Disentangling Language Disorder and Bilingualism in Children with Developmental Language Disorder and Autism Spectrum Disorder: Evidence from Writing

Research in second language (L2) writing demonstrates that bilingual children draw on their native language to fulfil writing requirements, with the strength of cross-language transfer being modulated by the length of the children’s L2 exposure and proficiency (see Williams & Lowrance-Faulhaber, 2018 for a review). Cross-linguistic influence has been found to be particularly strong for spelling as well as orthography-specific skills during handwritten production in the L2, especially in children speaking alphabetic languages in which phonological as well as orthographic overlap boosts transfer of spelling strategies across the two languages (Sun-Alperin & Wang, 2011; Yeong et al., 2014). The majority of the studies found that the bilingual groups exhibit weaker spelling skills in the L2 than their TD monolingual peers (e.g. Caravolas et al., 2020; Guimaraes & Parkins, 2019; Li et al., 2012a, 2012b), even when monolingual and bilingual children have comparable amounts of writing experience in their main language of instruction (Caravolas et al., 2020). Spelling errors in both Caravolas et al.’s (2020) and Li et al’s. (2012a, 2012b) studies did not differ across the children’s L1 and L2, while word spelling ability in L1 emerged as a significant predictor of spelling skills in the L2, thus, suggesting transfer of orthographic properties across the two languages.

While most work investigating the writing development of TD bilingual children has focused on their code-related (i.e. letter knowledge, spelling) skills, very little attention has been paid to the direct empirical investigation of prosodic patterns underlying the use of punctuation in bilingual children’s written texts. There are studies with monolingual adults that have mainly focused on the potent disambiguating role that punctuation may play in the processing of locally ambiguous sentences using self-paced reading paradigms (Drury et al., 2016; Niikuni & Muramoto, 2014; Steinhauer & Friederici, 2001). These studies converge to show that punctuation marks (especially commas) play a critical role in cueing types of parsing decision at particular points in sentences with local processing difficulty; in a sense, they guide readers through surface text organization by facilitating written text chunking into coherent blocks and pulling the interpretation of separate punctuation-marked blocks towards a globally coherent parse. Punctuation marks in written production may thus be regarded as visual markers of coherent syntactic and conceptual dependencies that would otherwise be unrelated. In this sense, they may provide the same facilitatory benefits in chunking written discourse as explicit prosody (i.e. the intonation patterns of pitch accents and phrase boundaries and rhythmic timing) provides during listening comprehension, at least in neurotypical individuals (Steinhauer, 2003). Indeed, in Heggie and Wade-Woolley’s (2018) study, adults with high prosodic awareness skills in oral language were more efficient to correctly apply punctuation in written speech, over and above the influence of their baseline punctuation knowledge, as compared to their peers with lower prosodic awareness skills. Similarly, Calet et al. (2017) found that TD children that have been trained on manipulating expressive prosody to encode information structure in a reading-aloud task tended to perform more accurately than untrained children in a written prosody task, in which they had to insert punctuation marks in unpunctuated sentences in order to assign to them a particular meaning. The overall findings support the hypothesis that the ability to use prosodic ‘boundaries’ to capture the information structure of sentences in oral speech may have important consequences for the appropriate use of punctuation marks in writing, which may, in turn, reflect an implicit or internal prosodic representation, also known as implicit prosody (Fodor, 2002). The way TD bilingual children, as well as bilingual children with ASD and DLD use implicit prosodic cues while conveying meaning in text production is considerably underexplored.

Though the effects of DLD and ASD on the oral language abilities of school-aged monolingual and bilingual children have been extensively reported across various methodological designs and languages (e.g. Broc et al., 2021; Durrleman & Delage, 2016; Goldman, 2008; Huang & Finestack, 2020; Peristeri et al., 2017; Schaeffer et al., 2014, among many others), writing skill in the two disorders, especially autism, has received relatively little attention.

