The term “crosslinguistic influence” describes cognitive and linguistic processes that allow to apply knowledge of one language to another language (Moattarian, 2013). Cross linguistic influences in language production among speakers of multiple languages have been attributed to learner characteristics, including number of languages spoken and linguistic proficiency. Cummins' Interdependence Hypothesis in bilingualism, for example, claims that second language (L2) skills are dependent in part on L1 language abilities, and that language skills will transfer from L1 to L2 given sufficient exposure and level of proficiency in L2. Thus, well-developed skills in L1 should support the development of similar skills in L2 (Cummins, 1979; Verhoeven, 1994). Within a broader framework of third language acquisition, it has been suggested that third language learners have a larger linguistic repertoire than second language learners, have stronger metalinguistic skills (Jessner, 2008; Huang, 2018), are more experienced with language acquisition processes, and thus, may have developed a unique set of language learning strategies that they can access (Jessner, 2008; Cenoz, 2013). Moreover, as a result of their wider linguistic repertoire, multilinguals may activate both direct routes and indirect routes to additional language learning (Hirosh and Degani, 2018). According to Hirosh and Degani (2018), direct routes include transfer of linguistic skills and knowledge, whereas indirect routes represent cognitive factors such as metalinguistic awareness and working memory.

In addition to learner attributes, cross linguistic relationships are also affected by properties of the languages in contact, such as general typological proximity or more specifically structural similarities. According to the Linguistic Proximity Model (Westergaard et al., 2017), structural similarities among the languages of a multilingual learner may lead to facilitative cross linguistic influences, particularly when the learner is aware of fine-grained variations across the grammars of their language repertoire. Non-facilitative influences may occur when the learner does not have a solid grasp of particular linguistic input in the target language and erroneously attributes shared properties between the target language and any of the already learned languages. Within this model, the specific patterns of influence will be determined by the areas of cross language overlap (Westergaard et al., 2017). However, recent research indicates that there is a bidirectional transfer of skills from L1 to L2 (forward transfer) and from L2 to L1 (reverse transfer) (Jarvis and Pavlenko, 2008; Kim and Piper, 2019). Pavlenko and Jarvis (2002) examined bidirectional influences of L1 (Russian) and L2 (American English) oral production and found that not only did Russian influence English, but there was also a direct influence of English (L2) on L1 (Russian) oral production. Kim and Piper (2019) found bidirectional influences between Kiswahili and English (official languages in Kenya) literacy skills. The present study examines cross linguistic influences in EFL oral narrative production among monolingual Hebrew speakers and multilingual Arabic speakers.

Oral narratives have long been recognized as a valid measure of linguistic growth in monolingual and bilingual children from different linguistic backgrounds (Berman and Slobin, 1994; Pavlenko, 2002; Soodla and Kikas, 2010). The interest in narratives is fueled by different disciplines: linguistics, sociology, cognitive psychology (Iluz-Cohen and Walters, 2012). Producing a narrative requires the integration of different cognitive and linguistic skills, e.g., interaction of lexical, morphosyntactic, and general discourse knowledge (Boerma et al., 2016), along with metacognitive skills (Cortazzi and Jin, 2007; Kupersmitt et al., 2014).

Growing narrative abilities also coincide with the development of executive functions, in this instance the ability to plan, i.e., organizing the story in sequential order (Friend and Bates, 2014), making narratives a perfect vehicle to access linguistic and cognitive growth in children. Expressing one's own and others' perspectives through lexical diversity and application of appropriate morphosyntactic structures (Moonsamy et al., 2009) involves both production and comprehension of multiple utterances and represents the current level of linguistic and conceptual knowledge of an individual (Justice et al., 2010). For these reasons, narratives have been examined from the point of view of (1) global characteristics, i.e., macrostructure, or producing a narrative based on the understanding of the thematic orientation, as it accounts for the “mental representation of events” (Berman, 1995, p. 287), and (2) overall grammatical complexity, e.g., microstructure, which includes lexical diversity as well as morphosyntactic knowledge, to represent the meaningful use of grammatical structures to allow listener's understanding (Justice et al., 2010). Moreover, strong correlations were found between these two structures of narratives, indicating that better lexical and morphosyntactic knowledge results in better global representation of narrative (Terry et al., 2013). Research also suggests that there is a strongly implied interaction between lexical and grammatical knowledge across different languages, such as Italian, Hebrew, Icelandic, etc. (Thordardottir et al., 2002). Moreover, Thordardottir et al. (2002) suggested that this interaction is due to a single mechanism which supports the development of both lexical and grammatical knowledge. Speakers of multiple languages also showed interaction between lexical and grammatical knowledge, when they produce narratives in their non-native languages (Marchman et al., 2004; Simon-Cereijido and Gutiérrez-Clellen, 2009; Gagarina et al., 2015; Gagarina, 2016). However, there is no consensus regarding which particular factors, e.g., lexical, morphosyntactic, or general cognitive skills, may play mediating roles among learners of foreign languages in general, and among learners of EFL specifically, in their attempts to produce cohesive oral narratives in English. Despite the proposed interaction between lexical and morphological knowledge, there is also evidence that lexical and morphological knowledge may separate among second language learners, depending on the morphological tasks' reliance on general vocabulary (Shahar Yames et al., 2018).

