The linguistic expression of causal relations in picture-based narratives: A comparative study of bilingual and monolingual children with TLD and DLD

Abstract

The present study examines the linguistic expression of causal relations between the motion events within the main episode in a picture-based narrative. One hundred and fifty children aged 5–7 were asked to narrate a story based on a series of pictures: 45 Hebrew monolinguals (19 with Developmental Language Disorders [DLD]), 57 English–Hebrew bilinguals (20 with DLD) and 48 Russian–Hebrew bilinguals (21 with DLD). The narratives told by bilingual children with Typical Language Development (TLD) were as complex as those of their monolingual peers in occurrence of causal relations to establish a goal-oriented episode, and in the use of language forms as cohesive devices. By contrast, bilingual and monolingual children with DLD showed lower performance on expression of causal relations, particularly those involving more complex scenes that demand higher levels of linguistic complexity and content elaboration. The form–function analyses enabled an exploration of cognitive and language abilities in interaction in the context of narrative discourse production. The study reinforced the differences between children with TLD and children with DLD, showing that typical language development rather than the proficiency in a particular language is necessary for generating causal relations and the particular linguistic forms that together yield a coherent narrative.

Keywords

Bilingualism causal relations cohesion DLD motion events narrative discourse

Introduction

Narrative practices require the use of language to transform experience into a unified piece of discourse that is ruled by organizational principles underlying the genre (Berman & Slobin, 1994). Children telling stories need to be equipped with top-down knowledge for organizing and connecting the story events and with bottom-up knowledge of language to select the appropriate lexical and morpho-syntactic forms in the process of storytelling (Berman, 1988, 2008; Karmiloff-Smith, 1979). The present study integrates these two levels of knowledge into one analysis, by looking at an array of linguistic forms at the clausal and inter-clausal levels, and their relation to creating four types of causal links that lead to a coherent episode. The study compares the narratives produced by monolingual and bilingual children with Typical Language Development (TLD) vs. Developmental Language Disorder (DLD), adding another perspective to the growing body of literature that analyzes the macro- and micro-levels in these populations.¹ Developmental Language Disorder (DLD) is used in this article for Specific Language Impairment (SLI), following Bishop, Snowling, Thompson, Greenhalgh, and the CATALISE-2 Consortium (2017).

The sections below report on relevant studies about narrative development and present the conceptual basis of the present research stemming from well-established psycholinguistic perspectives.

Narrative abilities in monolingual and bilingual children

The production of coherent narratives demands top-down, conceptual knowledge to assemble narrative events into a causally related, goal-oriented episode, basically composed of a goal, an attempt to reach the goal and a final outcome (Stein & Glenn, 1979; Trabasso & Van Den Broek, 1985). Numerous studies that focused on episodic structuring showed a progression after age 5 in the number of components mentioned by children and their hierarchical organization around a goal plan, which becomes well established around age 9 (Aksu-Koç & Tekdemir, 2004; Berman, 1988; Berman & Slobin, 1994; Mandler, 1978; Peterson & McCabe, 1983; Stein & Glenn, 1979, among others). Fewer studies reported on the ability of children to establish causal relations between the events, an ability that increases from age 5 through 9 and further develops in more subtle ways until adulthood (Brown, Lile, & Burns, 2011; Sah, 2015; Trabasso & Nickels, 1992). These studies have paved a vast area of research in the development of narrative macro-structure, as top-down cognitive-driven schemata led by universal, language-independent knowledge of representing complex sequences of events in stories (Hickmann, 2004; Peterson & McCabe, 1991).

