finland strategy use increased as a function of schooling level, while second-language learners Cristina Burani, had a worse performance throughout, but no group showed a specific pattern of Italian National Research Council, Italy results. In conclusion, the paper offers substantial evidence that non-phonological *Correspondence: Naymé Salas strategies are paramount to learning to spell at least during the early and intermediate nayme.salas@uab.cat elementary school years. It is suggested that the teaching of writing should therefore Specialty section: be multidimensional in nature and target particularly the strategies with which children This article was submitted to struggle the most: knowledge of morpho-graphemic mappings and word-specific lexical Educational Psychology, representations. Theoretical implications are also discussed. a section of the journal Frontiers in Psychology Keywords: spelling, morphology, orthographic constraints, orthographic representations, Catalan Received: 25 December 2019 Accepted: 27 April 2020 Published: 05 June 2020 INTRODUCTION Citation: Salas N (2020) Non-phonological Learning to spell is a process of a phonological nature (Read, 1971; Treiman, 2004). There Strategies in Spelling Development. is substantial evidence, however, that accurate spelling also requires accessing and applying Front. Psychol. 11:1071. other types of knowledge beyond phonology, such as knowledge of morpheme-to-grapheme doi: 10.3389/fpsyg.2020.01071 mappings (e.g., Nunes et al., 1997a; Pacton and Deacon, 2008) and of orthographic patterns Frontiers in Psychology | www.frontiersin.org 1 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies (e.g., Treiman, 1994; Cassar and Treiman, 1997; Pacton et al., consistent orthographies, such as Spanish (e.g., Defior et al., 2008: 2002; Deacon et al., 2008). This holds true particularly in highly Suárez-Coalla et al., 2017), Italian (e.g., Angelelli et al., 2014, inconsistent orthographies, like English (Borgwaldt et al., 2004; 2017), or Finnish (e.g., Lehtonen and Bryant, 2005). Defior et al. Caravolas et al., 2012), that often need to resolve the ambiguities (2008), for example, studied the spelling strategies of children generated by the multiple alternative spellings for a single who speak a regional dialect of Spanish (Andalusian) in which phoneme. Nonetheless, users of more consistent orthographies some consonant endings are not pronounced (e.g., /s/ in coda have also been shown to apply non-phonological strategies position). They asked children in grades 1–3 to spell words in during spelling (e.g., Defior et al., 2008; Carrillo et al., 2013; two conditions: one in which the final /s/ belonged to a verbal Alegría and Carrillo, 2014; Carrillo and Alegría, 2014; Rothe et al., morpheme, as is the case with the in tienes “have_2nd 2014; Marinelli et al., 2017; Angelelli et al., 2018; Zarić et al., person singular” and a control condition in which the /s/ did 2020). In this paper, we examined the use of non-phonological not have any morphological bearing, as is the case with the spelling strategies in Catalan-speaking children in the early and in lunes “Monday.” Children across grades were more prone to intermediate elementary school years. spelling the silent in the morphologically bound condition than the reverse. Models of Spelling Development In sum, there is abundant evidence that children are sensitive Early models of spelling were concerned with identifying the to regularities beyond phonology and that they apply them various phases in its development. A common trait in stage- from very early on. However, only a few studies have addressed like theories of spelling was that the various stages were the extent to which these strategies are applied successfully articulated around the role of phonology. For example, most at different educational levels. Moreover, a majority of studies models distinguished between a pre-phonological phase (i.e., pre- on the use of non-phonological spelling strategies has been literate), where the nature of the link between language and conducted in English or French, both languages with highly graphemes is unknown to the child; a phonological phase, where inconsistent phonographic mappings (Borgwaldt et al., 2004; the child has grasped the alphabetic principle (Byrne, 1998) and Caravolas et al., 2012), which could arguably make such strategies becomes increasingly more able to represent the phonological indispensable, in contrast to more consistent orthographies that structure of words; and a “beyond phonology” phase, in which could rely on phonological representations to a much larger the child recruits the necessary (non-phonological) knowledge degree. In this study, we examined the use of non-phonological to arrive at the conventional spelling of words (e.g., Gentry, spelling strategies in early- (grade 2) and intermediate-level 1978; Henderson and Beers, 1980; Frith, 1985). By collapsing all (grade 4) speakers of Catalan, a Romance language spoken non-phonological aspects into a single, later stage, these theories in Barcelona (Spain), with a semi-transparent orthography did little to accommodate their precise nature and their role in (Llauradó and Tolchinsky, 2016). learning to spell. Currently, the widely accepted view on spelling development Spelling Development in a Second is that children are sensitive to non-phonological information right from the start of the learning process and that different Language spelling strategies overlap throughout development (Rittle- Literacy skills have often been regarded as transferable across Johnson and Siegler, 1999; Treiman, 2017). What is more, the languages spoken by an individual (e.g., Cummins, 1979). sensitivity to non-phonological aspects of spelling, such as the However, research on second-language (L2) spelling has reported “outer form” of words, actually precedes phonological processes conflicting findings. On the one hand, L2 learners show low (Treiman, 2017, p. 4). For example, even pre-phonological spelling accuracy levels and appear to develop at a slower rate, in spellers develop ideas about the number of graphemes that words comparison with their L1 peers, but their spelling performance have, which are consistent with the average word length in the appears to be driven by similar skills (Geva et al., 1993; language to which they are exposed (e.g., Ferreiro and Teberosky, Verhoeven, 2000). Moreover, L2 spelling can be explained, to a 1979). They have also been found to develop orthographic great extent, by L1 spelling skills (Sparks et al., 2008), in line with awareness skills prior to their understanding of the alphabetic Cummins’ (1979) assertion. Conversely, some studies reported principle; for instance, they are sensitive to the positions at which that the skills underlying L1 and L2 spelling differ over time the orthography allows letter doubling and those at which it does (Jongejan et al., 2007), where at least part of the differences may not (Cassar and Treiman, 1997). be related to the characteristics of the L1 writing system (Martin, Children appear to be sensitive to morpho-graphemic 2017). Importantly, studies of brain dynamics have shown regularities in spelling as well (e.g., Nunes et al., 1997a,b; varying patterns of EEG activity as a function of population type Sénéchal, 2000; Deacon et al., 2008). Treiman and Cassar (1996) (i.e., L1 vs. L2), especially during the spelling of words that require showed that young children, aged 5–9 years, were more prone non-phonological strategies (Weber et al., 2013). In this study, to spelling complex consonant clusters correctly when one of the we compared children with (L1) and without (L2) exposure to consonants involved a past tense morpheme (e.g., passed > past). Catalan, the language of instruction, outside school. Although it could be argued that complex spelling strategies, such as morphological knowledge, are mostly applicable in highly Assessing Spelling inconsistent orthographies, such as French or English, there The analysis of misspellings allows understanding the type of is evidence of the use of morphological strategies in highly strategies that children apply during spelling. Several schemes Frontiers in Psychology | www.frontiersin.org 2 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies have been proposed that have different outcomes and goals. et al., 2009, 2012). Investigations using the POMAS scheme One type of assessment focuses on giving children (partial) classify misspellings as phonological when they omit, add, or credit for misspellings that to some extent reflect the underlying replace a letter, so that the resulting word does not preserve the phonological structure of the target words. For example, phonological structure of the target word, as in ∗ for some schemes evaluate whether children’s written productions worked. Orthographic errors are those in which the misspelling represent all phonemes in the target word, using a letter or does not observe orthographic patterns or constraints, such grapheme that could represent the intended phonemes in some as ∗ for worked; they may also include ambiguous context, even if the resulting production is unconventional, letters, such as writing ∗ for worked. Finally, when such as writing ∗ for fight (e.g., Ritchey et al., 2010; the child’s production has ignored a morpho-graphemic mapping Treiman et al., 2019). Other assessment schemes evaluate the that could resolve an ambiguity, such as ∗ for worked, degree of phonological proximity of the misspelling to the the error is classified as morphological. Morphological errors target word. Caravolas et al. (2001) scored each sound segment are evaluated on both inflectional and derivational morphemes, and their