The effect of the disorder on the writing skills of monolingual children with DLD is declared to be robust. Especially, spelling has been shown to be a particularly vulnerable domain for children with DLD, who seem to fall behind their age-matched peers on applying orthographic and morphological spelling conventions (Brizzolara et al., 2011; Broc et al., 2013; Cordewener et al., 2012; Critten et al., 2014; Larkin et al., 2013). According to Graham et al.’s (2020) recent meta-analysis, when language was controlled across studies, DLD children’ s spelling difficulties were significantly greater than their language-matched peers. Crucially, a number of studies (e.g. Dockrell & Connelly, 2015; Dockrell et al., 2007, 2009; Mackie et al., 2013; Puranik & Lonigan, 2012) have demonstrated a strong influence of DLD children’s oral language ability (i.e. expressive vocabulary, phonological awareness, recalling sentences, oral speech perception) on their spelling skills. These findings indicate that oral language difficulties in monolingual children with DLD can drive their spelling performance in written text production. These results also align well with established findings in relevant research in TD monolingual children that spelling is backed up by the child’s phonological awareness skills which become more strengthened as breadth of expressive vocabulary increases in order to avoid confusion of lexical items having similar phonological representations (Cassano & Schickedanz, 2015; Metsala, 1999). At the same time, these findings add to our growing understanding of how code-related aspects of writing, and especially spelling, in DLD may be affected by the children’s oral language difficulties.

Studies on the writing skills of monolingual individuals with ASD have been considerably fewer relative to children with DLD, and reveal asymmetries across code- and meaning-related aspects of writing. Specifically, at the level of written word representation, monolingual autistic children and adults have been found to exhibit poor handwriting legibility and letter formation (Kushki et al., 2011), while their performance at the macro-organizational level of writing is characterized by poor global and local coherence relations between the ideas (Brown & Klein, 2011; Dockrell et al., 2014). On the other hand, there is evidence showing that syntactic complexity and lexical diversity in written production are well-preserved in autism, and that autistic children’s rates of spelling mistakes are comparable to those in TD children (Hilvert, 2018). While there is no robust support to link ASD children’s spelling performance to their oral language skills, there has been some evidence for a correlation between word spelling accuracy and phonological awareness skills, but not oral vocabulary knowledge (Bailey & Arciuli, 2018). Phonological awareness has been found to be an important emergent literacy predictor for kindergarten children with ASD (Dynia et al., 2017), and it has been associated with their weak sound-to-letter mappings. On the other hand, vocabulary knowledge appears to be a preserved component of autistic children’s oral language competency, and even seems to be boosted by the autistic condition (e.g. Davidson, 2021; Kissine et al., 2019; Durrleman et al., 2022).

Furthermore, monolingual children with ASD have been reported to differ from TD peers on explicit prosodic properties of their speech. In fact, deficits in speech prosody in individuals with ASD are one of the earliest symptoms of the disorder (Tager-Flusberg et al., 2005) and are reported to persist even if other areas of language improve (McCann & Peppé, 2003). Prosodic impairments in ASD have been mainly observed in language production, and more specifically, in the use of prosody to convey phrase-level stress (McCann et al., 2007), in slow syllabic speech (Baron-Cohen & Staunton, 1994) and in increased pitch range in conversation (Nadig & Shaw, 2012). The functional role of prosody in perception has been found to be relatively more preserved than production, since monolingual children with ASD have been shown to discriminate word pairs with distinct lexical stress patterns (Grossman et al., 2010), to recall better stressed than unstressed words (Fine et al., 1991), and to distinguish between sentence types, such as statements and questions (Paul et al., 2005; Peppé et al., 2007, 2011). Interestingly, Geelhand et al. (2021) have recently investigated syntax-prosody mappings in the speech production of monolingual autistic and neurotypical adults, and found that the two groups differed in the strategies they employed to combine syntactic and prosodic information; crucially, the autistic adults were reported to condense significantly more syntactic units into a single prosodic unit (delineated by silent pause) relative to neurotypical adults who were found to use considerably more discourse markers (i.e. lexemes that serve a structuring function, e.g., well, you know, I mean) than the autistic group. Geelhand et al. (2021) took this evidence to suggest that explicit prosody in the delivery of oral speech can distinguish autistic adults’ discourse coherence management strategies from their neurotypical peers. Differences in speech prosody perception have also been identified as a characteristic of children with DLD. For example, Richards and Goswami (2015) have found that monolingual children with DLD fall behind their TD peers on stress perception at both the lexical and phrasal level, while Cumming et al., (2015) show that children with DLD face general perceptual difficulties with the global prosodic structure of spoken language. Though prosodic deficits in spoken language have been the focus of research in populations with DLD and ASD, the way these deficits map into the use of punctuation marks in written speech production in the same populations is as of yet not well understood.

