Purple | My Site 6

Towards an Optimal Bilingual Language Instruction Model in Preschool Years (BSF)

Sharon Armon-Lotem, Laida Restrepo, & Carmit Altman

1. A brief description of the subject and the scientific and technological background

Bilingualism is the daily experience of more than half of the children worldwide. Yet, most educational services in Israel and the US are provided in the majority societal language to children who speak another language at home. Bilingual children need to master the societal language, their second language (L2) in order to succeed academically, but need to maintain the home language, their first language (L1), for functional communication and transmission of cultural values, in order to enjoy the benefits of bilingualism. While bilingual education might be a potential solution, it is hardly available (Goldbenberg & Wagner, 2015), especially in preschool years. This challenge has been exacerbated by the recent pandemia as children have limited access to schooling whether they are monolingual or bilingual. The poor valorization of the home language in healthier times and the difficulty to support the societal language in these unstable times that will generate innovative means to enhance the societal and the home language.

To meet this challenge, the proposed study examines the impact of different instructional modules of narrative intervention on cognitive-linguistic skills in their two languages in English-Hebrew bilingual preschool children, ages 4;6-5;6.. The novelty of the proposed study is in the investigation of the differential effect of two instructional modules -- block vs. mixed -- within and across languages on micro- and macro-communication skills. Micro- and macro-communication skills are essential components of a narrative that we use for telling stories. We further examine the effects of language order in block designs. Intervening and testing in two different language conditions (L1 first followed L2 vs. L2 first followed by L1), we will examine transfer of cognitive-linguistic skills trained in one language to the other language across four micro-and macro-communication skills. Impact of both questions will be tested for immediate post, learning curve, and retention outcomes across the two languages.

Micro- and macro-communication skills are cognitive-linguistic skills that affect comprehension and production of narratives at the word-, sentence-, and discourse-levels. Micro-communication skills refer to vocabulary and syntactic knowledge, while macro-communication skills refer to the global narrative structure. These skills emerge alongside pre-literacy skills in the early preschool years and are critical for academic achievement. Bilingual narrative intervention can be provided online and offline and supports both languages; however, but research on the optimal instructional model is limited, especially during the critical preschool years. We aim to determine which instructional model are more efficient in enhancing learning of the different aspects of narrative within and across languages.

Different types of instructional model were tested in monolingual children and on a limited basis in bilingual children. Research indicates that teaching one skill at a time, in a block design, leads to better immediate outcome results, but poorer retention effects (Hodges et al., 2014; Soderstrom, et al., 2015). On the other hand, mixed learning designs, where more than one skill is taught or where there is contextual interference, may lead to better learning retention (Hodges et al., 2014). For bilingual children, these instructional modules have only been tested for different skills within the same languages, or for the same skill across both languages, but not for multiple skills within and across both languages. These instructional modules can be best tested in bilingual children because the language not receiving intervention acts as a control for the language receiving intervention. Only studies comparing between the different models, testing for immediate gains, learning curves, and retention effects, could provide an optimal model for learning the different micro-and macro-communications skills in the two languages. The proposed study rises to this challenge aiming to yield novel evidence from bilingual preschool children at the critical age that prepares them for academic success.

1.1 Instructional modules in intervention for bilingual children

Cognitive principles focused on communication indicate that the variability in the input impacts learning. Specifically, high variability in communication input leads to better learning and retention than low variability (Soderstrom, et al., 2015). This variability is also known as contextual interference (Battig, 1979; Lee & Magill, 1983; Lee et al., 1992). Studies have shown that the contextual-interference effect is low in a block design when a single skill is practiced until acquired, before a new skill is introduced in a separate block (Hodges et al., 2014). The cognitive learning principle of variability has been supported by findings from vocabulary learning (Perry et al., 2010). However, the effect of contextual interference has not been tested for other micro-communication skills, such as sentence complexity, and on macro-communication skills, such as narrative structure.

Similarly, research on foreign language learning (Schneider, et al., 2002) indicates that certain types of difficulties can facilitate learning in a foreign-language. These “desired difficulties'' (Bjork, 1994) include mixed sessions or providing contextual interference. Bjork et al. (2015) concluded that introducing bilingual learning contexts may be an option for “desired difficulties”. Such challenges to language processing may yield cognitive gains at later stages. Therefore, , learning a skill simultaneously in two languages could be considered a higher variability input than learning the same skill in only one language. The proposed project attempts to test this by examining the extent to which low variability (a single language block module), and a high variability (mixed languages module) affect. In addition, within the block module, we examine whether a block L1 first followed by a block of L2 leads to better learning than block L2 first followed by a block of L1.