Since the time when Labov (1972) delineated structural and functional aspects of narratives as macro- and micro-structure, or cohesion and coherence, researchers have consistently used specific markers to account for the completeness of narratives. The macrostructure assessment includes story setting (time and place), identifying the protagonist, references (use of pronouns with antecedents), and attempts to solve the problem (Pearson, 2002; Uccelli and Păez, 2007; Heilman et al., 2010). At the same time, when producing an oral text, a narrator in EFL must also show the ability to use appropriate lexical and morphosyntactic structures (microstructure, or coherence) in the non-native language in order for the narrative to be understood by the listener (Iluz-Cohen and Walters, 2012). This includes extensive vocabulary, appropriate knowledge of morphological inflections, as well as sentence complexity, e.g., use of cohesive devises, such as conjunctions. While research showed that the stories produced by multilingual speakers in their L1 and L2 may not really differ in the macrostructure, there were substantial differences in the appropriate use of the linguistic elements in L2 production (Pearson, 2002). Moreover, neither Pearson (2002), nor Uccelli and Păez (2007) later found any cross-linguistic transfer in children's narratives. On the other hand, Castilla et al. (2009) did find correlations across L1 and L2 (Spanish and English in this instance) in morphosyntactic knowledge, which indicated that there are some aspects of linguistic knowledge that actually can be transferred cross-linguistically. More recent research with Cantonese-English speaking preschoolers showed that Cantonese micro- and macrostructure predicted English micro- and macrostructure, again suggesting cross-linguistic transfer, as well as bidirectional influences (Rezzonico et al., 2016). This particular study is important, as it examined cross-linguistic transfer among typologically different languages. There is also substantial evidence that the transfer can occur on the level of morphosyntax, e.g., development of relative clauses in Cantonese-English bilinguals (Yip and Matthews, 2007), as well as in morphological awareness. For example, Pasquarella et al. (2011) showed bidirectional transfer of morphological awareness between English compound morphology and Chinese vocabulary acquisition, as well as English compound morphology predicting Chinese reading comprehension. However, the abovementioned studies examined written language. Nevertheless, based on these results, we may infer which L1 skills might contribute to producing L2 narratives among bilingual individuals, and individuals who are learning EFL with typologically different L1s. In other words, it is possible to postulate that there is a mediating effect of specific L1 skill that will account for cross-linguistic transfer even in the oral language domain.

Arabic and Hebrew are Semitic languages. As such, they are both morphologically rich with words being formed by combining roots comprised of two to four letters with patterns or templates: these are usually inserted between the root letters in order to create words with meaning. Some words are also formed by affixation at the beginning or end of the word. Verbs are inflected for number, gender, person (first, second, and third), and tense, or aspect to create three tenses: past, present, and future. Nouns are inflected for number and gender. Pronominal suffixes can be attached to verbs and nouns, thereby adding an additional layer of morphological complexity to words, so that most words are multi-morphemic. Research indicates that as a result of the morphological complexity of the language, Arabic and Hebrew speaking children show signs of morphological sensitivity at an early age and are well able to attend to internal word structure (Ravid, 2001; Gillis and Ravid, 2006; Saiegh-Haddad and Taha, 2017; El Akiki and Content, 2020). Moreover, there is empirical evidence that young bilingual Hebrew and Arabic speakers outperform monolinguals on tasks of derivational morphology, as a sign of positive cross linguistic influences (Asli-Badarneh and Leikin, 2019). These findings suggest that cross-linguistic influences can be traced among languages that belong to the same typological group, such as Hebrew and Arabic (Schwartz et al., 2016; Asli-Badarneh and Leikin, 2019). It has been further suggested that since Arabic morphology is more complex than Hebrew morphology, Arabic speaking bilinguals may have an advantage when they transfer skills from their more complex L1 morphological system to the less complex L2 Hebrew system (Chen and Schwartz, 2018). However, despite shared linguistic and structural features between Arabic and Hebrew, Arabic is unique in that it is a diglossic language, characterized by linguistic distance between the spoken dialects and the Modern Standard form (Saiegh-Haddad, 2003). Moreover, there may be great variance across spoken dialects in the areas of phonology, morphology, and semantics (Saiegh-Haddad and Henkin-Roitfarb, 2014), while the written form is stable. Children begin to learn the Modern Standard form formally when they enter school, although they are exposed to it informally in their environment from an early age though literature and media. Thus, Arabic speakers study academic subjects in school in Modern Standard Arabic (MSA), a language that is not their home language (spoken dialect). As a result of their exposure to two forms of the language from an early age, some researchers suggest that Arabic speaking children are, in essence, can be considered bilingual (Eviatar and Ibrahim, 2000), e.g., the spoken form is their L1 and the Modern Standard form is their L2, even before they begin to study Hebrew. Studies have also indicated that the linguistic distance between the two forms of Arabic language may result in difficulties with the acquisition of the Modern Standard form, and other literacy related skills (Saiegh-Haddad, 2003, 2007). Moreover, these difficulties persist beyond the initial years of literacy acquisition (Abdelhadi et al., 2011).