The existing literature on the language development of bilingual children with DLD and ASD suggests that bilingualism does not seem to have a negative impact on the language abilities of these children, who can successfully acquire two languages. Studies indicate that, when provided with adequate language exposure, bilingual children with ASD develop language similar to their monolingual non-verbal IQ matched peers (Beauchamp et al., 2020; Gonzalez- Barrerro & Nadig, 2018, 2019; Lund et al., 2017). In addition, Dai et al., (2018) found that, though bilingual children with ASD showed greater language deficits than their bilingual peers with DLD, there was no adverse effect of bilingual exposure on any of the groups’ language performance. Bilinguals and monolinguals have been found to perform comparably on various language tasks (e.g., Drysdale et al., 2015; Meir & Novogrodsky, 2019, for children with ASD; Paradis et al., 2003, 2022; Schwob & Skoruppa, 2022; Tsimpli et al., 2016 for children with DLD), while it is often the case that lower performance in bilinguals for vocabulary can be attributed to the fact that this measure has been only administered in a single language and thus does not represent a bilingual’s total lexicon (Bedore & Peña, 2008). Previous work on ASD, specifically, has revealed that bilinguals who master fewer words than their monolingual peers when the lexicon is measured in a single language in fact have a larger total number of words when this is measured across both languages (Petersen et al., 2012). Beauchamp and MacLeod (2017) have stressed the urgent need to sensitize professionals to existing literature, and to formulate guidelines to assist the decision-making process for selecting language(s) for exposure in children with ASD growing up in bi/multilingual environments. To the best of our knowledge, no study has investigated the written text production of bilingual children with ASD and DLD, or the contribution of vocabulary to bilingual DLD and ASD children’s writing performance.

Since writing performance in the current study was assessed in the bilingual children’s L2/Greek, the current section presents an overview of the orthographic and stress assignment properties of the Greek language, as well as information on the developmental trajectory of spelling, stress and punctuation use in TD monolingual Greek-speaking children.

Greek is considered to be an orthographically transparent language. Specifically, grapheme-to-phoneme consistency is around 80% in the sound-to-spelling direction (Protopapas & Vlahou, 2009), with graphemes representing either single letters or letter combinations that behave like single letters, for example the ει/i/ in Greek. Phoneme-to-grapheme mappings in Greek are more transparent than English, yet, less transparent than Hungarian, Dutch and German. The moderate transparency of the Greek orthographic system mainly stems from irregularities in vowel-to-grapheme mappings. For instance, the vowel phoneme /ε/ is represented by the graphemes αι and ε, while the vowel phoneme /ι/ is graphically represented by η, ι, υ, οι, ει, and υι. Vowel spelling irregularities are the main factor influencing Greek-speaking children’s spelling performance, since the overwhelming majority of spelling errors consist of selections of mutually exclusive graphemes that give raise to phonologically acceptable representations (e.g. target: γράφει/γrafi/“(s/he) writes—spelling error γράφι/γrafi/“(s/he) writes; the spelling error is underlined” (Diamanti et al., 2014; Prootopapas et al., 2013).

Since orthographic representations in Greek arise as a by-product of relatively consistent sound-spelling mappings, Greek-speaking children develop phonological strategies, i.e. they tend to rely on sound-letter correspondences, at the phoneme and syllable level at relatively early stages of literacy development, specifically at the age of 5 years (Porpodas, 2001; Protopapas et al., 2013). For instance, the word γυναίκα/γineka/“(woman) is frequently misspelled as γινέκα, implying that children fail to consistently integrate orthographic information at this developmental stage. The repertoire of strategies to support spelling abilities is then gradually expanded to include word-specific orthographic information as well as morphological knowledge mainly related to the inflectional endings of various grammatical categories, e.g. active voice verbs are composed of the stem and the inflectional ending —ω/o/, which signifies first person singular, while feminine nouns are composed of the stem and the inflectional ending –ες/es/, which signifies plural nominative case. Inflectional types play a critical role for the correct spelling of the inflectional endings (also known as suffixes) in verbs, nouns, adjectives and pronouns (Grigorakis & Manolitsis, 2016; Protopapas, 2017; Tijms et al., 2020), so knowledge of morphological rules allows the child to capture the spelling of word forms in a fine-grained manner (Diakogiorgi et al., 2006; Nunes et al., 2006). In fact, Diamanti et al.’s (2017) study has shown that Greek children as young as 5 years can be sensitive to the grammatical role of word parts, and that they are able to use this morphological information to cope with the spelling demands of Greek suffixes introduced at later grades. Taken together, research on the time-course of the spelling development of Greek-speaking children suggests a feed-forward relationship between phonology and grammar in spelling: the phoneme-to-grapheme correspondences are retrieved, and then the phonological strategies combine with suffixes subject to a set of well-formedness principles, overridden only by lexically-specific information. Convergence to the final stage of appropriate spelling takes place at the age of 11 years (Georgiou et al., 2008). This is in line with the objectives of the official programs of studies in primary education, as well as school books, which have been recently updated by the Greek Institute of Educational Policy in Greece. By the age of 10 years, students attending Greek public schools are expected to have fully acquired spelling strategies and expand on the textual features of various texts and genres (Greek Government Gazette, 2021). Though the expository written discourse development in Greek-speaking children remains underexplored, there is limited evidence showing that it is a dynamic process that develops gradually and is academically attained in adolescence. Specifically, Kantzou’s (2019) study has focused on the syntactic complexity of the expository texts of Greek-speaking children aged 10, 13, and 16 years, and found that children at the age of 10 years are able to produce coherent expository texts, however, they follow a linear organization of information as compared to the older groups that organize their text hierarchically.