Despite a relatively large number of intervention studies focused on monolingual children, only a moderate number are available for bilingual children with typical language development (e.g. Cruz de Quirós et al., 2012). Most of these studies conducted vocabulary interventions in bilingual children in the two languages (Armon-Lotem et al., 2020; Ebert et al., 2014; Restrepo et al., 2013) and reported the efficacy of bilingual intervention on vocabulary in a block design protocol. Research on vocabulary in a mixed design protocol indicates that alternating L1 and L2 facilitates learning in both; whereas learning in a block condition of L2 only, facilitates L2 only (Restrepo et al., 2013). In the even fewer bilingual narrative intervention studies, such as Puente de Cuentos (Spencer, et al, 2019; 2020), the intervention aims to support both languages using a mixed design condition to support micro- and macro-communication skills. Such a design showed cognitive-linguistic gains in both languages, but made it difficult to estimate the transfer of skills from one language to the other language.

In a block design intervention, Armon-Lotem et al (2020) examined whether a bilingual narrative intervention on micro-communication using a block design could yield crosslinguistic gains in both languages regardless of the language of intervention. They found that children made significant progress in vocabulary in the language of intervention. In addition, they found that children showed transfer of skills across languages from L1 to L2, but not vice versa. Altman et al (2019), using a similar block design, examine transfer of macro-communication skills. They found that story grammar transferred from L1 to L2 after six sessions of intervention in the L1. Progress in the L1 was not observed in the immediate post. Gains in L1 were only observed after the retention period reflecting the complexity of these macro-communication skills. Figure 1 shows the differential progress patterns in a specific skill in the two languages.

Figure 1. Differential progress patterns for gains in micro-& macro-communication skills following the pilot intervention at four data elicitation points

Since most studies of bilingual intervention use only a single instructional module or intervene in a single language or a single direction, they are unable to compare how the differences between instructional conditions affect the different skills in the two languages. Crucially, while mixed design might show better retention and better learning curves, block design might show better immediate-post-test results and transfer of skills across languages (Petersen et al., 2016). Studies that compare between the different modules, as the proposed study aims to do, can contribute to our understanding of the mechanisms that enhance learning in bilingual settings, and provide novel evidence for an optimal model of intervention in the different skills in the two languages.

1.2 Rationale for using narrative intervention

Narrative (story telling) skills are authentic, socially relevant discourse that reflect general language learning ability (Bishop & Adams, 1992; Boudreau, 2008; Hughes, et al., 1997; McCabe & Marshall, 2006). Narrative production relies on micro-communication skills (including vocabulary, mental state terms and complex syntax) and macro-communication skills (story structure).To produce a narrative, children need to plan and produce contextualized and cohesive discourse (Petersen et al., 1999) and organize ideas into a unified-coherent whole (Stein & Glenn, 1979). The formation of a coherent narrative enables the child to share and relate inner thoughts and feelings beyond the facts and events. This is especially important in periods of instability (Restubog et al., 2020), where the child is limited in social interaction . Narrative skills are also one of the best predictors of later academic outcomes (Paul & Smith, 1993; Bishop & Edmundson, 1987; Fazio et al., 1996; Wellmann et al., 2011).

Narrative development in preschool years is characterized by rapid changes in the micro- and macro-communication skills. Micro-communication skills include vocabulary knowledge, and grammatical knowledge that can be measured by sentence complexity and the use of sentence connectors within the story. Preschoolers mainly use basic connectors, such as 'and' 'so', while older children use more sophisticated connectors that are associated with contrast, such as ‘but’ and causality ‘because’, both within and between clauses (Berman & Slobin, 1994). With age, children produce more lexically rich and complex narratives (Justice et al., 2006; Pearson, 2002). Protocols and guidelines of intervention improving oral language in bilingual children are still limited and do not clarify the best instructional module for word learning.