Psycholinguistic studies of narrative development in monolinguals (Berman, 1996, 2008; Berman & Katzenberger, 2004; Berman & Slobin, 1994; Hickmann, 2003, 2004; Kupersmitt, 2016, among others) report that these top-down representations are activated in the process of selecting lexical and grammatical forms to serve as cohesive ties (Hickmann, 2003, 2004). In this sense, linguistic forms (e.g., grammatical morphemes and constructions, content and function words, and syntactic processes) gradually mark and organize relations within the discourse (e.g., anaphoric, temporal or causal relations) as structuring abilities develop. For instance, clauses are connected by means of deictic markers (e.g., here, now) at preschool, through use of chronological markers (e.g., and then) at early school-age, to the use of coordination and subordination later on for expressing causal and temporal relations (Berman & Nir-Sagiv, 2009; Berman & Slobin, 1994; Jisa, 1984; Peterson & McCabe, 1991). Referential strategies develop from deictic to anaphoric (Hickmann, 2003; Karmiloff-Smith, 1979, 1981), thus creating tighter links between the entities in the discourse; and temporal forms such as past or present tenses, simple or progressive aspect, serve to indicate chronological sequences at the younger ages and develop later to express simultaneity and foreground–background distinctions throughout the story (Aksu-Koç & von Stutterheim, 1994; Berman, 2014; Kupersmitt, 2015). Lexical and morpho-syntactic constructions that express motion events become more specific with age and provide more information about location, change of location, path and manner of motion in the narratives (Hickmann, 2003, 2006; Slobin, 1996). This last domain is particularly relevant for the present study, and will be further specified in the methods section.

Studies performed with simultaneous or early sequential bilinguals investigated the relationship between knowledge of language forms and narrative organization abilities, focusing on connectivity (e.g., Akinci, Jisa, & Kern, 2001; Kupersmitt, 2004; Minami, 2008), the expression of evaluative content such as false belief and characters’ thoughts (e.g., Pearson, 2001), or the management of referential links (e.g., Chen & Pan, 2009; Serratrice, 2007). These studies found a similar development of macro-structures in monolinguals and bilinguals, as guided by universal, language-independent strategies. They also found a similar development of form–function relations when both languages of the bilingual had available means to express a certain function (e.g., referential links), but less so when they differed in a certain domain (e.g., narrative temporality). Nonetheless, bilinguals often produced less sophisticated morpho-syntactic constructions and used fewer language-specific resources (e.g., aspectual markers, temporal connectors) to express various narrative functions, particularly at school-age when bilingual children may be less exposed to reading and writing practices in the school language (SL) prior to school enrolment (Akinci et al., 2001).

Fewer studies have investigated the narrative productions in sequential bilinguals who are learning the SL after age 3. In a study with Finnish-speaking preschool and school-age children learning Swedish, Viberg (1994, 2001) showed a language difference between bilinguals and monolinguals in Swedish when using connectors and adverbs for combining clauses, verbs to express major events in the plot, and tense forms to express temporal relations. Studying younger Spanish–English bilinguals, Montanari (2004) argued that impoverished linguistic resources in the SL may have a detrimental effect on the ability to produce a well-organized account of the events and on the use of more complex forms of cohesion in the domains of reference and temporality, but less so on the ability to express evaluative functions. More recently, Kupersmitt, Yifat, and Blum-Kulka (2014) showed that preschool learners of Hebrew as SL were largely delayed in measures of coherence that included episodic structuring and causal relations, particularly in scenes depicting complex motion events in picture-based stories. After four years of exposure to the SL, at the age of 8, measures of coherence were similar in monolinguals and bilinguals but cohesive means showed a different path of development in the two groups, at both preschool and school-age. The constraints imposed by language use in organizing the discourse resulted in poorer connectivity between the clauses, and in less accurate lexico-grammatical encoding of the events in the bilingual narratives.

Narrative abilities of children with DLD

During the last decade a number of studies have compared the narrative abilities of bilingual and monolingual children with TLD with those produced by their peers with DLD, in order to differentiate between language patterns that are typical of acquiring a second language and language disorders that may reflect a linguistic deficit (Bedore, Peña, Gillam, & Ho, 2010; Terrell & Terrell, 1983). Iluz-Cohen and Walters (2012) investigated the narrative production abilities of English–Hebrew bilingual preschool children with and without DLD telling familiar stories. Their results showed similar patterns of narrative structuring abilities (macro-structure level), with a disadvantage for DLD bilinguals in language-related (micro-structure) measures. By contrast, in a longitudinal study of Spanish–English bilinguals with and without DLD, Squires et al. (2014) reported that the TLD group outperformed children with DLD in macro- and micro-structure elements in both languages, at both preschool and first grade.