Several studies have investigated the spelling error patterns in Greek-speaking children. Across these studies, spelling errors have been classified into three categories, namely, phonological, grammatical and orthographic errors. Phonological errors are defined as those that bring about a change in the phonemic makeup of a word and are mainly derived from grapheme omissions, additions and transpositions (e.g. στ()ατός, instead of στρατός/stratos/‘army’). Grammatical errors are derived from erroneous spelling of inflectional endings that do not, however, distort the phonological identity of the words (e.g. e.g. παίρνι, instead of παίρνει/perni/‘takes’; the spelling error is underlined). Finally, orthographic errors include misspellings of word stems that do not modify the pronunciation of the target word (e.g. ορέος, instead of ωραίος/oreos/‘beautiful’). Protopapas et al. (2013) have examined spelling errors in large samples of 8-, 9-, and 12-year-old children who were tested on word and passage spelling. As expected, 12-year-old children committed considerably fewer spelling errors than the younger groups, indicating that the older group’s system was mature enough to integrate multiple cues to guide spelling processes; however, these mechanisms still undergo development since even 12-year-old children make at least some spelling errors (see also Diamanti et al., 2014 for similar findings). In Protopapas et al.’s (2013) study, there was a significant trend towards fewer phonological errors as compared to grammatical and orthographic errors across all experimental groups, while older children exhibited higher rates of grammatical vs. orthographic errors. Interestingly, phonological and punctuation errors were negligible across all age groups. Diamanti et al.’s (2014) study has also shown significant grammatical category effects on the spelling error patterns of 10–13 year-old Greek-speaking children, since suffixes in nouns were found to be more accurately spelled than stems, yet, spelling errors in the verbs’ suffixes were significantly more than in nouns. This suggests that children’s spelling performance may be explained by processing constraints posited by the syntactic properties of the grammatical categories the words belong to.

Stress assignment constitutes an additional important aspect of writing competence in Greek. Greek is a relatively free-stress language in which the location of stress on a word is restricted to the last three syllables. Researchers (Botinis, 2011; Revithiadou, 1999) have claimed that stress assignment in Greek is planned on the basis of both bottom-up and top-down cues, starting from the segmental phonological properties of the word and spanning the word’s ‘neighborhood’, since stress assignment is also sensitive to the syntactic and prosodic properties of the word’s adjacent items. A characteristic example of the synergy between lexical and syntactic cues in stress assignment is’host-and-clitic’ phrases in Greek, such as [το μάθημά μου]/to_{ARTICLE.NEUTER} máθimá_NOUN.NEUTER mu_CLITIC/‘my lesson’, where the noun appears with two stresses, one on the first syllable and another one on the third syllable of the noun. According to Botinis (2011), the enclitic stress pattern has nothing to do with the enclitic [mu] as such but rather with the phrase boundary. As a core feature of the phonological specification of words, stress assignment is essential in word production. Because of the existence of stress minimal pairs in Greek, such as μόνος/mónos/‘alone’—μονός/monós/‘single’, in which stress alone disambiguates the meaning of the two words in printed text, children can flexibly learn lexical meaning distinctions based on stress assignment, which further contributes to efficient reading performance. In spelling, stress is indicated by a diacritic over the vowel of the stressed syllable, and it is mandatory on every word with two or more syllables. Hence, any omission or misplacement of the stress diacritic is a spelling error.