In contrast to Semitic languages, English is a Western Germanic language with a relatively simple morphological structure. Morphemes are affixed onto base words producing multimorphemic words, which can be inflectional or derivational in nature. Inflectional morphemes change tense or number of free morphemes, whereas derivational morphology can change the part of speech or the meaning of the base word. In addition to the three basic tenses (past, present, future), English verb tenses are complicated by the use of aspectual forms, which contribute an additional level of linguistic complexity. And yet, English has been described as a morphologically impoverished language due to the fact that, in comparison with Semitic languages, there are hardly any markings of agreement for gender, number, or person (DeKeyser et al., 2010; Tsarfaty and Sima'an, 2010).

In addition, word order within a sentence in Semitic languages, is rather flexible. This is because syntactic characteristics are determined based on morphological information embedded within words, above and beyond word order within a sentence (Tsarfaty and Sima'an, 2010). Thus, while the dominant word order for Hebrew sentences is SVO, and VSO for Arabic sentences, variability in word order appears and is accepted in both languages (Schwartzwald, 2011). Moreover, verbless constructions exist in both languages. In contrast, English word order within sentences is relatively fixed according to the SVO pattern so that grammatical patterns are reliant on syntax along with meaning. The present study examines oral language skills in EFL among Arabic and Hebrew speaking pupils in sixth grade, thus, these particular differences between Hebrew, Arabic and English language structure provide an important frame of reference.

English is a global language (Crystal, 2012). It is also the most widely taught foreign language today. In Israel, English holds a unique status. It is a semi-official language and the first foreign language studied in all schools. It is the key to international economic growth, and social communication, as well as the gatekeeper to academic advancement. It is also a compulsory subject for the high school matriculation certificate. Moreover, English proficiency is required as a prerequisite for entrance into higher education.

Based on the language learning policy of the country, Arabic speakers are taught in school in Arabic, and Hebrew speakers are taught in Hebrew. Arabic speaking pupils must also learn Hebrew as their second written and third spoken language, beginning in the second grade or even earlier (Amara, 2018). Hebrew speaking pupils only begin to study Arabic in the 7^th grade, if at all. Formal instruction in English as the first foreign language (EFL) for both populations begins in 3^rd grade.¹ Thus, while Hebrew speakers study EFL as their second language, native Arabic speakers study EFL as the 3^rd written language and fourth spoken one. Moreover, the status of each of the languages is evident from their appearance in the linguistic landscapes of the community and the school settings. Among the Hebrew speaking communities, Hebrew is dominant in the linguistic landscape, followed by English. In most Arabic speaking communities, Arabic is the dominant language in the linguistic landscape followed by Hebrew. English is hardly present, except for the region of East Jerusalem where it is the second most prominent language after Arabic (Amara, 2018). A similar situation exists within the school linguistic landscapes, at least in the Arabic speaking population, where Arabic is the most prominent language followed by Hebrew with little or no representation of English (Amara, 2018).

Based on the national guidelines, all pupils in Israel learn EFL according to the same English curriculum and teaching materials, regardless of their L1 backgrounds. The first year of study (3^rd grade) is dedicated to building basic oral skills and introducing the letters of the alphabet. Pupils in 3^rd grade study English for two 45-min lessons a week. From 4^th–6^th grades pupils receive four 45-min EFL lessons a week. In the 4^th grade, pupils continue to build their lexical knowledge while beginning to learn how to read. By the last year of elementary school (6^th grade), pupils are expected to reach a basic level of oral and written language proficiency.

In reality, however, after the first year of oral language instruction, there is little, if any time specifically devoted to the further explicit development of authentic oral language skills (Al Hosni, 2014). While the revised elementary English curriculum¹ relates to oral language skills, the expectation is that the pupils will be able to manage short, rehearsed utterances, and not necessarily produce spontaneous language. Moreover, the reality of large numbers of pupils in each class does not really allow for individual practice or use of oral language skills in school. Additionally, while the revised curriculum highlights the importance of English oral skills in the classroom, many teachers prefer to use their L1 (Orland-Barak and Yinon, 2005; Timor, 2012), which further limits exposure to spoken English.