Various studies using LITMUS-MAIN (Multilingual Assessment Instrument for Narratives, Gagarina et al., 2012) confirmed previous results on similar macro-level abilities in monolingual and bilingual children regardless of their language abilities (Kunnari, Välimaa, & Laukkanen-Nevala, 2016) while micro-structure often distinguished bilingual children with TLD from those with DLD (Boerma, Leseman, Timmermeister, Wijnen, & Blom, 2016). To refine the analysis, Altman, Armon-Lotem, Fichman, and Walters (2016) added an exploration of mental state terms, claiming that cognitive mental terms (e.g., know, decide, think) distinguished between TLD and DLD children in their home language (HL) and school language (SL) narratives, a finding replicated in Tsimpli, Peristeri, and Andreou (2016) with bilingual speakers of Greek as the SL. A recent study with Russian–Hebrew bilinguals showed that a more detailed macro-structural analysis that includes story grammar elements and causal relations was effective in distinguishing bilinguals with TLD and DLD (Fichman, Altman, Voloskovich, Armon-Lotem, & Walters, 2017). These studies provide valuable information about the linguistic and cognitive-related abilities of children, but fail to examine the use of language forms for the purposes that are intrinsically constrained by the type of discourse that children are producing, i.e., ‘the purposes of constructing a text that is cohesive and coherent at all levels: within the clause, between adjacent clauses, and hierarchically relating larger text segments to one another’ (Berman & Slobin, 1994, p. 4).

The present study

This study proposes an analysis that ties the macro- and micro-levels of the narrative, rather than treating them as two separate domains. From this perspective, language forms are considered in the context of the functions they serve in the narrative (Berman & Slobin, 1994), as described in the categories of analysis section. This is critical when studying bilingual populations, and particularly bilinguals with DLD, whose language abilities might have different impact on linguistic forms and functions. Assuming that bilingual children perform differently from monolingual children in language-related tasks including lexical diversity or morpho-syntactic forms in narrative contexts (Altman et al., 2016), it is important to examine whether top-down knowledge of narrative organization boosts language usage at the clause and inter-clausal levels, and, alternatively, whether the limitations in language skills affect narrative coherence. In bilinguals with DLD, tapping the relation between micro- and macro-levels is particularly relevant for shedding further light on the debate regarding the underlying difficulties encountered by children with DLD (see, for instance, Bishop et al., 2017). Taking monolingual and bilingual children with TLD and DLD in narrative production, a continuum of language abilities can be examined alongside discourse organizational abilities needed for integrating events into a unified episode. Against this background, the present study has two objectives:

To examine the effect of language background (monolingual vs. bilingual) and language ability (TLD vs. DLD), as well as measures of language proficiency (e.g., scores of screening test of language abilities) and of bilingual experience (length of bilingual exposure) on the occurrence of causal relations between the episodic components, at the top-down level of narrative construction.

To examine the effect of language background (monolingual vs. bilingual) and language ability (TLD vs. DLD) on the array of linguistic forms used at the clausal and inter-clausal levels, from a bottom-up perspective, and their relation to expressing causal relations at the top-down level.

It is expected that language ability, measures of language proficiency and bilingual experience will impact the occurrence of causal relations between the episodic components more than language background at the group level (Fichman et al., 2017; Kupersmitt et al., 2014). For the array of linguistic forms, monolinguals with TLD are predicted to perform better than the rest of the groups at the clausal level, particularly in the lexical domain. A difference is also expected between TLD and DLD in use of syntactic and referential strategies at the inter-clausal level (Altman et al., 2016; Boerma et al., 2016; Iluz-Cohen & Walters, 2012).

Finally, it is expected to find a relation between the use of particular language forms and the production of inferable causal relations (Berman & Slobin, 1994; Hickmann, 2004). Since the study investigates preschool children, it is tentative to suggest that emerging complex forms at the inter-clausal level will better reflect this relation than intra-clausal lexical items which in turn are more susceptible to the impact of bilingualism.