Mental State Terms (MST) are central to children’s narratives connecting micro- and macro-communication skills. They express a storyteller’s understanding of why characters acted, how they felt, and what they thought (Burns et al., 2012; Ukrainetz et al., 2005). Evaluative devices (Labov & Waletzky, 1997) such as MSTs, reflect a child’s interpretation of the events and indicate the reflection on characters’ thoughts and motives.. The ability to utilize MSTs can further help the child express her social-emotional state. Mental state terms include motivation verbs (e.g. ‘want’), perception verbs (e.g. ‘see’), adverbs (e.g. ‘suddenly’) and cognitive states (e.g. think) (Füsté-Herrmann et al., 2006; Greenhalgh & Strong, 2001; Johnston et al., 2001). Representation of mental states requires a certain level of linguistic development because inferences about characters’ intentions (see macro-communication skills) are linked to vocabulary in micro-communication skills. Ukrainetz et al. (2015), for example, showed a developmental increase in the acquisition “evaluations” mental state terms (as part of evaluations) from age 5 age 12.

Macro-communication skills refer to story grammar structure, the story parts and their organization of that provides the key features for understanding and following a story. Story grammar analysis is the realization that narratives are made up of one or more episodes with a unique internal structure (Mandler & Johnson, 1977; Stein & Glenn, 1979; Trabasso & Nickels, 1992). An episode is composed of a Goal, Attempt, and Outcome. The explicit or implicit goal reflects a need or desire of a character; the attempt is his/her action to achieve the goal, and the outcome is the effect of the action. In all narratives, there is a setting, which gives the context of the story. Preschool children are expected to generate and retell stories. Children tell stories with increasingly more complex episodes, with more complex causal and temporal relations that reflect the child’s cognitive development (Berman & Slobin, 1994; Pearson, 2002). They show increasing understanding of a more complex narrative schema, along with perspective taking and theory of mind (Trabasso & Nickels, 1992). At age three, children’s narratives are short and include few story grammar elements; instead they provide lists of characters, events, and objects. Their stories are missing the causal and temporal relations of slightly older children (Berman & Slobin, 1994; Peterson & McCabe, 1983), and often lack the description of settings and goals, and mental states of characters (e.g., feelings displayed) (Trabasso & Nickels, 1992; Trabasso & Rodkin, 1994). Between the ages of four to nine, story grammar increases, and children start including settings and full episodes with goal-attempt-outcome sequences (Trabasso & Nickels, 1992; Trabasso & Rodkin, 1994).

Narrative interventions aim to help children recognize, internalize and produce components of micro- and macro-communication skills. Causality has been targeted in intervention in the form of causal subordinate clauses that used the conjunction “because” (Petersen et al., 2016). Narrative intervention potentially targets a child's thought processes, emerging skills of storytelling, and learning potential (Peña et al., 2006). Narrative intervention is often incorporated into treatment plans in children with disabilities because of its functional nature, its adaptability to different media- offline or online- and its relationship to academic demands (Davies, Shanks, & Davies, 2004; Gillam, McFadden, & van Kleeck, 1995; Hayward & Schneider, 2000; McCartney et al., 2004; McFadden, 1998; Peterson, Jesso, & McCabe, 1999; Schoenbrodt, Kerins, & Gesell, 2003; Stiegler & Hoffman, 2001; Ukrainetz, 1998). Recently studies began to address oral language skills, specifically, micro- and macro-communication skills in bilingual narratives (Altman et al., 2016; Boerma et al., 2016; Fichman et al, 2017; Fichman et al., 2020). Yet very few narrative intervention studies focus on bilingual children. The proposed study will measure the micro and micro-communication gains following intervention among bilingual children with typical development, addressing the critical need to better serve bilingual children and capitalize on their strengths.

1.3 Transfer of macro- and micro-communication skills across languages

Transfer of skills across languages is an important consideration when examining learning outcomes in bilingual children because some instructional modules may be more efficient in improving both languages than others, due to language facilitation effects (See Restrepo et al, 2013). Some studies show facilitating effects for a single skill. For example, Armon-Lotem et al (2020) found that intervention on vocabulary in L1 is carried over to L2. In addition, L2 narrative skills, which are strongly supported during preschool, may transfer from L2 to L1 (Petersen et al, 2016).