The present study explores the contribution of cross linguistic influence to oral narrative production in EFL among a sample of Hebrew and Arabic speakers. To our knowledge, this is the first study that directly examines cross language influences and possible specific influentual factors between Semitic L1 and EFL through oral narratives. Previous studies concerned with narrative production among bilingual individuals compared macro- and micro-structures in relation to the differences of the output between L1 and L2 of participants (Uccelli and Păez, 2007; Gagarina, 2016; Lucero, 2018). However, there was no specific data relating to which L1 skill(s) had a mediating effect on L2 narratives, as well as the possibility that different EFL skills may prove to be more influential for oral EFL production among Arabic and Hebrew speakers. The present study addresses this gap by examining oral narratives produced by 6^th grade speakers of Arabic and Hebrew in their 4th year of learning EFL, within a cross-linguistic framework, in search for possible mediators between Semitic languages and English. For these purposes, we did not compare L1 and EFL narratives, but rather examined the proficiency of the participants in linguistic and literacy domains in L1 Hebrew and L2 Modern Standard form of Arabic (MSA) and EFL, as well as general cognitive skills, as possible contributing factors to the quality of EFL oral narratives.

Our research questions were as follows:

The participant pool was comprised of two linguistic groups: 86 native mono-literate sixth grade Hebrew speaking pupils for whom English is the second language being learned in school (43 females), and 85 native 6^th grade Arabic speaking pupils, who learn spoken language as L1, MSA as L2, learn Hebrew as L3 and then begin learning EFL. Therefore, this group can be considered “multilingual,” as English is third written language and the fourth spoken language being learned in school (44 females). All of the participants were between the ages of 10 and 11 at the time of data collection. The Arabic speaking pupils were chosen from four different schools, and the Hebrew speaking pupils were chosen from six different schools in the central area of Israel. As the Arabic speaking participants were from several different cities, their spoken dialects varied in relation to the city where the participants lived. The average socio-economic index for the Arabic speaking school was 5.66, and for the Hebrew speaking schools 2.95 on a scale of 1–10 where 1 is the highest and 10 is the lowest. However, very large discrepancies can be found within the Arabic speaking populations that were included in the study. As a case in point, in our groups, there were significant individual differences on task results, as can be seen in the very large standard deviations in Table 1². The protocol of this study was approved by the Ministry of Education Chief Scientist Bureau. Pupils chose to participate on a voluntary basis and all parents signed a consent form. Pupils with learning disabilities or pupils who are fluent in languages in addition to Hebrew or Arabic were excluded from the sample.

Language tasks in Arabic and Hebrew were adapted from existing standardized tools (Arabic—Asadi et al., 2015; Hebrew—Shany et al., 2006). All the Arabic tasks were administered in the MSA (which, as mentioned earlier, is considered the first written language that the Arabic speakers acquire but the second spoken language). Arabic speakers were also tested in Hebrew—their second written and third spoken language. While most of the Arabic and Hebrew tests have norms, some of the tasks were shortened for the purposes of the present study and therefore the norms can only be used to give an indication of performance trends. Further, the Hebrew tasks were not designed to be administered to non-native Hebrew-speaking pupils, so the scores for the Arabic speakers cannot be evaluated with relation to the Hebrew norms. All tests, in both languages, were administered by native speakers of the language of the participants. In what follows, the tasks will be described, according to language of administration.

The first research question addressed the contribution of different cognitive and linguistic skills in Hebrew, as L1 to the ensuing EFL narratives. Our second research question addressed the contributions of cognitive and linguistic skills in Arabic as the first written language of the participants, along with Hebrew as the second written language and third spoken language acquired by Arabic speakers, and micro- and macro-structures of EFL narratives. The third question explored the contribution of EFL linguistic skills (ERC, EMA) to EFL oral narratives among the monolingual Hebrew and multilingual Arabic speakers. In what follows, results will be presented for each language group separately.

Our final research question addressed the possible role of L1/L2/L3 skills in the production of EFL narratives. In essence, we were looking for a specific L1/L2/L3 skill that may give additional support to the quality of EFL oral narratives, based on the MacKinnon et al. (2000) suggestion that after establishing the relationship between predictor and outcome variables it is common to identify a third variable (e.g., specific skill) that may play a role in that relationship. That skill (or variable) should have an influence on both, the predictor variable and outcome variable. This third variable should relate to both factors and possibly enhance the relationship between them. If we identify the predictor variable as X and the outcome variable as Y, variable Z (confounding variable) should have an effect on both, X and Y, therefore signifying cross linguistic transfer from L1(L2/L3) to EFL. These relationships are represented in Figure 1. In what follows, we present the results for the regression analysis for confounding factors.