corresponding grapheme on a 0–4 scale, so that as well as in word roots (Bahr et al., 2012). Besides the any plausible spelling of a specific sound got the maximum classification of spelling errors under these three categories, score, and close approximations (e.g., graphemes that represent the POMAS allows further specifications within categories, a phoneme that differs from the target one in a single feature) identifying specific mistakes of each kind. In the examples above, were scored lower, while omissions were given a score of 0 would be further classified as a problem with an initial (Caravolas et al., 2001, p. 758). Phonology-centered assessments obstruent (Silliman et al., 2006, p. 111). In the case of , of spelling have been particularly insightful when evaluating it would be classified as an error that uses an ambiguous letter: children in the early stages of learning to spell, when they are both and can represent /k/, as in cattle and kettle. The expected to use a phonological strategy to spell both known case of is a morphological error that would be further and unknown words. Even most spelling instruction programs classified as a problem with (past tense) inflections (Bahr et al., recommend beginning with phonologically based strategies [e.g., 2012, p. 22). National Reading Panel, 2000; Alves et al., 2018]. Typically, One of the advantages of a framework such as the POMAS these assessment schemes do not penalize misspellings for not is that it allows for a comprehensive analysis of both the observing orthographic constraints, since the aim is to determine phonological and non-phonological strategies that children use the extent of children’s phonographic skills. for spelling. Another advantage is that it allows using fine-grained As reviewed above, there is abundant evidence of children’s analysis criteria as much (or as little) as the research requires. early sensitivity to orthographic regularities. Therefore, For these reasons, the POMAS should be helpful in accumulating assessment proposals that focus on evaluating children’s data from different languages, given that the definition of the knowledge of orthographic patterns or their familiarity with main categories should be applicable and comparable cross- the “outer-form” of words are highly valuable (Treiman, 2017). linguistically, while the specific phenomena evaluated within each Letter-based schemes are characterized by giving children category may be adjusted to each language. credit for partial success in representing the conventional form of printed words. One of such attempts, for example, consists in awarding points for each two-letter sequence that Adapting POMAS to Assess Catalan is accurately (i.e., conventionally) represented, plus points Spelling for conventionally represented initial and final letters (Frisby, In this study, we made two adaptations to the POMAS scheme to 2016). This type of schemes is useful to tap into children’s suit the specific characteristics of Catalan, as well as to answer knowledge of orthographic patterns and to test the strength of specific research questions. Catalan is a Romance language their orthographic representations in greater detail than a simple spoken in the region of Catalonia, Spain, where the vast majority correct/incorrect spelling measure. of people are speakers of, at least, Spanish and Catalan. Catalan is Other spelling assessment proposals evaluate a combination also the language of instruction throughout preschool, primary, of phonological and orthographic knowledge. For example, and secondary education. It is also used in most university Treiman et al. (2016) assessed children’s spelling considering education. Catalan’s morphosyntax is similar to Spanish, in that both phonographic and orthographic skills. Words spelled they both have a rich morphological system, particularly in with all conventional letters received more points than verbs, and relatively free word order with subject elision (pro- productions that were spelled with phonologically plausible, but drop nature, Bel, 2003). Syllable structure allows, like Spanish, unconventional, letters. only up to two consonants in onset position, and it is slightly Yet other spelling assessment schemes pay attention to more complex in coda position than Spanish, allowing up to children’s use of knowledge sources beyond phonology and three consonant sounds, as in boscs/bosks/“forests.” In the region orthography (e.g., Masterson and Apel, 2010; Bahr et al., 2012; where the data were collected, Barcelona, Catalan uses eight Lee and Al Otaiba, 2017). Bahr et al. used a framework for vowel sounds (three more than Spanish) and has vowel-reduction assessing spelling rooted on triple word-form theory, which takes processes, such that in unstressed syllable position only three into account phonological, orthographic, and morphological vowel sounds may occur: /@, i, u/(Prieto, 2004). This leads to strategies, the Phonological, Orthographic, and Morphological several spelling ambiguities, because there are multiple possible Assessment of Spelling (POMAS; Silliman et al., 2006; Bahr representations for each of these three phonemes. In addition, Frontiers in Psychology | www.frontiersin.org 3 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies some consonants are silent: is always silent, is silent knowledge. This category has been proposed in previous studies after a nasal sound at the end of words, as in vent, “wind”; of spelling in Dutch and Hebrew (Gillis and Ravid, 2006). caminant, “walking”; and estudiant, “student.” The letter Morphophonological representations occur in languages that is also silent when word final, as in primer, “first,” or cantar, have productive phonological processes that are triggered as a “sing.” Several consonant sounds may be represented by two or result of a morphological change. In Catalan, some words include more graphemes. For example, /b/may be written as or a final occlusive phoneme that is dropped from pronunciation , as in vaixell, “boat,” and bèstia, “beast”; likewise, the palatal, (i.e., it has a zero-realization allophonic variant) only in the fricative, voiced sound/Z/may be represented by , as in case of masculine, singular nouns and adjectives, but that is albergínia, “aubergine,” and as , as in jove, “young.” Some of pronounced in other forms of the word. For example, vent “wind” these inconsistencies can be resolved by applying morphological is pronounced [ben] in its masculine, singular form, although it strategies; for example, the silent in caminant indicates that is spelled with a silent . In other, usually derived, forms of the the word is a verb in gerund form. Therefore, if the speller is word, such as ventós, “windy,” or ventet, “wind.small,” the is aware of the link between this morpheme, /-an/, and its spelling, pronounced. This means that the phonological representation of <-ant>, the ambiguity disappears. Other inconsistencies can be the masculine, singular word must carry the final stop phoneme, resolved by applying orthographic knowledge, that is, knowledge /bent/, which is dropped only in this version of the word, but of orthographic patterns and constraints. For example, the use that resurfaces in every other word form. This category entails of or to represent phoneme /Z/ can be disambiguated both access to phonological representations, in combination with by taking into account the following vowel sound: front vowel knowledge of morphological processes; thus, because it involves sounds /e, i, E/ combine only with , while all other vowel the recruitment of strategies beyond phonology, we included it in sounds, combine only with . Thus, the correct spelling the present study. of /Z/ in albergínia and jove requires accessing and applying knowledge of orthographic patterns. Finally, certain words or parts of words require a full orthographic representation that, Previous Research on Spelling often, may need to be paired with a semantic representation. As Strategies such, they require memorizing word-specific spellings, since they Some studies using POMAS or evaluating both phonological and are not governed by a morphological or an orthographic rule, and non-phonological spelling strategies have examined the relative phonographic knowledge is necessary but insufficient to arrive at difficulty of applying them. In general, all strategy types are the conventional spelling. This is the case of in vaixell, of used even by the youngest participants (e.g., first graders), but in bèstia, the first in cantar, and many others. there are differences in the rate and developmental route for Previous studies using POMAS collapsed irregular or each type of strategy. Orthographic errors, such as word-specific spelling knowledge together with knowledge of or for worked, predominate across grades, while orthographic patterns into a single category of “orthographic morphological and phonographic errors are the least frequent knowledge” (e.g., Silliman et al., 2006; Bahr et al., 2012). (e.g., Bahr et al., 2012; Benson-Goldberg, 2014; Llauradó and A first adaptation of the present study is, thus, to differentiate Tolchinsky, 2016; Joye, 2019). Llauradó and Tolchinsky (2016), between these two types of knowledge sources. We will classify who analyzed the spelling errors of 225 Catalan-speaking as “orthographic” those errors in which the ambiguity in the children in grades 1–5, found that, from grade 2 onward, representation of a phoneme can be resolved by resorting orthographic () and lexical () mistakes to knowledge of legal orthographic patterns, and of context- were the most frequent and were produced at a similar rate, while dependent rules. We will classify errors as “lexical” when they differed significantly