Method

Participants

The data presented in the present study come from a subset of 150 children out of 600 bilingual children, half with HL-English and the other half with HL-Russian and a similar number of monolingual children who were tested for studies conducted at the lab of the second author. Children were recruited from Hebrew-only mainstream preschools as well as language preschools for children with language difficulties in which the majority of the children came from Hebrew-speaking homes. Both communities value their HL and support it at home and in afternoon activities, and most homes maintain a HL-only policy or a bilingual policy, with both parents being speakers of the HL (Altman, Burstein-Feldman, Yitzhaki, Armon-Lotem, & Walters, 2014; Armon-Lotem, Altman, Joffe, Oz-Abutbul, & Walters, 2015). Data collection took place over a period of 10 years (2005–2015) as part of the screening procedure for children. Upon inclusion in these studies, language proficiency was assessed for monolingual children in Hebrew and for bilingual children in both the HL (Russian or English) and the SL-Hebrew. Parents filled in parental questionnaires providing information on maternal and paternal education and occupation, child age and length of exposure to Hebrew and concerns about language development and general development.²

One hundred and fifty children aged 5–7 years were randomly selected to participate in the present study: 45 Hebrew monolinguals (19 with DLD), 57 English–Hebrew bilinguals (20 with DLD) and 48 Russian–Hebrew bilinguals (21 with DLD).³ All children selected came from mid-high Socio Economic Status (SES), operationalized as more than 12 years of maternal education, and lived in English–Hebrew and Russian–Hebrew speaking communities in urban neighborhoods in which the majority language was Hebrew. The children with Developmental Language Disorder (DLD) had a language impairment with no known differentiating condition and no co-occurring disorder. All children had normal IQ and no report of hearing difficulties in the past or present, no observed neurological impairments and no atypical social behavior, as reported by parents. As non-verbal IQ is no longer a criterion in the classification of DLD the children in the present study are a subset of children with DLD, i.e., children with a non-verbal IQ in the normal range. Most children with DLD attend language schools while others are treated by a speech-language pathologist (SLP) in the clinic. Of the 105 bilingual children, 40 were exposed to Hebrew at home or within the early educational system before the age of 2 (25 were exposed to Hebrew from birth) and the remaining (65 children) at a later age. These 40 children were distributed proportionally across the different groups. Length of exposure varied from a year to five years (for those with early exposure). Current ways of conceptualizing the nature of the bilingual experience suggest that a dichotomous distinction between bilinguals and monolinguals is somewhat out of step and a more continuous measure should be used, especially since some of the participants are simultaneous bilinguals while others are sequential bilinguals. Yet, initial analyses that divided the bilinguals by type of bilingualism or by language dominance (measured by performance gaps between the languages) showed that most children showed better performance in Hebrew (e.g., were Hebrew dominant) and their Hebrew narrative abilities (tested in this article) were similar within each group. Thus, the dichotomous approach was maintained here.

SL-Hebrew proficiency was measured by the Goralnik Screening Test for Hebrew (Goralnik, 1995) that consists of six subtests for vocabulary, sentence imitation, comprehension, expression, pronunciation and storytelling. Each subtest has a maximum score of 30, and the maximum score for the whole test is 180. Raw scores were converted to z-scores using population adequate norms. HL-English proficiency was measured by the Clinical Evaluation of Language Fundamentals Preschool 2 (CELF; Wiig, Secord, & Semel, 2004), and the standardized scores from the Sentence Structure, Word Structure and Expressive Vocabulary subtests were used to determine a norm-referenced Core Language Score (CLS) for which the standard mean (SD) is 100 (15). HL-Russian proficiency was measured by the Russian Language Proficiency Test for Multilingual Children (Gagarina, Klassert, & Topaj, 2010) that includes expressive (noun/verb naming, production of case and verb inflections) and receptive measures (comprehension of grammatical constructions, receptive vocabulary). The maximum score of the test is 112, and raw scores are converted to z-scores for the adequate population mean. In opting for a cut-off point of −1.25 SD, the cut-off point for the CELF was a core score of 81 and above, the cut-off point for the Russian language test was −1.25 SD from the provisional bilingual norms proposed by Armon-Lotem and Meir (2016), and the cut-off point for the Goralnik was set to −1.25 SD below the provisional bilingual norms proposed by Altman et al. (2016). Monolingual norms of the Goralnik were used for monolingual children with a cut-off point of −1.25 SD as well. These measures of language proficiency together with parents’ reports of concerns regarding their children’s language development enabled a reliable classification of children by their linguistic abilities into children with TLD in at least one of their languages and children with DLD, who show delays in both languages (Håkansson, Salameh, & Nettelbladt, 2003).