Despite the importance of stress assignment in both receptive (reading) and expressive written language (spelling), little attention has been paid to the linguistic and cognitive processes involved in lexical stress assignment, the vulnerability of these processes in the case of disorders affecting written language, as well as to the potential usefulness of assessing stress assignment performance for the evaluation of writing competence. In a study of word and pseudoword reading, Protopapas (2006) found that Greek-speaking children made frequent stress assignment errors in pseudowords (i.e. strings of letters that resemble a real word) but not in words, and that such errors were associated with the children’s reading ability, further suggesting that decoding the stress diacritic may be a demanding process and that stress assignment in reading may be largely lexical. Kotoulas and Padeliadu (1999) have examined stress assignment in spelling in Greek and found it to be a highly problematic domain of performance for TD children, while Protopapas et al. (2013) spelling study in 8-to-12 year old Greek-speaking children with and without dyslexia found that the rates of stress assignment errors were comparable to those of orthographic and grammatical spelling errors. Recently, Ralli et al. (2021) found that 8-year-old Greek-speaking children with DLD produced considerably higher rates of errors on the use of the stress diacritic than their TD peers in a short-story writing task. The overall evidence in Greek suggests that stress assignment is a domain of high vulnerability in the writing performance of children with and without disorders. The findings also imply that children’s phonological spelling strategies, which are recruited rather early in spelling development (Porpodas, 2001; Protopapas et al., 2013), may not encompass suprasegmental features, such as stress, at least not beyond the single-syllable level.

Motivated by previous findings in monolingual individuals with DLD and ASD, we investigated whether written expression may present even more of a challenge to children with DLD and ASD growing up in bilingual settings. In the current study, we examine for the first time the written text production of age-, PIQ-, and SES-matched bilingual Albanian-Greek children with DLD and ASD (along with TD bilingual peers), focusing on both micro-level properties of expository texts that relate to the spelling and stress assignment errors at the word level, and macro-level properties that expand across discourse units of text, such as punctuation errors. We also examine whether the children’s language ability (vocabulary), current language use and home language history measures affect their performance in each writing component.

The children were assessed on two expository essays, which were written by the children in Greek, and were coded for (a) spelling errors (phonological, orthographic, grammatical) on nouns, verbs and adjectives, (b) stress diacritic placement, and (c) punctuation errors. We have focused on 10-to-12 year-old Albanian-Greek children who were exposed to both languages since birth, they were thus simultaneous bilinguals. All bilingual children attended mainstream classes of state schools in central Greece wherein Greek is the only medium of literacy instruction with no heritage language support. This is partially explained by the Albanian parents’ misconception of the children’s home language as an obstacle to the acquisition of the dominant language, i.e. Greek (Gkaintartzi et al., 2015).

The main objective of the current study was to investigate whether spelling errors, stress assignment and punctuation errors differ across DLD, ASD and TD bilingual children. Based on previous research demonstrating several strengths in writing in autism, including the children’s spelling skills (Hilvert, 2018), and robust impairments across all the components of the writing performance in children with DLD (e.g. Graham et al., 2020; Joye et al., 2020), we predicted that the bilingual autistic children’s error rates in spelling would be similar to TD bilingual children and lower than their DLD bilingual peers. Regarding punctuation, based on previous research with individuals with ASD and DLD (Dockrell & Connelly, 2015; Favart et al., 2016; Geelhand et al., 2021), we assumed that both bilingual groups with DLD and ASD would exhibit more punctuation errors than the rest of the groups due to their difficulty in processing mappings between coherent units of meaning and ongoing written discourse.

Furthermore, the study has explored whether the children’s language ability measured through expressive vocabulary, and bilingual experience reflected in the children’s current language use and home language history, would account for the variance in each experimental group’s error rates in spelling, stress placement and punctuation. Deficits on expressive (oral) vocabulary has been proposed to impact a range of writing components, including spelling (e.g. Kim et al., 2013; Santos & Befi-Lopes, 2012) and stress assignment skills (Bellocchi et al., 2016), thus, we expected that the children’s performance in both domains would be affected by their expressive vocabulary performance. Finally, we predicted that the bilingual children’s errors rates in writing would be inversely related to their exposure to Greek, which was the language in which the children were tested.