Since typological differences between languages may affect transfer (Squires et al., 2014; Uccelli & Páez, 2007), the proposed research compares the directionality in two typologically different languages to explore transfer of features. Ebert et al. (2014) implemented a vocabulary intervention, 80% in L1/Spanish and 20% in L2/English, pretesting and posttesting receptive and expressive vocabulary in both languages. Since most of the intervention was in the L1, they found no transfer effect from L2, although found transfer effects from L1 to L2, since intervention was conducted primarily in L1. On the other hand, Restrepo et al, (2013) examined the use of a block of L2 or an alternating language mixed design on vocabulary outcomes. They found that the L2-only intervention did not lead to improvements in the L1, whereas the mixed alternating language condition led to improvements in both languages.

Transfer of narrative skills was investigated experimentally shows equivocal results. Petersen et al. (2016) examined transfer of narrative skills from L2/English to L1/Spanish after a L2 narrative intervention in Spanish-English bilingual children. Given that intervention was provided in L2/English, only unidirectional transfer from L2/English to L1/Spanish could be evaluated. The authors found transfer of micro- and macro-communication skills (subordinate clauses and story grammar elements, respectively) from L2 to L1.In contrast, Ebert et al (2014) found transfer only from L1 to L2.

Intervention studies that have focused primarily on vocabulary and not on other narrative skills, indicate that bilingual intervention has positive effects on both languages, whereas intervention only in the L2 impacts L2 only and may have detrimental effects on the L1 (Méndez et al., 2015; Restrepo et al., 2013; Rolstad et al., 2005). For one of the modules used in the proposed study, using an L1 block first followed by an L2 block, Armon-Lotem et al. (2020) found that all children gained in the L2 following intervention in the L1, but only children who were relatively stronger in the L1 gained in the L1 following the L2 intervention. This result is one of the motivations for testing both language orders in the block module of the proposed study. The impact of intervention in each language separately and the cumulative effect of intervention in two languages will be addressed to better understand the nature of transfer of micro- and macro-communication skills across languages.

In summary, research indicates that variability improves learning outcomes. We consider that using two languages for instruction is a type of variability in the language input provided for learning (Restrepo et al, 2013). In addition, research indicates that an alternating language design (mixed module) will lead to better outcomes than a block of intervention for each language (Armon-Lotem, et al 2020). Block design studies results indicate that language transfers from L1 to L2 in terms of vocabulary when the L1 is used first (Armon-Lotem et al, 2020), but not necessarily when the L2 is introduced first or used (Restrepo et al, 2013). In contrast, transfer effects are found from L2 to L1 in macro-communication skills (Petersen et al, 2016).

2. Objectives and significance of the research

Using, the two instructional modules in three practice conditions -- (1) block design with six sessions of L1 followed by six sessions of SL, (2) block design with six sessions of L2 followed by six sessions of L1 and (3) mixed design alternating L1 and L2 pseudo-randomly for 12 sessions -- the objective of the proposed study are twofold:

Investigate the impact of different instructional modules on each micro-communication skill (a) within language, (b) across languages via transfer, and (c) with regards to immediate effect, learning curves, and retention
Investigate the impact of different instructional modules on macro-communication skills (a) within language, (b) across languages via transfer, and (c) with regards to immediate effect, learning curves, and retention

In all three practice conditions, bilingual intervention is expected to improve the micro-and macro-communication skills. The gains yielded may show differential patterns in the different conditions in the two languages. Only block design allows one to separate and relate to relative gains in each language throughout the process. With mixed design, what is tested is the children’s progress in each language. Transfer gains cannot be measured in the mixed design module since languages are pseudo-randomly alternated and gains cannot be assumed to be linked to any one specific language. Our predictions relate to these three practice conditions.

For micro-communication skills, it is predicted that:

(a) Within language, intervention in a specific language will yield improvement in the language of intervention as shown in previous studies (Armon-Lotem et al., 2020; Restrepo, 2013; Petersen et al. 2016; Restrepo et al. 2010)