Table 1 presents the children’s demographic information and language scores.

Table 1.

Demographic information and language scores for the six groups of children (language proficiency in Hebrew is presented in raw scores).

	MoTLD	BiTLD-E	BiTLD-R	MoDLD	BiDLD-E	BiDLD-R
N (girls)	26 (13)	37 (21)	27 (19)	19 (9)	20 (11)	21 (7)	n.s.
Age	70.9 (1.9)	70.6 (3.0)	69.3 (2.6)	70.4 (5.7)	74.2 (6.0)	70.7 (2.9)	F = 4.158*
Age of L2 onset		24.1 (16.8)	22.3 (22.4)		32.3 (22.0)	31.8 (21.2)	n.s.
Hebrew score	156 (8.4)	145 (11.5)	142 (13.6)	132 (11.2)	131 (11.6)	123 (13.2)	F = 25.208**

p < .01, ** p < .001.

A one-way ANOVA showed a significant difference in chronological age, traced by post hoc tests to a significant difference between the group of English–Hebrew bilinguals with DLD that included a few older children and the other groups. No other differences in age emerged. The bilingual groups showed no significant differences in age of onset and length of exposure. As far as linguistic abilities in Hebrew were concerned, a one-way ANOVA showed a significant difference between the groups on the raw scores in the screening test with monolinguals with TLD outperforming both bilingual groups with TLD (p = .01 for English–Hebrew bilinguals, and p = .001 for Russian–Hebrew bilinguals), with no difference between English–Hebrew and Russian–Hebrew bilinguals (p = .936). There was no significant difference between the groups with DLD, but as expected each of these groups was significantly lower from language-matched TLD peers (MoDLD < MoTLD, p < .001; BiDLD-E < BiTLD-E, p = .003; BiDLD-R < BiTLD-R, p < .001).

As there was no difference between the bilingual groups with TLD and no difference between the bilingual groups with DLD, and both groups with DLD were significantly lower than the bilingual groups with TLD, the bilingual groups were collapsed into one group with TLD and one with DLD. Thus, for the rest of the article, two participant variables are considered – language background (monolingual vs. bilingual) and language ability (TLD vs. DLD) – yielding four groups: monolinguals with TLD and DLD and bilinguals with TLD and DLD.

Procedure

Narratives were elicited from all the participants using the narrative subtest of the Goralnik Screening Test for Hebrew (Goralnik, 1995), consisting of six pictures depicting a story about an intruding bee flying into the kitchen to interrupt a family at mealtime, and their sleeping cat who tries to get rid of the bee (see Appendix 1). This narrative was selected as it offers a rich array of causal relations within the main episode (see Figure 1). Data were collected from all children by native speakers of Hebrew in their preschool, after a short warm-up discussion with the child. After looking at the pictures presented in a book format, children were asked to tell a story using the cue paam axat … ‘Once upon a time …’ with the pictures present while they narrated the story. The story was transcribed during the session by the research assistant and the whole session was recorded.

Figure 1.

Causal network of the story in pictures 2 to 4.

Categories of analysis

The analytical approach proposed here considers the top-down level of causal relations between the episodic components and its relation to the bottom-up selection of particular language forms that serve to depict the components and relate between them into a cohesive and coherent episode. In this context, the clause (Berman & Slobin, 1994, pp. 660–663) was considered as the basic unit for the top-down and bottom-up analyses.

Causal relations between the episodic components

Following the causal network approach (Trabasso, van den Broek, & Suh, 1989), the present study focuses on four types of causal relations – motivational, physical, enabling and psychological – that can be inferred between the episodic components within a GAO unit (goal–attempt–outcome). Goals motivate actions or attempts, which can physically cause or enable a series of outcomes. Outcomes, attempts, or other story events can psychologically cause emotional or mental internal responses, which subsequently motivate other goals (Trabasso & Nickels, 1992; Trabasso & Van Den Broek, 1985). Focus is here on the causal relations between the components in episode 1, which comprises a chain of motion events – a bee flies into the kitchen, the cat jumps toward it, falls into the table and spills the soup (see Appendix 1). These relations are presented in Table 2 and represented by arrows in Figure 1, where episodic components are shown in capitals and causal relations in lower case letters.