from both phonographic () the ambiguity in the representation of a phoneme cannot be and morphological () mistakes, which were the least resolved by resorting to rules that can be generalized to several frequent and did not differ from each other. A recent study tokens and, rather, the speller needs to resort to word-specific that developed a pseudo-word spelling task based on triple spelling patterns. Often, this will involve an association between word-from theory reported that phonographic strategies showed the semantics or the lexical representation of a word and its little variation over time, while there were significant differences spelling. For example, /hi:l/ associated with the meaning, “to across grades for orthographic and morphological errors, which become healthy,” is spelled , while associated with the decreased over time. This general pattern, according to which meaning “back part of the human foot” is spelled . error rate is highest for orthographic errors, and lowest for A similar distinction was used in another study, also on Catalan morphological and, especially, phonographic errors, has received (Llauradó and Tolchinsky, 2016). In short, our adaptation entails support from cross-linguistic studies (e.g., Joye, 2019) and that the “orthographic” category in POMAS is subdivided into from investigations comparing dyslexic to typically developing “orthographic” and “lexical” strategies, where the former are controls (e.g., Baseki et al., 2016). characterized by context-dependent rules and knowledge of legal A common trait of previous research using POMAS is sequences of letters, whereas the latter require rote memorization that most studies analyzed naturalistic or semi-naturalistic text of words or parts of words, with or without association to a production data (e.g., Bahr et al., 2012, 2015; Llauradó and semantic representation. Tolchinsky, 2016). While ecologically valid, this procedure has Our second adaptation of POMAS in the present study the disadvantage that children may choose less complex or is the addition of a new category of morphophonological more familiar words (Graham and Harris, 2005), and a different Frontiers in Psychology | www.frontiersin.org 4 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies number of opportunities may be created for each type of strategy 2012; Llauradó and Tolchinsky, 2016). In addition, it has been to be applied. In addition, analyzing misspellings in such a suggested that speaking a morphologically rich language, as is context also poses the challenge of having to determine whether the case of Catalan, would facilitate the use of morphological one or more categories apply to a specific error (Bahr et al., strategies (Gillis and Ravid, 2006). In contrast, we expected that 2012, p. 8). For this reason, this study used a bespoke dictation the application of context-dependent rules (i.e., orthographic task, in which an equal number of opportunities were created for strategies) would be harder, because their overall incidence using each of the non-phonological strategies, while the internal is lower than both morphological and morphophonological characteristics of the word items (frequency, syllabic complexity, processes. Finally, we expected that word-specific spellings (i.e., length) were counterbalanced. lexical strategies) would be the type of strategy that would be used To the best of our knowledge, previous research using POMAS less successfully across grades and population types, given that has not been used to determine the relative contribution of they are, by definition, only learned through rote memorization. each type of knowledge source or strategy (i.e., phonographic, Our second research question was (RQ2), what is the orthographic, morphological) to conventional spelling. This is relative contribution of non-phonological spelling strategies to an important point, not only because spelling is paramount conventional spelling scores? In this sense, we expected that all to writing quality (e.g., Berninger and Winn, 2006) but also strategy types would have a significant contribution, in line with because conventional spelling is the most sensitive measure to modern views of spelling development that pose that children predict later literacy gains (Treiman et al., 2019). Arguably, are sensitive to various sources of knowledge from early on (e.g., then, understanding the impact that different types of strategies Treiman, 2017). have on children’s spelling skills should be instrumental in Our third research question was (RQ3), does the use of determining their importance across a key developmental stage non-phonological spelling strategies differ for L2 learners or as (beginning and intermediate elementary grades) and to orient a function of children’s educational level? With regard to L2 teaching practices into which strategies require the most attention spelling, we expected that children with no exposure to Catalan from practitioners. outside of school would have poorer linguistic and orthographic representations and, therefore, would show more difficulty across This Study all categories. Although the available evidence comparing L1 to The present study improves on previous ones in a number of L2 spelling is conflicting, we sided with previous literature that ways. First, it uses a controlled elicitation procedure, a dictation poses that, despite differences in performance, a similar pattern test, instead of naturalistic or semi-naturalistic text production, of spelling mechanisms underlies both L1 and L2 spelling (e.g., so that an equal number of opportunities for using each type Geva et al., 1993; Verhoeven, 2000). Given that both linguistic of strategy were created. In addition, our test measure allowed and orthographic representations would be less strong in L2 us to counterbalance word frequency and complexity. Second, than in L1 children, no single strategy type stood out as more it distinguished between rule-bound non-phonological strategies problematic for our L2 participants. Finally, we expected that from those that involve memorizing word-specific patterns, not fourth graders would use all non-phonological spelling strategies generalizable to other words. Third, in order to truly assess the more successfully than second graders, but that a similar pattern contribution of non-phonological spelling strategies to spelling of difficulty would emerge in both grade levels, in light of previous development, all analyses were carried out controlling for each studies on a similar population (Llauradó and Tolchinsky, 2016). child’s phonological accuracy skills. We also included some key demographic variables, sex and parents’ socioeconomic status (SES), as control variables, so as to determine their potential MATERIALS AND METHODS effect (Allred, 1990; Aram and Levin, 2001) and examine the unique contribution of spelling skills and strategies. Fourth, we Participants investigated a relatively unexplored language, Catalan, which has Students were 982 children (497 boys), attending second (494) a spelling system much more consistent than English, although it and fourth (488) grade at schools in the province of Barcelona is not exempt from complexities. Finally, we examined spelling- (Spain). All students were speakers of Catalan, and they were strategy use as a function of children’s exposure to the language of all assumed to be bilingual. Barcelona is a bilingual community, instruction, in order to contribute to the field of second-language where Catalan is the main language of instruction in elementary spelling and, more specifically, to the development and teaching education. All students were administered a sociolinguistic of spelling in different learning contexts. questionnaire, which they completed with the help of their The study was articulated around three main research teachers. We were particularly interested in the extent to which questions: (RQ1) What is the relative difficulty in the application children used Catalan outside school, so they were asked to of non-phonological spelling strategies? We hypothesized that declare the language(s) they used with each of their parents, morphophonological strategies would be easiest, because they with siblings, and with friends. Based on this information, we require accessing a productive process that is triggered very classified children as being exposed to Catalan outside of school frequently both in speech and spelling. We also expected or not. Table 1 provides the demographic information and morphological strategies to be among the easier ones, given that distribution of the sample. previous studies found that morphological errors were the least The study belongs to a larger project on writing development or second least frequent error type across grades (Bahr et al., that had obtained full clearance by the Ethics Committee of the Frontiers in Psychology | www.frontiersin.org 5 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies TABLE 1 | Participants’ distribution and demographic information. for frequency: 16 low-frequency items, 16 high-frequency items; for length, with short words consisting of one or two syllables and Group Number of Mean age (SD) Boys/girls Mean SES long words consisting of three to four syllables; and for syllabic participants (SD) complexity: simple words typically consisting of open syllables, Cat0-G2 238 7;5 (0;4) 113/125 41.24 usually CV in structure, while complex words had, at least, two (11.91) closed (e.g., CVC) syllables or at least one consonant cluster Cat0-G4 214 9;3 (0;8) 118/96 39.70 (e.g., CCV, VCC). All items were validated by two experienced (10.50) elementary-school teachers, who confirmed the perceptions of Cat1-G2 256 7;4 (0;4) 125/131 52.10 frequency