The current study investigated the writing abilities of bilingual children with DLD and ASD, along with their TD bilingual peers, with the aim of probing specific dimensions that would differentiate the DLD and ASD groups from TD children, as well as the two groups with DLD and ASD themselves. The study also set out to examine the effect of the current language use and home language history measures, as well as the role of language ability in the writing performance of the three groups and, thus, shed light on the mechanisms that could serve as potential factors regulating the children’s text production. The results of the study revealed discrepancies between the experimental groups across the various writing components that have been tested. More specifically, both bilingual groups with DLD and ASD exhibited more spelling and punctuation errors relative to their TD peers, yet, the writing profile of ASD children was found to be less severely affected than DLD children at least at the level of spelling content words. On the other hand, stress diacritic assignment and punctuation errors appeared to be the hallmark characteristic of ASD children’s textual output, since the latter group tended to misplace stress diacritics and punctuation markers, as well as inappropriately use superfluous stress diacritics on lexical items to a significantly greater extent as compared to both their TD and DLD peers. The results also suggest associations between the DLD and ASD groups’ punctuation errors and language exposure measures, while spelling and stress errors were rather associated with vocabulary knowledge for both groups with DLD and ASD. The overall evidence suggests the promising role of writing in the study of the cognitive profile of bilingual children with neurodevelopmental disorders. At the same time, the results highlight writing components that may serve as markers of ASD and DLD in school-aged bilingual children.

The objective of the study was to identify patterns of spelling, stress assignment and punctuation errors that would uniquely characterize each experimental group. Regarding overall spelling errors, the DLD group exhibited the most erroneous performance across the experimental groups, which aligns with developmental literature showing that language is the locus of predominant impairment in DLD (e.g. Leonard, 2014; Van der Lely, 1997). This finding seems to also agree with Ralli et al.’s (2021) study that found the spelling errors rates of Greek-speaking monolingual children with DLD to be considerably higher relative to their TD peers; however, information on the type(s) of spelling errors that drove the group effect is lacking in Ralli et al.’s (2021) study. In the current research, there was a predominance of grammatical and orthographic errors across the three experimental groups, and a very small proportion of phonologically implausible errors types. This suggests that all three groups tended to employ the developmentally sophisticated strategy of using phoneme-to-grapheme correspondences when attempting to spell words in their expository essays. Protopapas et al. (2013) have also reported a negligible proportion of phonological errors in their study of the spelling performance of 8-, 9-, and 12-year-old Greek-speaking monolingual children with and without dyslexia. The overall evidence indicates phonographemic mapping as a systematic and robust spelling strategy across the three groups of children that spoke languages with relatively transparent orthographies.

The experimental groups’ error rates in spelling seemed to diverge with grammatical and orthographic errors. More specifically, the DLD group exhibited higher rates of grammatical errors, i.e. erroneous spellings of the words’ inflectional endings without concurrent distortion of the phonological identity of the word, than their ASD and TD peers. Previous research (e.g. Goodwin et al., 2013; Larkin et al., 2013; Mackie & Dockrell, 2004) has shown that encoding inflectional endings in the written modality constitutes a considerably vulnerable domain in DLD, which could be attributed to a deficit in the children’s rule-based processing for inflectional markers. We also found that the DLD group’s grammatical error rates were modulated by the children’s home language history scores, meaning that the bilingual children that had more exposure in Greek at home up to the age of 4 years tended to make fewer errors in inflectional suffixes in writing as compared to the children with less exposure in Greek. This finding suggests that the amount of exposure in Greek in the early years affected bilingual DLD children’s grammatical awareness and morphological processing, as correct spelling of suffixes in Greek requires knowledge of inflectional morphology. This effect demonstrates that early oral exposure and use of Greek can add to the spelling skills of bilingual children with grammatical deficits.