Among the Hebrew speakers, this regression analysis indicated that ERC was the strongest EFL predictor for total oral narrative scores [R² = 0.28, F_(1,84) = 31.026, p ≤ 0.001]. Contrary to the possible assumption that HRC may have some influence on the predictor and outcome variables based on the significant relationship between these variables found in the correlational analysis, this was not the case for Hebrew speakers. While there was strong correlation between ERC and HRC, no correlation was found between the Total Narrative score and HRC, therefore, violating the assumption that the confounding variable should have an influence on both the predictor and the outcome variable (MacKinnon et al., 2000). However, HMA_pseudo was found to relate to both, ERC and Total Narrative score. The fact that HMA_pseudo can be the influential variable in EFL Total Narrative score is supported by numerous studies, that suggest strong relationship between MA and RC and MA and oral language development (Tomasello, 2005; Verhoeven, 2017). The choice of HMA_pseudo as possible variable that may enhance the strength of the relationship between EFL skills and EFL narratives was dictated by the strong metalinguistic component of this particular task: manipulation of pseudo words implies strong knowledge of the underlying mechanism in creating appropriate morphosyntactic elements (Shahar Yames et al., 2018). Therefore, based on the assumption of the confounding hypothesis regarding the effect of a third variable in enhancing the relationship between the predictor and outcome variable, we wanted to see if HMA_pseudo would have an effect on both, the ERC and Total narrative score. Indeed, the effect of HMA_pseudo on ERC was significant, suggesting that it would increase the ERC (predictor variable) score by 12% [R² = 0.122, F_(1,84) = 11.48, p < 0.001]. The effect of the HMA_pseudo on Total Narrative (outcome variable) score was also significant and suggested the increase on the Total Narrative score of 7% [R² = 0.073, F_(1,84) = 6.601, p < 0.01). Thus, we could say that HMA_pseudo had a significant effect on both, the predictor and the outcome variables, which was confirmed by bootstrapping procedures showing a confidence interval range from 0.01 to 0.13 and Sobel test (z = 2.08, p = 0.04) (See Figure 2).

Among the Arabic speakers, the regression analysis for confounding factors indicated that English MA was the strongest predictor for the Total oral narrative scores [R² = 0.302, F_(1,82) = 35.445, p ≤ 0.001], accounting for 30% of the variance. Again, our assumption that AMA may support EFL narratives did not yield the expected results. However, among Arabic speakers, AMA_choice and HMA_pseudo variables showed the strongest and most significant correlations. Moreover, significant correlations among this group were also found between HMA_pseudo and EMA, as well as Total Nrrative score. As stated above, the relatively stronger performance of Arabic speakers in the task of HMA_pseudo, as opposed to HMA_real signified their developed metacognitive and metalinguistic skills, based on the exposure to multiple languages (Jessner, 2008). Therefore, we assumed that HMA_pseudo may have an effect on EFL Total Narrative score among this group as well. Indeed, HMA_pseudo had a significant effect on EMA among Arabic speakers, accounting for 19% of the variance [R² = 0.194, F_(1,82) = 19.55, p < 0.001]. The effect of HMA_pseudo on the Total Narrative score in this group was even stronger than among Hebrew speakers, accounting for almost 13% of the variance [R² = 0.129, F_(1,82) = 12.003 p < 0.001] and suggesting the cross language influence. Again, the bootstrapping procedures confirmed the significance of the effect, with a confidence interval range from 0.023 to 0.177, as did the Sobel test (z = 0.2.27, p = 0.02). Therefore, we assumed that HMA_pseudo, was an influential variable that significantly impacted narrative performance in this group of participants (see Figure 3).

Our first and second research questions addressed the influence of cognitive skills, Hebrew and MSA, as well as Hebrew as L3 for Arabic speakers, on all aspects of EFL narrative production. Our hypotheses stated that we should see cross linguistic influences of native language skills on EFL narratives based on levels of language proficiency, as postulated by the Interdependence Hypothesis (Cummins, 1979, 2014). Absence of the observed transfers could be attributed to the tenet of the Linguistic Proximity Model (Westergaard et al., 2017), suggesting that based on the typological distances between language non-facilitative influences may occur when the learner does not have a solid grasp of particular linguistic input in the target language. Based on these theoretical assumptions, we first looked at the overall psycholinguistic profiles of both linguistic groups in our study, beginning with the Hebrew speaking sample.