and difficulty. (15.38) In the morphophonological category, all eight items had a Cat1-G4 274 9;4 (0;4) 141/133 53.05 (14.59) silent that could be recovered in derived words. In the morphological category, four items tested the silent in the G2, grade 2; G4, grade 4; Cat0, no Catalan exposure outside school; Cat1, some degree of Catalan exposure outside school (e.g., child speaks Catalan with mother). gerund form of verbs, and the other four items tested the spelling of the plural of feminine nouns, which contains an ambiguous vowel sound, /@/, that is always spelled . In the orthographic University. Children were recruited from 13 public schools. All category, seven items tested vs. alternation, which is children in the classroom were approached, and we collected regulated by the following vowel, and one item tested knowledge data only from those whose parents returned signed consent that the spelling of the sound /z/ after a consonant sound is forms. No children were excluded from the study on the basis always . Finally, the lexical category tested some of the most of learning disorders. common homophonous letters, when they are not governed by other rules: , , , . Task Design and Coding Several scores were obtained from the task: (1) a conventional All the children completed the same bespoke task, which involved spelling accuracy score, (2) phonographic plausibility, (3) use spelling 34 words that were dictated orally by the administrator. of morphological strategies, (4) use of morphophonological Only 32 items were used here, as two items targeted stress- strategies, (5) use of orthographic strategies, and (6) use of mark spelling, which was outside the scope of the present study. lexical knowledge. Each score was determined independently The items were selected from a corpus of Catalan children’s of the rest, in a binary fashion. Conventional accuracy and spelling (Llaurado et al., 2012), available online at http://clic.ub. phonological plausibility scores were determined for all 32- edu/corpus/en/cesca-en. This corpus includes the words that a word items, whereas for each type of non-phonological strategy, sample of more than 2000 children aged 5–16 wrote in response the score was the total correct out of the eight items in the to a semantic category (e.g., clothes, traits of character, natural category. Cronbach alpha reliability for the test was 0.92. An phenomena, food). Words are lemmatized and can be looked up external research assistant, uninvolved in the present research, according to their overall frequency, the educational level of the scored all words, and the author rescored 28% of a random writer, and they are moreover listed according to the number of sample. Inter-rater reliability [intra-class correlation (ICC)] was alternative (mis)spellings produced. Words that appeared in the excellent: 0.989. 90th percentile or higher were considered high-frequency words, and words in the 10th percentile or lower were considered low- Procedure frequency words, as long as they belonged to the corresponding Children were tested in their regular classrooms. They were given category within which they were produced1 . a lined paper, with a dot next to which they had to write each The task was designed as follows: 32 items were selected so word, one below the other. The administrator explained that that at least one phoneme was ambiguous (i.e., it had two or they were going to do a dictation task and that they would more alternative spellings). A single phoneme was targeted in hear each word three times: first in isolation, then in a carrier each item, whose ambiguity could be eliminated by resorting sentence, which would help identify the word used in context. to one of four different sources of knowledge: morphological, Finally, they would hear it one final time, again in isolation. morphophonological, orthographic, or lexical. There were eight For example, the administrator said, “Vent. Si fa vent, podrem items per category. In addition, all items were scored as anar a navegar. Vent.” “Wind. If there’s wind, we can go sailing. correct or incorrect in terms of their phonographic plausibility, Wind.” Children were given a few seconds to spell each word. that is, whether the spelling was an accurate phonological They were encouraged to write them as best they could, even if representation of the intended word, regardless of positional they were unsure, and told not to worry if they did not know or other orthographic constraints. Finally, items were scored as how to spell correctly. Testing lasted between 20 and 30 min, correct or incorrect in terms of conventional accuracy (i.e., as a time that all teachers reported they were used to engaging in they would appear in a dictionary). Items were counterbalanced writing activities. All test administrators received training as to how to deliver 1 Sometimes children would produce a word in response to, for example, traits of the sentences in terms of speed, rhythm, and emphasis. They character, that had an overall low frequency, such as beautiful. These instances practiced several times until it was clear that they were all reading were ignored, as the resulting word may not necessarily be a low-frequency token in absolute terms. Rather, they seem to be productions that failed to capture the the items using a similar, natural tone, and without specifically semantic field that was required. emphasizing any of the words. Frontiers in Psychology | www.frontiersin.org 6 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies RESULTS TABLE 3 | SEM comparative coefficients between types of strategies. Group L-O M-O MP-O L-M L-MP M-MP An inspection of the data showed that they were normally distributed, with skewness values below 3, and kurtosis G2 Cat0 −0.07* −0.10* 0.13* 0.04* −0.19*** −0.23*** values under 10 (Kline, 2011). We conducted a series (0.02) (0.02) (0.03) (0.02) (0.02) (0.02) of analyses dividing children into four groups: grade 2 G2 Cat1 −0.27* −0.21* −0.01 −0.03 −0.23*** −0.20*** (0.02) (0.02) (0.02) (0.02) (0.02) (0.01) without Catalan exposure, grade 2 with Catalan exposure, G4 Cat0 −0.11* −0.14* 0.12* 0.04* −0.22*** −0.26*** and analogous grade 4 groups. (0.02) (0.02) (0.02) (0.02) (0.02) (0.02) G4 Cat1 −0.14* −0.11* 0.12* −0.02 −0.25*** −0.23*** Development of Non-phonological (0.02) (0.02) (0.02) (0.02) (0.01) (0.01) Spelling Strategies G2, grade 2; G4, grade 4; Cat0, no Catalan exposure outside school; Cat1, We first addressed the difficulty of applying the various some degree of Catalan exposure outside school (e.g., child speaks Catalan with mother); O, proportion of correct use of orthographic knowledge; MP, proportion non-phonological spelling strategies in each group. We ran of correct use of morphophonological knowledge; M, proportion of correct use preliminary one-way, repeated-measures ANOVAs to determine of morphological knowledge; L, proportion of correct use of lexical knowledge. whether the type of spelling strategy that needed to be ***p < 0.001 and *p < 0.05. applied in each group of eight words affected performance. Results showed that the type of strategy did have a significant order of difficulty was the same reported in the ANOVAs effect on its successful application, with moderate effect (from easiest to hardest): morphophonological, orthographic, sizes across groups (Table 2). In general, application of a morphological, and lexical. All comparisons between each pair morphophonological or an orthographic strategy was easier, of strategies were significant across grade levels. In the case whereas lexical and morphological strategies were harder of children with exposure to Catalan outside school, there strategies across groups. It was apparent, however, that there were were only subtle variations. For second graders, letters that differences between groups. required a morphophonological or an orthographic strategy We next conducted a more comprehensive two-level were equally difficult; similarly, applying a morphological or structural-equation model (SEM), in which students were lexical strategy was equally hard. These two pairs of strategies, repeatedly measured on the four types of words. Thus, the level orthographic and morphophonological, on the one hand, and 1 data were the item-level performance and the level 2 data lexical and morphological, on the other, were significantly were the students. Moreover, this modeling strategy was run different from each other. In the fourth grade, the situation within a multiple-group comparison framework, where we used was similar, except that these children found that applying the division of children into the four groups specified above a morphophonological strategy was significantly easier than and tested the effect of Strategy Type on performance. The applying an orthographic one. model included sex, parents’ SES, and each child’s phonographic As the SEM model was built of two-level data, we estimated accuracy score as control variables. The overall unconstrained the impact of students’ characteristics on performance. The model goodness of fit was excellent, Chi2 (8) = 6.96, p = 0.541; ICCs of the success rate of Strategy Type were above 0.30, RMSEA = 0.000; CFI = 1.00; TLI = 1.00; SRMRwithin = 0.000; which emphasized the potential of variability across students. and SRMRbetween = 0.019. In each group, we were able to In all four groups separately, students’ sex and parents’ SES show pairwise comparisons between effects (Table 3). For did not have a significant effect on performance. Phonographic both grades 2 and 4 with no Catalan outside of school, the accuracy, in contrast, did have a significant effect, so that children with higher scores on phonographic accuracy tended to apply non-phonological