(b) Across languages effects examine the contribution of each practice condition to the transfer across languages. This can only be tested for the block design module. The Revised Hierarchical Model (Kroll et al., 2010), suggests that links are stronger from the stronger language to the weaker language and not vice versa. This was observed, at the word level (Armon-Lotem et al., 2020; Restrepo, 2013) and was explained by the relative high proficiency of the children in the L1 prior to the L2 intervention. At the word level, concepts are shared across languages and therefore may be more susceptible to transfer. If mental state terms (MST) present similar trajectories as other lexical items, it is expected that they follow the predictions above. Highly frequent MSTs are expected to show more transfer in both directions, since they are shared by both languages (e.g., ‘want’), while infrequent MSTs expressing complex emotions (e.g. ‘surprised’) are not expected to show transfer. Since MSTs are sensitive to culture (Grossmann, et al.,2012; Pavlenko, 2002), they may yield different patterns of transfer than standard lexical items. At the sentence level, syntactic structures are more language specific. Research shows that in bilinguals, positive transfer is observed for structures which are shared across languages and negative transfer is observed where language structure is different. Petersen et al. (2016) found transfer of causal subordination that is shared between English and Spanish after intervention in L2. Thus, transfer across languages is expected for syntactic structures which are shared. The mixed design protocol does not enable testing transfer of skills across language since languages are randomly alternated.

(c) Immediate effect, learning curves and retention. The block design is expected to show better immediate results as well as learning curves (Armon-Lotem et al., 2020; Lugo- Neris, 2016; Petersen et al., 2016; Restrepo, 2013). As for retention gains, there are very few studies that observed children’s performance over time to examine whether gains have been preserved. For example, Armon-Lotem et al. (2020) found that lexical gains were retained six weeks after intervention with no further improvement. Similarly, it was found that L2 lexical gains were retained four and eight months following vocabulary intervention (Restrepo, 2013) and in L1 four months following syntactic intervention (Restrepo et al., 2010). This may be due to the use of a high variability design working on different micro-communication skills within each language block.

(2) For macro-communication skills, it is predicted that:

(a) Within language, intervention in a specific language will yield improvement in that language as shown in previous studies for the different modules and different languages (Cruz de Quirós et al., 2012; Lugo-Neris et al., 2015; Petersen et al., 2016; Lugo-Neris, 2016). For a mixed design, as mentioned above we expect progress in both languages yet cannot diffuse the differential language effects.

(b) Across languages, transfer of macro-communication skills is expected to show differential patterns in comparison to micro-communication skills (e.g., Altman et al., 2019; Petersen et al, 2016). Macro-communication skills, largely shared across languages (Boerma, Leseman, Timmermeister, Wijnen, & Blom, 2016) may be more prone to transfer than micro-communication skills, which is more language-specific. Macro-communication skills are based on more global cognitive linguistic skills that are needed in order to produce a coherent narrative in any language. The mixed design protocol does not enable testing transfer of skills across languages. Yet, following narrative bilingual mixed intervention, gains are expected in both languages (Spencer et al., 2018). We expect that the constant alternation of languages will highlight the similar macro-communication skills between the languages, therefore yielding gains in both languages.

(c) Immediate effect, learning curves and retention. Similar to what was found for micro-communication skills, we predict that the block design will show better immediate effects and learning curves (Lugo-Neris et al., 2016), while the mixed design is expected to show better retention effects. While immediate effect is documented in most studies, retention is often not documented (but see Altman et al., 2018; Spencer et al., 2018). Due to the disparity of studies, the proposed study aims to fill the gap by documenting immediate effect, learning curves, and retention in the three practice conditions proposed.

For both objectives, the relative gains in the two language orders -- L1 to L2 and L2 to L1 -- may be sensitive to language proficiency, which is still fluctuating among bilingual preschool children. Yet, this was only tested in a single study for children with language impairment (Lugo-Neris 2016). For bilingual preschool children with typical development, to the best of our knowledge, no study compared the two conditions of the block design and only a single study compared the mixed module with a single language block module. Using novel evidence following research objectives 1 and 2, we expect to better understand the mechanisms that enhance learning in bilingual settings, making it possible to identify the best instructional module for each of the skills.

3. Methodology and plan of operation.

The study proposes a pre- and post-narrative intervention design with three practice conditions in both languages for English-Hebrew preschool children with typical language development. The intervention procedure uses Story Champs (Peterson & Spencer 2012) and Puente de Cuentos structured activities (Spencer, Petersen, & Restrepo, 2018), which were found to have high efficacy for preschoolers with low language skills and for bilingual children (Petersen et al., 2014; Spencer & Slocum, 2010; Spencer et al., 2013, 2015; Weddle et al., 2016). The Puente de Cuentos structure activities were selected for the study because each intervention session focuses on micro-and macro-communication skills, obtaining high variability which is expected to maximize gains in bilingual narrative intervention. The English materials were adapted to Israel’s cultural context and Hebrew materials have already been developed within an ISF grant (no. 1716/19) to Altman, on which Restrepo is a consultant. Testing will take place in Israel to control for the possible effects of language prestige on the children’s learning motivation, given that in Israel the population values both languages the L1, English, has international prestige and the L2, Hebrew, has ideological and practical value. Hebrew data will be coded in Israel and English data will be coded in the US. Data analyses will be carried out in both locations in tandem using the expertise of the US partner in analyzing such data.