Table 2.

Causal relations between the episodic components (in caps) in episode 1 (pictures 2–4).

Causal relation	Role in the narrative
Motivational relation [mot]	Links the initiating event (IE) with the goal (G) leading to the attempt (A₁)
Enabling relation[enb₁][enb₂][enb₃]	Links the attempt (A₁) with the chain of events in outcome 1 (O_1a-c-Cat):‘jumping onto the bee’ enables ‘falling onto the table/dishes’‘falling onto the table/dishes’ enables ‘spilling the soup’‘spilling the soup’ enables ‘cat is dirty’.
Physical relation [phys]*	Links the attempt (A₁) with the outcome bee (O_1-Bee)
Psychological relation	Links the internal responses with the main episodic events:
[psy₁]	Links the initiating event (IE) with the internal response (IR₁) of the family/cat leading to the goal (G)
[psy₂]	Links the attempt (A₁) with the internal response (IR₂) of the family/cat
[psy₃]	Links the outcomes (O_1a-c-Cat) with the internal response (IR₃) of the family/cat

In a physical relation event A is both sufficient and necessary for B to occur, in an enabling relation, A is necessary but not sufficient for B to occur (Tapiero, Van den Broek, & Quintana, 2002).

The translated excerpt in (1) from a monolingual participant with TLD, aged 5 (taken from the corpus in Kupersmitt, 2016), further illustrates the four types of relations. In the example, the episodic components are in capitals, the clauses are shown between brackets, and the causal relations appear in lower case letters at the end of the clauses that are connected by such relations. In all the examples, the information between square brackets includes language background (Monolingual = Mo; Bilingual = Bi); language ability (TLD, DLD), L1 (Rus = Russian, Eng = English, only for bilinguals), and age.

Episodic component	Story [MoTLD, 5;10]
S (1–2)	Mother prepares [= is preparing] soup to-eat for dinner and they eat the soup.
IE (3)	And then came a bee _{psy1; mot}
IR₁ (4–5)	and wanted to-eat the soup because she thought that it was nectar.
IR₁;G (6–7)	And the cat got-angry with the bee _psy1, he wanted to-eat her. _mot
A₁ (8–9)	And the cat jumped and tried to-prey on the bee _{mot; enb1; enb2; phys}
O_1a-b-Cat (10–11)	but at the same moment the soup spilled on the table _{enb1;enb2; enb3} and all the soup spilled on the cat _enb3;psy3
O_1-Bee (12)	and the bee flew (away) _phys
IR₃ (13)	and mother was sorry _psy3

(1) Excerpt taken from a story produced by a monolingual child with TLD, aged 5

The events in this goal-oriented episode are causally related in various ways. A motivational relation between the initiating event in clause 3 and the goal in clause 7 leads to the attempt in clauses 8–9. Besides, psychological, enabling and physical relations can be inferred throughout the episode. These causal relations are implied by the content categories of the episode and they can be marked by specific causal terms, or unmarked, as in most cases (Sah, 2015).

Language forms: focus on motion events

The linguistic analysis focuses on the array of forms that serve to encode this complex chain of motion events in the first episode. Consider example (2), from a typically developing bilingual.

(2) ha-xatula kafca al ha-šulxan litfos et ha-dvora ve hi nafla al ha-marak šel ha-yeled ve ha-yalda.

‘the cat jumped onto the table to-catch the bee and she fell into the boy’s and the girl’s soup’ [BiTLD, Eng, 6;5]

In (2) the noun ‘cat’ represents the figure in motion, the nouns ‘table’ and ‘soup’ represent the grounds with respect to which the figure moves and provide the spatial anchors to the motion of both the cat and the bee (Hickmann, 2003). The prepositions ‘onto’ and ‘into’ mark the path of motion, i.e., the trajectory followed by the figure, the motion verb ‘jumped’ encompasses the manner in which the figure moved with respect to the grounds (as compared to ‘run’, ‘walk’ or ‘swing’), and the motion verb ‘fell’ encodes the end of the trajectory. The phrases comprising the preposition + ground encode the goal of the motion (Talmy, 2000).