spelling strategies more successfully. Adding TABLE 2 | Comparison between performance measurements of different word types by grade level and exposure to Catalan. Strategy Type as an independent variable considerably improved the explanatory power of the model, as shown by the notable Group Lexical Morph MPhon Ortho df F η2p increase pseudo R2 values of the final model vs. the model without level 1 indicators, from a.14 to a.21 increment (Table 4). G2 Cat0 0.35a 0.31a 0.53c 0.42b 3633 44.32*** 0.17 (0.23) (0.29) (0.29) (0.35) These findings were consistent across the four groups. Students’ G2 Cat1 0.32a 0.35a 0.55b 0.56b 3726 79.77*** 0.25 demographic characteristics did not impact their performance, (0.22) (0.28) (0.30) (0.32) while complementary indicators did. G4 Cat0 0.51a 0.48a 0.74c 0.62b 3579 72.12*** 0.27 Because there were remarkable similarities across groups in (0.22) (0.29) (0.24) (0.31) terms of (1) the pattern of relative difficulty for each spelling G4 Cat1 0.53a 0.55a 0.78c 0.67b 3783 96.98*** 0.27 strategy; (2) the influence of demographic variables; and (3) (0.22) (0.28) (0.23) (0.28) the role of phonological accuracy, we ran alternative, more G2, grade 2; G4, grade 4; Cat0, no Catalan exposure outside school; Cat1, some parsimonious models in which we constrained either Educational degree of Catalan exposure outside school (e.g., child speaks Catalan with mother); level or Exposure to Catalan to be equal across groups. These Ortho, proportion of correct use of orthographic knowledge; MPhon, proportion of models, however, were a significantly worse fit to the data, correct use of morphophonological knowledge; Morph, proportion of correct use of morphological knowledge; Lex, proportion of correct use of lexical knowledge; Chi2 (6) > 30.21, p > 0.05, supporting the between-group Small Latin letters for mean ranking, from the lowest (a) to the highest. ***p < 0.001. differences reported above. Frontiers in Psychology | www.frontiersin.org 7 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies TABLE 4 | Structural equation modeling results for level 2 student’s effects by grade and exposure to Catalan. Unconditional Conditional model with no Final model model strategy type Group Sex SES Phono. ICC Within Between Within Between Pseudo Within Between Pseudo Acc. variance variance variance variance R2 variance variance R2 G2 Cat0 0.03 −0.002 0.84*** 0.383 0.057 0.035 0.056 0.013 0.26 0.046 0.015 0.40 (0.02) (0.001) (0.06) (0.003) (0.005) (0.003) (0.003) (0.003) (0.003) G2 Cat1 −0.02 0.000 0.77*** 0.297 0.065 0.027 0.064 0.010 0.20 0.049 0.014 0.38 (0.02) (0.001) (0.06) (0.003) (0.004) (0.003) (0.002) (0.002) (0.002) G4 Cat0 0.01 0.00 0.97*** 0.367 0.051 0.030 0.052 0.006 0.28 0.038 0.010 0.49 (0.02) (0.001) (0.07) (0.003) (0.004) (0.003) (0.002) (0.003) (0.002) G4 Cat1 −0.02 0.001 0.93*** 0.336 0.050 0.025 0.050 0.004 0.28 0.037 0.008 0.48 (0.02) (0.001) (0.06) (0.002) (0.004) (0.002) (0.002) (0.002) (0.002) G2, grade 2; G4, grade 4; Cat0, no Catalan exposure outside school; Cat1, some degree of Catalan exposure outside school (e.g., child speaks Catalan with mother); SES, socioeconomic status; Phono. Acc., phonological accuracy. ***p < 0.001. The Contribution of Non-phonological TABLE 5 | Unstandardized estimates, standard errors, and significance values of Strategies to Conventional Spelling SEM on conventional spelling accuracy scores (final model† ). A second series of models examined the contribution of Variable Estimate2 SE p-Value non-phonological spelling strategies to explaining conventional accuracy scores. We ran a one-level, multigroup SEM, in which Lexical1 0.320 0.021 <0.001 conventional accuracy was the dependent variable and each Morphological1 0.326 0.020 <0.001 group was defined, as in previous analyses, according to the Morphophonological1 0.157 0.020 <0.001 educational level (grade 2 or 4) and exposure to Catalan outside Orthographic1 0.180 0.020 <0.001 school (exposure, no exposure). A baseline model that included Phonological accuracy 0.213 0.029 <0.001 sex and SES, as well as phonological accuracy, was an excellent Sex −0.002 0.028 0.087 fit to the data Chi2 (8) = 7.10, p = 0.530; RMSEA = 0.000; SES 0.026 0.036 0.232 CFI = 1.000; TLI = 1.002; SRMR = 0.027, and explained a 1 Strategy type; 2 standardized estimates; SE, standard error. † Model in which all significant proportion of the variance, with R2 values ranging groups were constrained to be equal. from 0.430 to 0.583. A model in which children’s performance on each type of non-phonological spelling strategy was added as an independent variable was also a great fit to the data, a dictation task that included words containing a target 1Chi2 (26) = 25.1, p > 0.05, and it explained a much larger inconsistent phoneme. Words were grouped according to proportion of variance: Rs2 = 0.848 and 0.865, for grades 2 and the type of non-phonological strategy that was required to 4 without Catalan exposure outside school, and Rs2 = 0.841, resolve an inconsistency: morphological, morphophonological, and 0.866 for grades 2 and 4 with Catalan exposure outside orthographic, and lexical. In addition, children’s production school, respectively. Finally, a model in which the influence of of each word was assessed in terms of conventional and non-phonological spelling strategies on conventional accuracy phonographic accuracy. We aimed to ascertain the relative was constrained to be equal across all four groups was an difficulty of applying different types of non-phonological excellent fit to the data as well, Chi2 (46) = 42.05, p = 0.638; strategies, and the contribution of each strategy to conventional RMSEA = 0.000; CFI = 1.000; TLI = 1.001; SRMR = 0.036; accuracy scores, above and beyond the effect of children’s 1Chi2 (12) = 9.95, p > 0.05. This final model indicated that, above phonographic skills. and beyond the significant effect of phonographic accuracy, conventional accuracy was affected by children’s ability to apply Development of Non-phonological non-phonological strategies. In contrast, children’s sex or their parents’ SES did not exert a substantial influence. Moreover, our Spelling Strategies Our first research question involved the relative difficulty results show that morphological and lexical skills contributed with which children applied the various non-phonological more to conventional accuracy scores than either orthographic strategies, controlling for their phonographic spelling skills. Our or morphophonological skills (Table 5). findings showed that, generally speaking, strategies that required morphophonological or orthographic knowledge were mastered DISCUSSION earlier than those requiring morphological or lexical knowledge. These results partially confirmed our initial hypotheses. We This study set out to examine the role of non-phonological had expected morphophonological strategies to be on the easier strategies in spelling development. A large number of side of the continuum, given that we hypothesized that even children in grades 2 and 4, speakers of Catalan, completed our youngest age group, the second graders, would succeed Frontiers in Psychology | www.frontiersin.org 8 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies in efficiently accessing the full phonological representation The most striking finding concerns our results for of words. In addition, the phenomenon that creates the morphological strategies. We expected children to be quite inconsistency, namely, the zero-realization of the final stop adept at using morpho-graphemic regularities to resolve very phoneme, although triggered by a morphological process, common spelling inconsistencies, such as plural or gerund is extremely common in everyday speech. Indeed, children formation. Our expectation was based on (1) the results from seemed to be able to extend their knowledge of phonological other studies using POMAS, which reported morphological representations to represent silent letters and to take advantage errors to be one of the least frequent (e.g., Bahr et al., 2012; of the phonology–morphology interface, in line with previous Llauradó and Tolchinsky, 2016; Daffern and Ramful, 2020) studies (Gillis and Ravid, 2006). and (2) the fact that children are speakers of a morphologically We had also expected that the lexical strategy would be rich language, which is a key factor in determining their among the hardest to apply, given that it involves word-specific sensitivity to morphological information (e.g., Gillis and Ravid, knowledge, so children would only be able to produce the correct 2006; Dressler, 2010). Nevertheless, in the present study, spelling if they already had an orthographic representation morphological strategies were one of the hardest to be applied of the word. Our findings for the lexical category supported across educational levels and population types. This was a this hypothesis. This means that, even in semi-consistent surprising result, especially because we only tested regular orthographies like Catalan, strong orthographic representations inflectional morphology (plural formation and gerunds), which are needed to spell words that are not entirely rule-bound has been reported to be the easiest context for the application (e.g., Carrillo et al., 2013; Marinelli et al., 2017; Angelelli of morphological knowledge in spelling (e.g., Llauradó and et al., 2018). It is likely that vocabulary knowledge and Tolchinsky, 2016; Daffern and Ramful, 2020). We believe that strong semantic representations are advisable teaching strategies the differences with past research are essentially methodological. to enhance this source of knowledge to improve children’s Most previous studies did not control for the occurrences of spelling skills. morphologically