Participants characteristic and criteria. 120 bilingual English-Hebrew preschool children, ages 4;6-5;6, all presenting with typical language, cognitive and neurological functions will participate in the study. Children will be functional bilinguals (Kohnert, 2007) meeting the following criteria: a) exposure to L1/English from birth, where the L1 was spoken at home as a mother tongue by at least one parent; b) exposure to L2/Hebrew for at least 24 months; c) ability to hold conversations in both languages; and d) attendance at preschool programs where L2/Hebrew is the language of instruction and social interaction.

Following informed parent consent and children’s oral assent, a parent questionnaire will be administered to elicit information on developmental milestones, exposure to Hebrew and English, language preferences, and concern regarding language development (Abutbul-Oz, Armon-Lotem, & Walters, 2012). Nonverbal intelligence will be tested using Raven (1998). Bilingual language proficiency measures with local standards will be administered in both languages before intervention (see Figure 2) to address the heterogeneity of the groups. For L1/English we shall use CELF-preschool 4 (Wiig, Semel, & Secord, 2004) with subtests for expressive language, receptive language, language content, and language structure. For L2/Hebrew proficiency, Goralnik Screening Test (Goralnik, 1995) with subtests for noun/verb naming, pronunciation, sentence comprehension and imitation, oral expression, and storytelling will be used. These two measures have local bilingual standards obtained in recent years (Altman et al., 2016; Armon-Lotem, Rose & Altman 2020). Only children who score on at least one Hebrew or English tests within age-appropriate standards and whose parents do not express concerns about their language development will be included.

. Bilingual narrative intervention will be conducted in three practice conditions: (1) Block L1-L2; (2) Block L2-L1 (3) Pseudo-random mixed design with both languages. To control for the intervention effects an attentional group will be added. A priori analysis of sample size given α = .05 and a power of .80 with a medium effect size (F= 25; η2p = .06), indicated that a sample size greater than 26 participants per group is required. Power analyses were made using G*Power (ver. 3.1.9.4). Each condition will include 30 participants. In block condition 1, the first block consists of six sessions in L1 (English) and second block six sessions in L2 (Hebrew) for a total of 12 sessions. In block condition 2, the first block consists of six sessions in L2 (Hebrew) and the second block six sessions in L1(English) for a total of 12 sessions. In the pseudo randomized mixed condition, children will receive 12 sessions of intervention, alternating between the L1(English) and the L2 (Hebrew). Pseudo randomization will take place holding constant the number of L2 and L1 sessions, with no more than two consecutive sessions in the same language. The control group will participate in an unrelated activity for 12 sessions. Gains will be measured at four data points (see progress monitoring and testing).

Figure 2

Intervention procedures. Intervention will be carried out in small groups of 3-4 children twice a week by research assistants with a degree in Education or Speech and Hearing. We will examine procedural reliability on 20% of the sessions. All interventionists will be trained to 90% procedural reliability. After 6 sessions, we will examine reliability in intervention and assessments to check for drift in reliability.Following Spencer et al. (2017), the procedures for each session include five steps: a) story modeling and introduction of icons for six targeted elements (character, problem, MST, goal, attempt and outcome); b) pairing icons with gestures; c) group (choral) retelling; d) explicit focus on target features; e) individual retelling. During each session, after the story is modelled, children repeat parts of the story with the experimenter four times, and each child has 4-6 opportunities to respond to different segments of the story that focus on the targeted features. In an effort to reduce the drill-like appearance of the activity and to add to children's engagement in the task, pictures, icons and gestures are accompanied by verbal encouragement throughout the intervention. While generating the story, special attention is dedicated to targeted vocabulary words and targeted syntactic structures. Such intense adult-child and peer interaction intends to raise awareness and elicit production of the target features crucial for the development of narrative skills. The procedure takes 20-30 minutes and is administered by a native speaker of the language of the session to guide and instruct the children. All sessions will be audio recorded.