As seen in the descriptive statistics, Hebrew speakers showed high scores on measures of L1 proficiency, with higher scores for morphological awareness of real over pseudo words. In EFL, scores on morphological awareness (EMA) were at a moderate high level of proficiency. Scores for reading comprehension (ERC) were the lowest of the English linguistic scores, possibly indicating that despite strong knowledge of individual words, these children still lack the ability to recruit their vocabulary skills in order to comprehend language in context (Masrai, 2019). The scores for EWR as a measure of PM, were close to ceiling. Of interest here is the suggestion proposed by Kaushanskaya et al. (2011) that bilingual individuals may rely on PM for vocabulary retrieval in L2, therefore boosting reading comprehension as well as oral narratives. Furthermore, the fact that the PM measure (EWR) correlated with Hebrew morphology tasks supports the notion that PM is important not only for vocabulary building, but also for the acquisition of morphology independent of native language (Williams and Lovatt, 2003).

Our subsequent analyses, related to our first research question, aimed to identify the specific skills that show cross linguistic influences from L1 to EFL, and to see how these skills support oral language narrative production in EFL. Of interest was the finding that whereas Hebrew morphological awareness tasks (HMA) (real and pseudo words), as part of within language associations, did not show any correlations with Hebrew reading comprehension (HRC), both HMA tasks significantly correlated with EFL reading comprehension. It may be inferred, then, that strong correlations between HMA and ERC suggest that skills in this area of L1 linguistic knowledge are very important for understanding the syntactic structure of L2 (EFL in this instance) in order to fully comprehend written text (Chen and Schwartz, 2018). This correlation exemplifies a cross-linguistic influence, and provides additional support to previous research, which indicated that morphological awareness could transfer from L1 to L2 even among typologically distant languages (Geva, 2006; Yip and Matthews, 2007; Pasquarella et al., 2011). Since HMA also showed strong association with EFL micro-, macro- and total score elements of EFL narratives, we can assume that L1 morphological awareness, as part of morphosyntactic knowledge (James et al., 2021) is a strong component of EFL oral production. Not surprisingly, the EWR as a measure of PM, showed strong associations with both HMA tasks, as well as all EFL narrative structures scores. This particular association provides strong support for the notion that PM is an important underlying cognitive/linguistic skill required for language acquisition. This relationship is seen not only across different L1 languages, but also in acquiring foreign languages as well (Service and Konohen, 1995; Masoura and Gathercole, 2005; Verhagen and Leseman, 2016). Moreover, the regression analysis highlighted the contribution of PM to the production of EFL narratives, solidifying our assumption that PM is important for acquisition of EFL oral language skills. Since Hebrew speakers showed high language proficiency in their ambient language, it supports the main tenet of Interdependence Hypothesis (Cummins, 1979, 2014). And yet, despite strong associations between the HMA tasks and all elements of EFL narratives, none of the Hebrew linguistic tasks added to the quality of EFL narratives beyond the contribution of EWR as a measure of PM.

Our second research question addressed the relationship between MSA (L2), L3 (Hebrew, for Arabic speakers), and EFL narratives among the Arabic speakers in the current study. It implied that there may be unique allocation of cognitive and linguistic resources among multilingual Arabic speakers. Our results indicated that our hypothesis was correct, as the results showed different within- and across-languages patterns of associations between tasks, as well as different cognitive and linguistic contributors to oral EFL narrative skills from those of the Hebrew speakers.

As a group, Arabic speakers showed average native language proficiency (according to existing norms, Asadi et al., 2015) and EFL proficiency within the average range, as measured by EMA and ERC scores. While the EFL tasks were not standardized as they were designed for the present study, achieving a result within the 33^rd-50^th percentile can be considered within average range³. This was not surprising given the diglossic nature of Arabic and the notion that the existence of two forms of Arabic (the spoken dialect and MSA form) may interfere in acquisitional processes (Saiegh-Haddad, 2003, 2007). Moreover, the Arabic speakers in this study also had to contend with Hebrew as their third spoken and second written language. However, Arabic speakers showed strong results on EWR task as a measure of PM, as could be expected from individuals who are exposed to multiple phonological systems (Bialystok et al., 2003). This group also showed different pattern of results on the HMA tasks. In contrast to the Hebrew speakers, the Arabic speakers scored higher on the pseudo word derivation task than on the real word derivation task. This could be due to the fact that deriving a real word is more reliant on lexical knowledge, whereas deriving a pseudo word is more dependent on the underlying morpho-syntactic knowledge and internalization of rules and patterns (Williams and Lovatt, 2003). Thus, their scores would indicate that while their word level knowledge in Hebrew may be lower, their understanding of how words are formed may be heightened as a result of their experiences with multiple languages (Kuo and Anderson, 2006). Nevertheless, we also saw the same inter-language relationships among Arabic speakers in MSA measures, general cognitive skills and EWR, as a measure of PM, as were found among Hebrew speakers again indicating the important role of PM in language acquisition across languages.