bound spellings. Particularly in studies that As for orthographic strategies, we had expected that they analyzed free writing samples (e.g., Bahr et al., 2012), each text would be relatively easy to apply across groups. Results supported will have created a different number of opportunities in which this hypothesis. Previous studies using POMAS had found application of a morpho-graphemic mapping was relevant (e.g., this category to be among the least successful (e.g., Bahr regular past tenses, derivates, plurals, and so on). Therefore, the et al., 2012; Llauradó and Tolchinsky, 2016; Daffern and low rates of morphological mistakes found in those studies might Ramful, 2020). However, we expected a different result based be merely reflecting instances in which morphological spellings on the fact that, in contrast to most previous research using were applicable (and, certainly, not resolved successfully). The POMAS, we had distinguished between types of misspellings, current study arguably provides a more reliable result, given that reserving this category only for spelling mistakes that involved children encountered the exact same number of opportunities for overlooking orthographic patterns and constraints, while word- the application of each strategy. Our findings thus indicate that, specific strategies were under the “lexical” category. Given the even in a language with rich morphology like Catalan, children fact that our orthographic category required applying constraints struggle to mobilize this aspect of their linguistic knowledge to to which children have been found to be sensitive from early use it for spelling. Previous research on morphological awareness on (Cassar and Treiman, 1997), our expectation was that indicated that it is, indeed, a protracted development (e.g., our participants would readily choose between homophonous Green et al., 2003), although intervention studies to improve letters using orthographic knowledge. It should be noted, it are generally successful (e.g., Devonshire and Fluck, 2010; however, that one study, also on Catalan, made the same Devonshire et al., 2013; Bowers and Bowers, 2017). A key distinction between orthographic and lexical errors (Llauradó educational implication is, thus, that children need to be taught and Tolchinsky, 2016) and found orthographic strategies to about the way words are formed and how these forms map onto be harder than phonographic and morphological strategies. the orthography. We would speculate that the differences with this previous study are due to the fact that Llauradó and Tolchinsky (2016) analyzed words that children had produced in a relatively free Contribution of Non-phonological writing context, while we analyzed only a specific segment Strategies to Conventional Spelling (phoneme) in a closed group of words, targeting a single A second research question concerned the impact of the ambiguous phoneme, /Z/. The orthographic rules that need various non-phonological spelling strategies on spelling accuracy to be applied to choose between homophonous letters in scores, over and above that of children’s phonographic skills. our case are taught at school, and, although children do A major novel outcome of the present study was the make substitution mistakes involving and , it is finding that all non-phonological strategies had a significant reasonable to think that a lot of them were aware of the and unique contribution to spelling conventionally across rule. Future studies should test a more comprehensive set educational levels and population types. In this way, we of orthographic-bound inconsistencies to determine whether have provided substantial support to theories that pose orthographic knowledge is indeed easier to apply than other types that learning to spell requires mobilizing various types of of strategies or whether it is highly dependent on the particular linguistic knowledge besides phonological skills (e.g., Rittle- context chosen. Johnson and Siegler, 1999; Bahr et al., 2009; Treiman, 2017). Frontiers in Psychology | www.frontiersin.org 9 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies Learning to spell involves phonological skills and knowledge the orthography (Alves et al., 2018). Lexical strategies should of letter-sound correspondences (e.g., Caravolas et al., 2012), be promoted by calling children’s attention to word-specific but it also requires knowledge of orthographic constraints, of spellings in meaningful contexts, with the overarching goal of morpho-graphemic regularities, and word-specific orthographic facilitating the creation of robust orthographic representations representations of tokens whose spelling cannot be ascertained by filled with semantic information (Treiman and Kessler, 2014; generalizable patterns. Treiman, 2017). The fact that the impact of non-phonological spelling strategies on conventional spelling accuracy was unaffected by children’s educational level indicates that these various types L1 and L2 Non-phonological Spelling of knowledge sources are operative over a wide developmental Strategies span. Similarly, the lack of an educational-level effect questions Second-language learners who do not have at-home support phonology-first approaches to literacy development [e.g., of the language of instruction found it harder to apply all National Reading Panel, 2000] and calls for further research spelling strategies, as expected. However, they did not show a on the importance of mobilizing all levels of linguistic unique pattern of strategy use and, just as the L1 participants, representations relevant for spelling from the earlier grades. found it easier to apply morphophonological and orthographic These findings are in line with previous studies claiming knowledge, than to apply morphological or lexical knowledge. that non-phonological spelling strategies are necessary also in Despite not showing a unique developmental route, we did not orthographies more transparent than English or French (e.g., find support for a common model of spelling strategy use (that is, Lehtonen and Bryant, 2005; Defior et al., 2008; Carrillo and one that did not distinguish between these two population types), Alegría, 2014; Rothe et al., 2014; Angelelli et al., 2017, 2018). It suggesting that performance differences were substantial, in line could be argued that users of orthographies with very consistent with previous studies (e.g., Geva et al., 1993; Verhoeven, 2000). phoneme-to-grapheme mappings would rely to a great extent Notably, this was true for children in grade 2 as for children on these simple associations, rather than apply a host of in grade 4, indicating that having little or no contact with the different strategies to spell accurately. However, spellers appear language of instruction outside school has a long-lasting impact to take advantage of additional knowledge sources that may help on these children’s literacy development. This means that L2 to disambiguate between alternative, homophonous spellings learners require extra support for spelling development beyond regardless, in principle, of how inconsistent the system is. Future phonology (e.g., Bar-Kochva and Hasselhorn, 2017; Bowers and research should strive to compare efficacy in strategy use (both Bowers, 2017), though such training can be similar in nature to phonological and non-phonological) across languages. training aimed to their L1 peers (e.g., Devonshire and Fluck, 2010; Devonshire et al., 2013; Alves et al., 2018). Without adequate support, however, these children could be at risk of academic Use of Non-phonological Strategies and failure, in view of the key role that spelling has on writing Grade development (e.g., Juel, 1988; Salas and Silvente, 2019). As expected, fourth graders outscored second graders in all The similarities found across L1 and L2 spelling strategies were strategy types, but the general pattern of difficulty applied across even clearer in the contribution of the various non-phonological grades: morphophonological and orthographic strategies were strategies to conventional spelling, above and beyond the generally easier to apply than both morphological and lexical contribution of phonographic strategies and demographic strategies. These findings contradict stage-like views of spelling factors. Results were consistent with studies reporting that L1 and that consider non-phonological strategies a later development L2 spelling have similar drivers of performance (Geva et al., 1993; (e.g., Frith, 1985). On the contrary, the present study provides Verhoeven, 2000), given that all non-phonological strategies further support to spelling development theories that claim that contributed greatly to explaining conventional spelling accuracy non-phonological spelling strategies are used from very early on across population types. The lack of qualitative differences in (e.g., Rittle-Johnson and Siegler, 1999; Treiman, 2017). the underpinnings of spelling development as a function of The current study also has clear educational implications. language status (i.e., L1 vs. L2) extends previous assertions about On the one hand, since all non-phonological strategies are potentially universal factors driving (early) literacy development, paramount for spelling accurately, the teaching of spelling should at least in alphabetic writing systems (Caravolas et al., 2012). strive to mobilize all relevant linguistic levels from very early on (e.g., Devonshire et al., 2013). On the other hand, at least from grade 2 onward, teachers should target lexical and morphological Limitations strategies in particular, in an overall multidimensional approach As with virtually every other developmental study, it would have to spelling instruction. This is