The narrative scripts, picture stimuli, and icons used in intervention are based on Spencer et al., (2017), and were adapted culturally (e.g., paper boats do not sail in ponds in Israel) for English and developed for Hebrew within an ISF funded project to PI2 (Altman) in collaboration with the US partner (PI3 - Restrepo). Materials were constructed by a team of special education and linguistics graduate students (native-speakers of/English/Hebrew) and submitted to a group of 10 teachers native in each language for judgment as to cultural and age appropriateness (Castilla-Earls, Peterson, Spencer, & Hammer, 2015). Narrative materials are comparable across the two languages. Targeted vocabulary is comparable in terms of parts of speech and syntactic positions, yet different words are used in each language to examine transfer effects. Identical numbers of vocabulary words, MSTs and complex clauses are used, as well as shared macro-communication measures.

Progress monitoring and testing: The impact of the bilingual intervention will be tested at four data points (as presented in Figure 2) in both languages: a) pre-intervention, b) mid-intervention after session 6, c) post-intervention after 12 sessions, and d) retention testing two months after the last intervention session. At each data point two narratives will be used for each language, to avoid story content effect, and the narrative with the higher score will be used for further analyses. Narratives for the four data points are constructed similarly to the stories in the intervention itself in order to examine potential gains for vocabulary, MSTs, syntax, and SG elements. Pre-intervention data points will provide a baseline of performance. The design also addresses the importance of retention of skills two months following intervention, addressing the concerns that it is not enough to show immediate gains and learning curves (Law et al., 2003; Petersen, 2011; Petersen, Gillam, Spencer, & Gillam, 2010),
. We will examine procedural reliability on 20% of the sessions. All interventionists will be trained to 90% procedural reliability. After 6 sessions, we will examine reliability in intervention and assessments to check for drift in reliability.
Data Analyses. All transcription and coding will be carried out by research assistants who will be blind to the data collection point or condition (pre-, post- or follow-up intervention). Hebrew data will be coded in Israel and English data will be coded in the US following joint consultation. Analyses of transcribed data will begin at the level of the individual child for the three targeted micro-communication skills - vocabulary, MSTs and complex syntax, and story grammar elements (macro-communication). We will run within-subject and group comparisons for the three conditions at each data collection point and across language (transfer) comparisons for each skill. Control in the design (Law et al., 2003) includes comparing gains for: a) the three experimental conditions b) within the two languages of the child c) across the two languages of the child and d) control group with unrelated activity. All comparisons will monitor gains in each language separately. Immediate effects for within and across languages are measured by comparing pre-intervention to post-intervention. The learning curve is measured by examining the growth rates (change in scores) across the four data points. Retention is measured by examining the changes from post test point three (the last post intervention post test) to the fourth testing point (two months later).

The proposed cooperation cuts across all phases of the proposed research. The design has been done in tandem, strongly relying on the expertise of the US partner in designing mixed narrative intervention for bilingual children and moving beyond the expertise of Israeli PI1 (Armon-Lotem) in typical bilingual acquisition and PI2 (Altman) in narrative development and the use of block design in bilingual narrative intervention. Testing will take place in Israel making use of the strong relations of the Israeli partners with the English-Hebrew speaking community using measures developed by the two partners. Training of RAs for test delivery will be supervised by both partners, and the delivery itself will be supervised by the Israeli partners with close online monitoring of the process through team meetings with the US partners, and if possible a visit of the US partner for kick-off and follow-up assessment period. Data transcription and coding will be done for Hebrew in Israel and for English in the US. Data analyses will be carried out in both locations in tandem using the expertise of the US partner in analyzing such data. Regular communication will take place using zoom for monthly team meetings with the RAs and bi-weekly updating meetings of the PIs, when necessary. The Israeli PIs expect to travel to the US to brainstorm on data analysis and paper writing.