Interestingly, the children with ASD did not differ from their TD peers in grammatical errors. This finding disagrees with Hilvert’s (2018) study, in which monolingual English-speaking children with ASD made significantly more grammatical errors in their expository essays than their TD peers. Besides testing monolingual autistic children, Hilvert’s (2018) coding design differs from the current study, since it targeted a considerably wider range of grammatical errors, including omitted obligatory tense markers, missing grammatical morphemes on nouns, wrong forms of verbs, and pronoun number or case errors, among others. Furthermore, children’s grammatical errors in Hilvert’s (2018) study mainly consisted of inflectional morpheme omissions which cannot be observed in the data of the current study, since, in a language like Greek, it is only after the affixation of the inflectional suffix that the root can be realized as a word. The finding that the ASD children in the current study did not diverge from their TD peers in grammatical spelling implies that rule-based processing for inflectional endings in writing was a relatively cost-free process for them. The fact that spelling inflectional suffixes was not challenging for the ASD group may be attributed to autistic traits, and more specifically, to autistic individuals’ high ability to engage in systemizing thinking (Baron-Cohen et al., 2009; Smith et al., 2010), which is characterized by a drive towards understanding and predicting rule-based systems that could potentially have boosted the spelling of inflectional suffixes. An alternative explanation for the ASD group’s error-free performance in suffix spelling is that these children had intact language skills which allowed them to compute the words’ grammatical endings correctly. Evidence in favor of the fact that the children with ASD had preserved language competence is that both their verbal intelligence and vocabulary scores were within the normal range and did not differ significantly from their TD peers. The fact that correct spelling of word stems cannot be predicted through rule inferencing but requires root-specific knowledge or/and skilled memory retrieval of the stems’ orthographic representation may be one of the reasons why the ASD group made significantly more orthographic errors than their TD peers. Importantly, orthographic error rates in the ASD and TD groups were significantly associated with the children’s expressive vocabulary skills, which further corroborates the relation between oral language skills and the maturity of the developing orthographic lexicon. This hypothesis proposes that the acquisition of a larger vocabulary boosts the development of phonological awareness by increasing children’s sensitivity to small changes in the words’ sound segments; enhanced phonological awareness in turn facilitates spelling (Metsala, 1999; Wang et al., 2013). The relation between oral language and written production also receives support from the fact that the children’s expressive vocabulary scores were found to be positively correlated with the number of content words in the expository texts of each experimental group. Overall, the significant correlations between spelling, vocabulary and exposure to the dominant language (i.e. Greek) across groups suggest that the differences between TD and ASD, as well as DLD children in the current study are unlikely to be explained by a possible negative influence of bilingualism in children, but rather by language exposure and vocabulary factors.

Stress diacritic placement, along with punctuation errors, appeared to be the hallmark features in the ASD group’s writing profile. Specifically, diacritic placement in roughly half of the content words in ASD children’s expository essays was erroneous (see Table 4). According to the error type analysis, the majority of the errors for the ASD group involved stress misplacement or/and inappropriate placement of more than one diacritic on more than one vowel of two-, three-, four- and five-syllable words. As such—and in spite of the fact that stress on a phonological word in Greek is restricted to the last three syllables—there were instances of written words with antepenultimate and anteantepenultimate stress in ASD children’s writing. According to the results, stress misplacement and superfluous diacritic addition errors for the group with ASD were significantly more than their TD and DLD peers. The DLD children, on the other hand, have made more stress omissions than TD and ASD children, and thus, DLD children’s divergence from the TD group was less striking as compared to the ASD group. DLD children’s use of stress in writing was significantly related to their vocabulary scores, which means that stress errors were mainly committed by children with low oral lexical skills. This seems to align with previous research claiming that the most important source of stress assignment information for Greek is lexical, and that lexical cues have a major contribution to children’s ability to detect the location of stress in a word (Protopapas, 2006; Grimani & Protopapas, 2017).

The evidence on stress errors suggests that stress diacritic placement was a persistent problem for DLD, and especially for ASD children, which is surprising given that the stress diacritic and its obligatoriness are taught at Greek schools as part of regular reading instruction from the beginning of first grade. The difference in the stress error pattern between the group with DLD and the group with ASD, i.e. prominent stress diacritic omission for the DLD children, and misplacement or/and superfluous diacritic misuse for the ASD children, speaks in favor of the view that stress assignment may be derived from information beyond the boundaries of the individual word. As already mentioned, stress assignment in Greek can be influenced by both the words’ segmental phonological properties and the phonological properties within the scope of the word’s adjacent items (Botinis, 2011; Revithiadou, 1999). If stress assignment operates at both the word and phrase level, one may assume that learning how to correctly use stress in writing is adaptive to lexical, syntactic and prosodic constraints which would increase demands for the ASD children whose language deficits go beyond the lexical domain. Future work should further investigate the manipulation of stress in monolingual and bilingual child populations with DLD and ASD to shed more light on the locus of impairment that underlies their erroneous performance.