Correlational analysis showed very strong associations between Raven (general cognition) and ARC, and every EFL measure, which was not the case for Hebrew speakers. Therefore, we can postulate that general cognitive abilities play an important role in language acquisition among this group of children. The reliance on general cognition for linguistic tasks has been postulated to be a driving force in language development (Clark, 2004; Tomasello, 2005). Tomasello (2005) specifically identified “pattern-finding…” as a “…cognitive skill involved in the abstraction process” (p. 193), which leads to integration of perceptual information into children's linguistic repertoire (Clark, 2004). However, Raven also showed strong correlations with ARC and HMA for real words (HMA_real). One possible explanation could suggest that Arabic speakers were relying on general cognitive resources to retrieve vocabulary items for reading comprehension in their native language, as well as recalling Hebrew vocabulary words to apply derivational process by association, rather than relying only on PM, although previous research found no connection between Raven and vocabulary (Ordónez et al., 2002). On the other hand, we also saw strong associations between PM and AMA and ARC, as well as with all aspects of narrative structure. This finding is not surprising since PM is strongly related to vocabulary development (Gathercole et al., 1999), and has been found to strongly associate with both MA and RC, a skill which is comprised of decoding abilities, lexical knowledge, as well as knowledge and understanding of morphosyntactic structures (Hipfner-Boucher et al., 2015). As Arabic morphology is very rich, PM provides the basis in the development of all of the abovementioned skills, and in turn may support the acquisition of MA in additional languages (Verhagen and Leseman, 2016). Indeed, there were significant correlations between HMA and AMA, as an example of cross-linguistic transfer between two typologically close languages. Additionally, there was a cross-linguistic transfer in MA between Arabic and EFL, a typologically distant language. These results could be interpreted in relation to the Interdependence Hypothesis. As PM was one of the strongest skills exhibited by Arabic speakers, we may postulate that the exposure to more than one phonological system has strengthened PM skills in this group. This conclusion gains support from numerous studies showing that bilingual children have stronger PM (Bialystok et al., 2003; Parra et al., 2011; Zaretsky, 2018).

Our assumptions from the onset of the study were that EFL skills will directly contribute to the EFL oral narrative production. The question was: will there be differences in these contributions among Hebrew and Arabic speakers. To answer this question, we conducted an additional regression analysis for both groups. Indeed, there were between group differences in how and which EFL skills supported oral language narratives. Among Hebrew speakers ERC made the largest contribution to the Total narrative score (combination of micro- and macro-structures). This contribution is not unexpected, as RC involves many elements that comprise narratives, as well as being a combination of lexical and morphosyntactic knowledge on its own. Moreover, Hebrew speakers showed above average scores on ERC (while there are no standard norms for this task since it was designed for this study, achieving a score between 60^th and 75^th percentile can be considered above average. See footnote 3 above), as well as high average score on HRC. This result is in line with the Interdependence Hypothesis, which postulates that L1 proficiency can support L2 skills. The fact that PM also supported EFL narratives in this group, was an additional conformation that this cognitive/linguistic skill is intimately involved in every aspect of language acquisition, from early acquisition of vocabulary (Gathercole et al., 1999) to supporting MA and metalinguistic skills (Kupersmitt et al., 2014; Boerma et al., 2016), as well as being an important skill in the acquisition of narrative abilities and promoting grammatical competence in later stages of L2 acquisition (O'Brien et al., 2006).

As we hypothesized, Arabic speakers relied on different EFL skills to support EFL oral narratives. The multilingual Arabic speakers were strong on the tasks that measured AM. Moreover, based on their HMA performance, specifically on the HMA pseudo word task, we can infer that they have better understanding of morphological procedures, above and beyond specific lexical representation at the word level. This finding is a direct conformation of the proposal presented by Shahar Yames et al. (2018), that suggested a separation of lexical and morphological knowledge among learners of additional language, who show much better performance on morphological tasks that are not dependent on extensive lexical knowledge, as measured in their case by pseudo word tasks. As further evidence of strong morphological awareness skills, the Arabic speakers performed within average range on EMA task (see footnote 3). Thus, it was no surprise to see that EMA was the largest contributor to the EFL Total narrative score. This particular finding was of importance as it provided additional support to previous research findings that MA is essential for lexical development (Zhang, 2015), which in turn impacts oral narrative production. Moreover, the correlations between MA tasks in Arabic, Hebrew and EFL support the idea that cross-linguistic influences can be traced between languages that belong to the same typological group, such as Hebrew and Arabic, as well as languages that are typologically distant (Schwartz et al., 2016; Asli-Badarneh and Leikin, 2019), and further highlights the potential for cross linguistic influences across all the languages in the linguistic repertoire for multilinguals (Cenoz, 2013).