4. Risk analysis and alternative paths that will be followed if the suggested research plan fails (only in those fields in which it is relevant);

We have identified several potential risks that required our attention. i. No differences will be observed between the different instructional modules or between the different cognitive linguistic skills. While such results will undermine the ability to identify the optimal instructional module, our pilot results (see below) are encouraging as far as the different skills are concerned. Finding no difference between the instructional modules will have a major scientific impact. ii. Recruiting the children might take a long time. Our contacts in over 60 preschools with all approval needed in advance is expected to shorten the process. We already have a waiting list of children who would like to join. iii. Length of the intervention with 22 sessions per child, is prone to attrition difficulties. We have minimized attrition in our pilot study by ensuring that children are engaged in the sessions and like meeting with us. Parents have our personal contact information in case of specific difficulty. In our second pilot with over 100 children, only two children left due to their moving away due to the pandemic, yet they offered to continue via zoom. iv. Children at risk for language impairment may be identified and referred for services. To allow such cases we opt for 30 children per group while power analysis using G-Power suggests that 26 are enough. In such a case, we continue with the intervention protocol, but exclude the children from the data analysis. v. Consistency across experimenters, transcriber and coders– we train our experimenters first together to 90% reliability in a group, so they can learn from each other, including simulations in pairs so they can get feedback. For the intervention ,we go to the preschools and observe the assistants. Only when we see they are familiar with the experimental protocol and ethical guidelines we allow them to collect data. vi. COVID-19 pandemia - we are well aware that COVID-19 may be here for a while and therefore, we have also piloted the materials online. We found that 90% of children said that working on line was better or similar to working with us face to face. 10% said it was more difficult working online.

We have two sets of pilot data that were collected prior and within the ISF grant to Altman. (1) Seventeen bilingual children aged 5;6-6;6 participated in the pilot bilingual narrative intervention study testing the impact of condition 1 ( L1 block first followed by a block of L2). Results show that micro-communication skills improved differently at the word and sentence levels. At the word level, intervention in L1 had an impact on both L1 and L2 showing clear transfer of skills, while L2 intervention yielded gains for L2 only. At the sentence level, gradual, yet insignificant gains were observed in both languages as children began to use more sentence connectors, such as because, or after. Macro-communication skills improved significantly in L2 following L1 intervention, but L1 presented gains only when retention was measured. (2) For Altman’s ISF funded project, data were collected from 95 children ages 5;6-6:6 (a year older than the participants of the proposed project), using condition 1 as in the pilot with the exact protocol proposed for the proposed project (22 sessions, 4 data points). This experience already reduces the risks of the proposed project. Preliminary findings show the bilinguals who were stronger in L1 benefited more from the L1 block, showing more transfer of micro-communication skills (including internal state terms) across languages than those stronger in L2. The latter had relatively fewer significant cross-language associations. These findings strengthen the need of testing the effect of the two language conditions for the block module, in order to understand whether different conditions are better for different children as a reflection of their proficiency in the two languages. Data collection for this project continued during the pandemic, moving online when necessary, which the children and families appreciated. This variation in administration of the intervention, will be addressed when analyzing the findings. This experience reduces , however, the concerns raised by the sixth risk mentioned above.

6. An account of available U.S. and Israeli resources.

The three PIs have worked closely together for the past three years. All participated in developing and writing this research proposal. We have the personnel, space, equipment, and expertise to perform the proposed study. The Israeli partners have lab space with laptops for data collectors and the relevant infrastructure for storing and analyzing the data. The study will be conducted in the preschools in Israel where contacts have already been established. Prof. Sharon Armon-Lotem and Dr Carmit Altman will supervise the administration of the studies and manage data collection. Since the team in the Israeli lab works on a regular basis ( three prior ISFs, one MoE grant, BMBF, and GIF) with bilingual children, personnel is highly-trained. As demonstrated in our pilot data (see also Armon-Lotem et al., 2020) in the past three years, this team has tested a large sample of bilingual children on narrative tasks. Prof. M. Adelaida Restrepo, who was the first to carry out the bilingual intervention in the US, will constantly advise professionally. Data transcription in English and statistical analysis will be carried out under the supervision of Prof. Restrepo. She has 450 square ft lab with 4 workstations and software installed for transcription, a small conference table and headsets for transcription. All transcription data will be stored in a password secured server where only the investigators have access. Finally and most importantly, not only we established access and received all ethics approvals to test bilingual children in the 60 preschools we have contacted, we have started testing older bilingual children and aim to reach a total of 60 children prior to the beginning of the proposed study. This initial recruitment ensures a smooth start of our research project and guarantees that the project will be conducted in a timely manner.