Finally, both DLD and ASD groups exhibited significantly higher error rates in punctuation than their TD peers, yet, the two clinical groups selectively struggled with distinct aspects of punctuation use. More specifically, the children with DLD tended to omit punctuation markers considerably more than TD and ASD children, while the group with ASD exhibited significantly more extensive misuse of full stops, question marks and commas in their expository essays than the rest of the experimental groups. As already mentioned, punctuation in writing serves a pragmatic function since it contributes to the chunking of text into coherent units on the basis of syntactic and discourse knowledge, and thus aids the extraction of pragmatically felicitous meaning. The appropriate use of punctuation shows that an individual has good syntactic parsing and pragmatic abilities, and it helps to bring thought into writing for which voice, intonation, volume, tone, pauses are used while speaking (Ferreiro & Pontecorvo, 1999). We assume that the inappropriate use of punctuation marking, which was a hallmark feature of ASD children’s performance in the current study, reflects a more deviant or improper implicit prosodic marking in written production as compared to punctuation omission, and that this pattern may be attributed to pragmatic deficits, which have been widely acknowledged to characterize autistic children’s language skills, or/and to ASD children’s putting less weight on implicit prosodic cues to convey meaning in the essay (Durrleman & Delage, 2016; Eigsti et al., 2007; Walenski et al., 2006). Interestingly, both DLD and ASD children’s punctuation performance correlated with their current language use scores, further suggesting that language exposure and use of Greek over time has significantly modulated the children’s manipulation of implicit prosodic cues in their expository essays.

The overall evidence of the study confirms that writing is more challenging for DLD and ASD children relative to their TD peers at several levels of the writing process, including spelling, stress and punctuation. Crucially, the children’s error rates and patterns across the various writing constructs were able to distinguish between the DLD and the ASD group. Specifically, DLD children exhibited greater vulnerability in spelling as opposed to ASD children who were more erroneous in stress and punctuation. Crucially, omission of stress and punctuation was the dominant error type for DLD children, while the children with ASD tended to use stress diacritics and punctuation markers in their writing, but were prone to either misplacing or/and overusing them, which further implies that the children with ASD possessed basic knowledge of suprasegmental features but could not integrate them in writing. The findings suggest that the deficits underlying the writing skills of the DLD and ASD groups may be distinct—for example, that the writing performance of the ASD group is rather affected by deficits that span discourse units rather than single words, while DLD children’s writing deficits are rather reflected in low-level properties of the text, such as spelling.

Turning to the implications of this research, these can be considered at two levels. From a theoretical perspective, the findings may inform hypotheses about the mechanisms underlying writing performance in bilingual children diagnosed with DLD and ASD, and allow us to explain the specificities of spelling, stress and punctuation errors across the two disorders. More specifically, the present study provides evidence that error types in writing can distinguish the deficits of autistic bilingual children from those of children with DLD or/and their TD peers, mainly by being based on stress and punctuation use. Crucially, the group with ASD tended to use superfluous stress diacritics and punctuation in words and discourse units, respectively, considerably more than the rest of the experimental groups, while the bilingual group with DLD faced difficulty mainly with spelling. The specific patterns suggest that the deficits in ASD and DLD surface at distinct levels of writing, namely, at the word and the discourse organization level, respectively. At a more practical level, examining writing skills is important to develop appropriate, targeted writing interventions across the school-age years that meet the literacy needs of bilingual children with DLD and ASD. As such, the findings of the current study may be beneficial to educators developing and implementing instructional practices to support writing and broader literacy development in the specific populations.

The current study has four limitations. First, because of the limited number of standardized language ability tests in Greek, it is unknown whether a more fine-grained approach to the children’s language profile would reveal more associations between language skills and the children’s writing performance. Second, the small sample size of the children that have participated in the current study may have limited the statistical power of the results. Third, the children’s bilingual experience could have been more reliably reflected through calculating the actual percentage of time during a typical week that the children heard/spoke a given language across numerous interlocutors and settings, rather than through computing the percentage of interlocutors with whom the children spoke the two languages. The fact that language exposure and use were conflated in the bilingual questionnaire, and that time-based measures of language exposure at home and at school were not calculated, did not allow us to scrutinize the effect of each factor on the bilingual children’s writing performance; also, the lack of weighting across the various interlocutors the bilingual children interacted with, prevented us from assessing their loading on the participants’ bilingual experience. Finally, more measures which include theory-of-mind and executive functions could have shed more light on the evaluation of the bilingual ASD and DLD children’s writing performance. Therefore, further studies are warranted to investigate the extent to which the children’s writing performance is affected by their executive functions and theory-of-mind skills, as well as how their writing skills might differ across writing genres. In fact, the current research is followed-up by a study of the same children’s writing skills in personal, autobiographical topics, and the preliminary evidence is promising in highlighting significant effects of writing genres on DLD and ASD children’s spelling, stress and punctuation errors.