because it was precisely these been ideal to test the same hypotheses on longitudinal, rather strategies that not only proved to be the hardest but also than cross-sectional data. Future studies should try to inquire were the ones that made the largest contribution to spelling whether the same pattern of results is replicated in longitudinal conventionally. Morphological strategies could be taught by datasets, particularly encompassing a larger developmental span. raising children’s levels of morphological awareness, explicitly Another shortcoming of the present study is that it included teaching them about word formation, while showing the a limited number of use cases within each strategy type. This specific (and consistent) way in which morphemes map onto affects particularly the application of orthographic knowledge, Frontiers in Psychology | www.frontiersin.org 10 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies in view that some orthographic patterns might be more difficult DATA AVAILABILITY STATEMENT to learn than others. Similarly, research on morphological spelling strategies would benefit from investigations that include The datasets generated for this study are available on request to a larger set of contexts, with not just inflectional but also the corresponding author. derivational morpho-graphemic correspondences. Finally, the present findings may only be applicable to the language under examination, Catalan, although they can be accumulated to ETHICS STATEMENT previous studies that used a similar error-analysis approach. Cross-linguistic studies should be able to shed light on whether The studies involving human participants were reviewed and the trends presented are part of a language-general pattern of approved by Comissió d’Ètica en l’Experimentació Animal i spelling development. Humana, UAB. Written informed consent to participate in this study was provided by the participants’ legal guardian/next of kin. CONCLUSION Spelling is a multidimensional skill that involves phonographic AUTHOR CONTRIBUTIONS knowledge, but to which non-phonological strategies make a large contribution early on in development. We have NS designed and conducted the study, analyzed the spelling shown that, even in a relatively consistent orthography, samples, carried out statistical analyses, and wrote the manuscript strategies beyond phonology that involve morphophonological, in its entirety. morphological, orthographic, and lexical knowledge are instrumental to spell accurately. In particular, knowledge of morpho-graphemic correspondences and word-specific FUNDING orthographic representations made the largest contribution to conventional spelling scores in early and intermediate This research was supported by grant 2015ACUP 00175 to NS elementary school levels. L2 learners had more difficulty and grant EDU2015-64798-R to Teresa Ribas. to apply all strategy types than peers who have target language exposure outside school, but they seem to follow the same developmental route. Across population types and ACKNOWLEDGMENTS regardless of children’s sex or their parents’ SES, applying morphological and lexical strategies was more challenging than I would like to thank all the children and teachers who applying morphophonological and orthographic regularities. participated in the study. I am also very grateful to Mariona Spelling instruction should therefore strive to adopt a Pascual and Sara Silvente for their invaluable help collecting data. multidimensional approach from the earlier grades and provide Special thanks to Sílvia Paré for her attention to detail coding part extra support to children without exposure to the language of of these data and to Dr. Gabriel Liberman for his guidance in the instruction outside school. statistical approach. REFERENCES Aram, D., and Levin, I. (2001). Mother–child joint writing in low SES: sociocultural factors, maternal mediation, and emergent literacy. Cogn. Dev. 16, 831–852. Alegría, J., and Carrillo, M. S. (2014). Learning to spell words in Spanish: doi: 10.1016/S0885-2014(01)00067-3 a comparative analysis/La escritura de palabras en castellano: un análisis Bahr, R. H., Silliman, E. R., and Berninger, V. (2009). “What spelling errors have comparativo. Estud. Psicol. 35, 476–501. doi: 10.1080/02109395.2014.978544 to tell about vocabulary learning,” in Contemporary Perspectives on Reading and Allred, R. A. (1990). Gender differences in spelling achievement in grades 1 through Spelling, eds C. P. Wood and V. Connelly (London: Routledge), 121–142. 6. J. Educ. Res. 83, 187–193. doi: 10.1080/00220671.1990.10885955 Bahr, R. H., Silliman, E. R., Berninger, V. W., and Dow, M. (2012). Linguistic Alves, R. A., Limpo, T., Salas, N., and Joshi, R. M. (2018). “Handwriting and pattern analysis of misspellings of typically developing writers in grades 1– spelling,” in Best Practices in Writing Instruction, eds S. Graham, C. MacArthur, 9. J. Speech Lang. Hear. Res. 55, 1587–1599. doi: 10.1044/1092-4388(2012/10- and M. Hebert (New York, NY: Guilford Press), 211–239. 0335) Angelelli, P., Marinelli, C., and Burani, C. (2014). The effect of morphology on Bahr, R. H., Silliman, E. R., Danzak, R. L., and Wilkinson, L. C. (2015). Bilingual spelling and reading accuracy: a study on Italian children. Front. Psychol. 5:1373. spelling patterns in middle school: it is more than transfer. Int. J. Biling. Educ. doi: 10.3389/fpsyg.2014.01373 Biling. 18, 73–91. doi: 10.1080/13670050.2013.878304 Angelelli, P., Marinelli, C. V., De Salvatore, M., and Burani, C. (2017). Morpheme- Bar-Kochva, I., and Hasselhorn, M. (2017). The training of morphological based reading and spelling in Italian children with developmental dyslexia and decomposition in word processing and its effects on literacy skills. Front. dysorthography. Dyslexia 23, 387–405. doi: 10.1002/dys.1554 Psychol. 8:1583. doi: 10.3389/fpsyg.2017.01583 Angelelli, P., Marinelli, C. V., Putzolu, A., Notarnicola, A., Iaia, M., and Baseki, J., Andreou, G., and Tzivinikou, S. (2016). Dyslexia and spelling in two Burani, C. (2018). Learning to spell in a language with transparent different orthographies (Greek vs. English): a linguistic analysis. J. Adv. Linguist. orthography: distributional properties of orthography and whole-word lexical 7, 1172–1191. doi: 10.24297/jal.v7i1.4622 processing. Q. J. Exp. Psychol. 71, 704–716. doi: 10.1080/17470218.2016.127 Bel, A. (2003). The syntax of subjects in the acquisition of Spanish and Catalan. 5715 Probus 15, 1–26. doi: 10.1515/prbs.2003.003 Frontiers in Psychology | www.frontiersin.org 11 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies Benson-Goldberg, S. (2014). Spelling of Derivationally Complex Words: The Role of Gillis, S., and Ravid, D. (2006). Typological effects on spelling development: a Phonological, Orthographic, and Morphological Features. Doctoral dissertation, crosslinguistic study of hebrew and dutch. J. Child Lang. 33, 621–659. doi: University of South Florida, Tampa, FL. 10.1017/S0305000906007434 Berninger, V., and Winn, W. (2006). “Implications of advancements in brain Graham, S., and Harris, K. R. (2005). “The impact of handwriting and spelling research and technology for writing development, writing instruction, and instruction on the writing and reading performance of at-risk first grade educational evolution,” in Handbook of Writing Research, eds C. MacArthur, writers,” in Proceeding of the Paper presented at the Pacific Coast Research S. Graham, and J. Fitgerald ( New York, NY: The Guilford Press), 96–114. Conference, Coronado, CA. Borgwaldt, S. R., Hellwig, F. M., and de Groot, A. M. (2004). Word-initial entropy Green, L., McCutchen, D., Schwiebert, C., Quinlan, T., Eva-Wood, A., and Juelis, in five languages: letter to sound, and sound to letter. Writ. Lang. Literacy 7, J. (2003). Morphological development in children’s writing. J. Educ. Psychol. 95, 165–184. doi: 10.1075/wll.7.2.03bor 752–761. doi: 10.1037/0022-0663.95.4.752 Bowers, J. S., and Bowers, P. N. (2017). Beyond phonics: the case for teaching Henderson, E. H., and Beers, J. (1980). Developmental and Cognitive Aspects of children the logic of the English spelling system. Educ. Psychol. 52, 124–141. Learning to Spell: A Reflection of Word Knowledge. Newark, Del: International doi: 10.1080/00461520.2017.1288571 Reading Association. Byrne, B. (1998). The Child’s Acquisition of the Alphabetic Principle. London: Jongejan, W., Verhoeven, L., and Siegel, L. S. (2007). Predictors of reading and Psychology Press. spelling abilities in first-and second-language learners. J. Educ. Psychol. 99, Caravolas, M., Hulme, C., and Snowling, M. J. (2001). The foundations of spelling 835–851. doi: 10.1037/0022-0663.99.4.835 ability: evidence from a 3-year longitudinal study. J. Mem. Lang. 45, 751–774. Joye, N. (2019). Spelling With a Developmental Language Disorder: Predictors, doi: 10.1006/jmla.2000.2785 Strategies and Error Patterns in French-and English-Speaking Students at the End Caravolas, M., Lervåg, A., Mousikou, P., Efrim, C., Litavský, M., Onochie- of Primary School. Doctoral dissertation, University College London, London. Quintanilla, E., et al. (2012). Common patterns of prediction of literacy Juel, C. (1988). Learning to read and write: a longitudinal study of 54 children from development in different alphabetic orthographies. Psychol. Sci. 23, 678–686. first through fourth grades. J. Educ. Psychol. 