These results also highlighted skills that can be considered as contributors which serve to increase the level of oral language production among Hebrew and Arabic speaking children. It has been empirically shown that bilinguals and multilinguals are influenced by one of their languages in activating the processing of another (Dijkstra, 2005), and this activation is seen for different types of information, including syntactic structures (Macizo et al., 2010). So, it is possible to postulate that ERC is the important influencer for EFL oral narratives for Hebrew speakers, i.e., the stronger the ERC scores, the better the EFL narratives, especially since RC and narratives rely on similar component skills. Further, ERC scores for this group were very strong. However, there was a different influence of EFL narrative scores for Arabic speakers, namely EMA. Influence of MA on narratives skills are well-documented, therefore there is no surprise that there should be an influence of EMA on EFL narratives, particularly in this group. This relationship in this case may be explained by the fact that the Arabic speakers may have heightened morphological awareness skills as part of their cross linguistic metacognitive repertoire (Bialystok et al., 2003; Hirosh and Degani, 2018), resulting from their prior experiences with two Arabic morphological systems (spoken dialect and MSA form), as well as Hebrew MA. This could explain how increases in EMA may reflect an increase in the EFL Total narrative score among this population.

Our last research question explored the possibility that there may be a specific L1/L2/L3 skill, that could play an influencing role in enhancing the relationship between the predictor and outcome variables. We chose the confounding hypothesis as an explanatory framework in order to exlore this possibility. The confounding hypothesis proposes a relationship whereby an additional variable should be related to the factors of interest (predictor and outcome). Namely, it should be correlated with predictor and related to the outcome. Moreover, this particular hypothesis would strongly support the tenets of Interdependence Hypothesis.

As seen in the strength of the EFL skills as predictors for EFL narratives, ERC was the predictor variable for Hebrew speakers. The correlational analysis suggested that HMA_pseudo (as the third variable) strongly correlated with ERC, our independent variable (IV, or predictor), therefore fulfilling the first assumption of the confounding hypothesis, namely that the confounding variable and IV should correlate. At the same time, HMA_pseudo was related to the EFL Total Narrative score, i.e., it increased the quality of the narrative, fulfilling the second assumption of the confounding hypothesis. Moreover, this explanation aligns with the tenet of the Interdependence Hypothesis, in suggesting that strong L1 skills will support L2 skills, indicating a CLI.

Among Arabic speakers, EMA was the predictor variable for the Total narrative score, as evidence of stronger metalinguistic knowledge in this multilingual group as opposed to weaker specific linguistic knowledge. However, AMA was not found to explain the relationship between the predictor (EMA) and the outcome (EFL Total narrative score), e.g., it was not a factor in enhancing the outcome. On the other hand, the strong performance of Arabic speakers on HMA_pseudo task also suggested the possibility that HMA may be the 3^rd variable that would explain the relationship between IV (EMA) and DV (EFL Total narrative). Our analysis indicated that this was the case for Arabic speakers as well: HMA_pseudo was a confounding variable that explained the increase in EFL oral language performance among Arabic speakers. This finding is important because it expands the significance of previous data regarding the role of morphological awareness as L1/L2 linguistic skills (Zhang, 2015), to include additional languages. Moreover, this particular finding provided strong support for previous research findings that MA is essential for vocabulary development (Zhang, 2015), and suggested that it can play a strong role in acquisitional processes.

The findings for both language groups are also in line with Hirosh and Degani (2018) proposition that multilinguals may activate both direct routes and indirect routes to additional language learning, where direct routes include transfer of linguistic skills and knowledge, while indirect routes represent cognitive factors such as metalinguistic awareness and working memory, which was the case among our participants. The present findings gain support from recent research which indicates the importance of crosslinguistic transfer of skills, particularly MA, from L1 to L2 (forward transfer) (Jarvis and Pavlenko, 2008; Kim and Piper, 2019). Moreover, we found that a specific MA skill in one language may enhance narrative production among two groups of speakers of different Semitic languages. This skill was HMA_pseudo, which clearly influenced the EFL Total narrative scores for both linguistic groups, despite typological distance of the languages. It is possible then to postulate that the assumptions of Linguistic Proximity Model (Westergaard et al., 2017) that crosslinguistic transfer between typologically distant languages will be determined by the areas of cross language overlap may not necessarily account for the application of all cognitive and linguistic resources among multilingual speakers, since our findings indicated transfer among typologically different languages. Although we did not observe bidirectional transfer of specific skills, the finding that Hebrew MA support EFL narratives not only for native Hebrew speakers but also for Arabic speakers, increases our understanding of the role of CLIs in oral language production.