80, 437–447. doi: 10.1037/0022- doi: 10.1177/0956797611434536 0663.80.4.437 Carrillo, M. S., and Alegría, J. (2014). The development of children’s sensitivity Kline, R. B. (2011). Principles and Practice of Structural Equation Modeling, 3rd to bigram frequencies when spelling in Spanish, a transparent writing system. Edn. New York, NY: Guilford Press. Read. Writ. 27, 571–590. doi: 10.1007/s11145-013-9459-y Lee, J. A. C., and Al Otaiba, S. (2017). End-of-kindergarten spelling outcomes: how Carrillo, M. S., Alegría, J., and Marín, J. (2013). On the acquisition of some basic can spelling error analysis data inform beginning reading instruction? Read. word spelling mechanisms in a deep (French) and a shallow (Spanish) system. Writ. Q. 33, 226–238. doi: 10.1080/10573569.2016.1165639 Read. Writ. 26, 799–819. doi: 10.1007/s11145-012-9391-6 Lehtonen, A., and Bryant, P. (2005). Active players or just passive bystanders? The Cassar, M., and Treiman, R. (1997). The beginnings of orthographic knowledge: role of morphemes in spelling development in a transparent orthography. Appl. children’s knowledge of double letters in words. J. Educ. Psychol. 89, 631–644. Psycholinguist. 26, 137–155. doi: 10.1017/S0142716405050113 doi: 10.1037/0022-0663.89.4.631 Llaurado, A., Martí, M. A., and Tolchinsky, L. (2012). Corpus CesCa: compiling a Cummins, J. (1979). Cognitive/academic language proficiency, linguistic corpus of written Catalan produced by school children. Int. J. Corpus Linguist. interdependence, the optimum age question and some other matters. Work. 17, 428–441. doi: 10.1075/ijcl.17.3.06lla Pap. Bilingualism 19, 198–205. Llauradó, A., and Tolchinsky, L. (2016). The developmental pattern of spelling in Daffern, T., and Ramful, A. (2020). Measurement of spelling ability: construction Catalan from first to fifth school grade. Writ. Syst. Res. 8, 64–83. doi: 10.1080/ and validation of a phonological, orthographic and morphological pseudo- 17586801.2014.1000812 word instrument for students in Grades 3–6. Read. Writ. 33, 571–603. doi: Marinelli, C. V., Cellini, P., Zoccolotti, P., and Angelelli, P. (2017). Lexical 10.1007/S11145-019-09976-1 processing and distributional knowledge in sound–spelling mapping in a Deacon, S. H., Conrad, N., and Pacton, S. (2008). A statistical learning perspective consistent orthography: a longitudinal study of reading and spelling in dyslexic on children’s learning about graphotactic and morphological regularities in and typically developing children. Cogn. Neuropsychol. 34, 163–186. doi: 10. spelling. Can. Psychol. Psychol. Can. 49, 118–124. doi: 10.1037/0708-5591.49. 1080/02643294.2017.1386168 2.118 Martin, K. I. (2017). The impact of L1 writing system on ESL knowledge of vowel Defior, S., Alegría, J., Titos, R., and Martos, F. (2008). Using morphology when and consonant spellings. Read. Writ. 30, 279–298. doi: 10.1007/s11145-016- spelling in a shallow orthographic system: the case of Spanish. Cogn. Dev. 23, 9673-5 204–215. doi: 10.1016/j.cogdev.2007.01.003 Masterson, J. J., and Apel, K. (2010). The spelling sensitivity score: noting Devonshire, V., and Fluck, M. (2010). Spelling development: fine-tuning strategy- developmental changes in spelling knowledge. Assess. Effect. Intervent. 36, use and capitalising on the connections between words. Learn. Instruc. 20, 35–45. 361–371. doi: 10.1016/j.learninstruc.2009.02.025 National Reading Panel (2000). Teaching Children to Read: An Evidence-Based Devonshire, V., Morris, P., and Fluck, M. (2013). Spelling and reading Assessment of the Scientific Research Literature on Reading and Its Implications development: the effect of teaching children multiple levels of representation in for Reading Instruction (National Institute of Health Pub. No. 00-4769). their orthography. Learn. Instruc. 25, 85–94. doi: 10.1016/j.learninstruc.2012. Washington, DC: National Institute of Child Health and Human Development. 11.007 Nunes, T., Bryant, P., and Bindman, M. (1997a). Learning to spell regular Dressler, W. (2010). “A typological approach to first language acquisition,” in and irregular verbs. Read. Writ. 9, 427–449. doi: 10.1023/A:100795121 Language Acquisition across Linguistic and Cognitive Systems, eds M. Kail and 3624 M. Hickman (Amsterdam: John Benjamins). Nunes, T., Bryant, P., and Bindman, M. (1997b). Morphological spelling strategies: Ferreiro, E., and Teberosky, A. (1979). Los Sistemas de Escritura en el Desarrollo del developmental stages and processes. Dev. Psychol. 33, 637–649. doi: 10.1037/ niño. Melbourne VIC: Siglo, 21. 0012-1649.33.4.637 Frisby, C. (2016). An empirical comparison of the Words Spelled Correctly and Pacton, S., and Deacon, S. H. (2008). The timing and mechanisms of children’s Correct Letter Sequence spelling scoring methods in third-and fourth-grade use of morphological information in spelling: a review of evidence from classrooms. J. Appl. Sch. Psychol. 32, 101–121. doi: 10.1080/15377903.2016. English and French. Cogn. Dev. 23, 339–359. doi: 10.1016/j.cogdev.2007. 1151847 09.004 Frith, U. (1985). Beneath the surface of developmental dyslexia. Surface Dyslexia Pacton, S., Fayol, M., and Perruchet, P. (2002). “The acquisition of untaught 32, 301–330. orthographic regularities in French,” in Precursors of Functional Literacy. Studies Gentry, J. R. (1978). Early spelling strategies. Elemen. Sch. J. 79, 88–92. doi: 10.1086/ in Written Language and Literacy 11, eds L. Verhoeven, C. Elbro, and P. Rietsma 461136 (Amsterdam: John Benjamins), 121–138. Geva, E., Wade-Woolley, L., and Shany, M. (1993). The concurrent development Prieto, P. (2004). Els Sons del Català. Malappuram: UOC. of spelling and decoding in two different orthographies. J. Read. Behav. 25, Read, C. (1971). Pre-school children’s knowledge of English phonology. Harvard 383–406. Educ. Rev. 41, 1–34. doi: 10.17763/haer.41.1.91367v0h80051573 Frontiers in Psychology | www.frontiersin.org 12 June 2020 | Volume 11 | Article 1071 Salas Non-phonological Spelling Strategies Ritchey, K. D., Coker, D. L., and McCraw, S. B. (2010). A comparison of metrics for Treiman, R., and Cassar, M. (1996). Effects of morphology on children’s spelling of scoring beginning spelling. Assess. Effect. Intervent. 35, 78–88. final consonant clusters. J. Exp. Child Psychol. 63, 141–170. doi: 10.1006/jecp. Rittle-Johnson, B., and Siegler, R. S. (1999). Learning to spell: variability, choice, 1996.0045 and change in children’s strategy use. Child Dev. 70, 332–348. doi: 10.1111/ Treiman, R., Kessler, B., and Caravolas, M. (2019). What methods of scoring young 1467-8624.00025 children’s spelling best predict later spelling performance?. J. Res. Read. 42, Rothe, J., Schulte-Körne, G., and Ise, E. (2014). Does sensitivity to orthographic 80–96. doi: 10.1111/1467-9817.12241 regularities influence reading and spelling acquisition? A 1-year prospective Treiman, R., and Kessler, B. (2014). How Children Learn to Write Words. Oxford: study. Read. Writ. 27, 1141–1161. doi: 10.1007/s11145-013-9479-7 Oxford University Press. Salas, N., and Silvente, S. (2019). The role of executive functions and transcription Treiman, R., Kessler, B., Pollo, T. C., Byrne, B., and Olson, R. K. (2016). Measures skills in writing: a cross-sectional study across 7 years of schooling. Read. Writ. of kindergarten spelling and their relations to later spelling performance. Sci. 33, 877–905. doi: 10.1007/s11145-019-09979-y Stud. Read. 20, 349–362. doi: 10.1080/10888438.2016.1186168 Sénéchal, M. (2000). Morphological effects in children’s spelling of French words. Verhoeven, L. (2000). Components in early second language reading and spelling. Can. J. Exp. Psychol. Rev. Can. Psychol. Exp. 54, 76–86. Sci. Stud. Read. 4, 313–330. doi: 10.1207/S1532799XSSR0404_4 Silliman, E. R., Bahr, R. H., and Peters, M. L. (2006). Spelling patterns in Weber, P., Kozel, N., Purgstaller, C., Kargl, R., Schwab, D., and Fink, A. (2013). First preadolescents with atypical language skills: phonological, morphological, and second language in the brain: neuronal correlates of language processing and orthographic factors. Dev. Neuropsychol. 29, 93–123. doi: 10.1207/ and spelling strategies. Brain Lang. 124, 22–33. doi: 10.1016/j.bandl.2012.11. s15326942dn2901_6 010 Sparks, R. L., Patton, J., Ganschow, L., Humbach, N., and Javorsky, J. (2008). Early Zarić, J., Hasselhorn, M., and Nagler, T. (2020). Orthographic knowledge predicts first-language reading and spelling skills predict later second-language reading reading and spelling skills over and above general intelligence and phonological and spelling skills. J. Educ. Psychol. 100, 162–174. doi: 10.1037/0022-0663.100. awareness. Eur. J. Psychol. Educ. 1–23. doi: 10.1007/s10212-020-00464-7 1.162 Suárez-Coalla, P., Martínez-García, C., and Cuetos, F. (2017). Morpheme-based Conflict of Interest: The author declares that the research was conducted in the reading and writing in Spanish children with dyslexia. Front. in Psychol. 8:1952. absence of any commercial or financial relationships that could be construed as a doi: 10.3389/fpsyg.2017.01952 potential conflict of interest. Treiman, R. (1994). “Sources of information used by beginning spellers,” in Handbook of Spelling: Theory, Process and Intervention, eds G. D. A. Brown and Copyright © 2020 Salas. This is an open-access article distributed under the terms N. C. Ellis (Hoboken, NJ: Wiley), 76–91. of the Creative Commons Attribution License (CC BY). The use, distribution or Treiman, R. (2004). “Phonology and spelling,” in Handbook of Children’s Literacy, reproduction in other forums is permitted, provided the original author(s) and the eds T. Nunes and P. Bryant (Alphen aan den Rijn: Kluwer), 31–42. copyright owner(s) are credited and that the original publication in this journal Treiman, R. (2017). Learning to spell words: findings, theories, and issues. Sci. Stud. is cited, in accordance with accepted academic practice. No use, distribution or Read. 21, 265–276. doi: 10.1080/10888438.2017.1296449 reproduction is permitted which does not comply with these terms. Frontiers in Psychology | www.frontiersin.org 13 June 2020 | Volume 11 | Article 1071