Since motion events within and between episodes have a strong causal role (Kupersmitt, 2016; Kupersmitt et al., 2014), selecting particular language forms for their expression should reflect the interaction between bottom-up and top-down levels of narrative construction. In example (2), the manner verb ‘jumped’ comprises the attempt of the protagonist to reach the goal ‘catch the bee’, leading to a negative outcome encoded by the path verb ‘fell’. The path prepositions (‘onto’ and ‘into’) are indicating goal-oriented motion, i.e., motion with a clear endpoint reference (e.g., ‘he walked to the tree’ as compared to ‘he walked around’) (Muentener & Lakusta, 2011). Thus, in this particular story, the series of motion events in the first episode represents a causally related GAO.

For the purpose of the linguistic analysis, the forms considered include specific motion verbs of path (e.g., fall, land) and manner (e.g., jump, pounce), prepositional phrases to encode the paths of motion with their corresponding grounds (e.g., at the bee, onto the table) to serve as the goals of the motion, and three types of clause connectivity adapted from Berman and Nir-Sagiv (2009), as presented in Table 3, and further illustrated in (3) from a TLD monolingual.

Table 3.

Types of clause connectivity.

Isotaxis [ISO]	An isolated main clause (a) with no overt syntactic connection with the surrounding clauses; or (b) headed by an overt marker of temporal sequentiality (e.g., ve az ‘and then’) or by the conjunction ve ‘and’ where the function is not necessarily syntactic or semantic but rather marks the linear progression of the narrative.
Parataxis [PAR]	Two or more clauses juxtaposed or coordinated by means of ve ‘and’, o ‘or’, aval ‘but’, or by zero. This category includes coordinated clauses with subject ellipsis as well as complement (dependent) clauses.
Hypotaxis [HYPO]	Two or more clauses related by subordination, including finite and non-finite adverbial or relative clauses.

(3) ha-xatul kafac litfos et ha-zvuv aval bimkom ze hu kafac al ha-šulxan ve kol ha-marak nišpax

‘the cat jumped [HYPO] to-catch the-bee [PAR] but instead he jumped onto the table [PAR] and all the soup spilled’ [MoTLD, 5;05]

The first clause is followed by a subordinated clause [HYPO] that carries the goal in path (‘to-catch the-bee’). These are followed by two coordinated clauses [PAR] expressing the second phase of the motion. This compact syntactic arrangement of the motion information across clauses plays a significant role in the inference of motivational and enabling causal relations within the GAO.

Coding and statistical measures

All narratives were analyzed and coded by the first author. Each of the above categories were coded as present (1) or absent (0). A ratio of causal relations out of a total of 8 was calculated (see Table 4) along with the percentage of children in each group who expressed each type of relation (Table 5) and each of the language forms that serve to encode the motion event (Table 6).

Table 4.

Ratio of causal relations, by group.

Group		Mean	SD	N
Monolingual	TLD	.33	.21	26
	DLD	.08	.12	19
	Total	.22	.22	45
Bilingual	TLD	.32	.21	64
	DLD	.16	.15	41
	Total	.26	.21	105
Total	TLD	.32	.21	90
Total	DLD	.13	.14	60

Table 5.

Per cent of narrators who expressed the four different types of causal relations, by group.

Causal relations	Group
	MoTLD	BiTLD	MoDLD	BiDLD	χ²
mot	58	52	21	29	11.107*
psy₁	46	59	5	32	20.312**
psy₂	35	30	11	17	5.570
psy₃	42	27	16	22	4.869
enb₁	23	28	5	5	11.770*
enb₂	39	20	0	12	12.316*
enb₃	12	14	5	2	4.516
phys	12	27	0	5	13.844*

p ⩽ .01, ** p ⩽ .001.

Table 6.

Per cent of narrators using different language forms to express the core motion event, by group.

	Linguistic forms		MoTLD	BiTLD	MoDLD	BiDLD	χ²
Motion event – Phase I	Motion-Bee	V (+path)	54	63	37	37	8.396
	Motion-Bee	V (-motion)	35	31	26	49	4.328
	Motion-Cat	V (+manner)	69	64	42	68	4.525
	Motion-Cat	V (other)	23	23	21	15	1.304
	Goal (in path)	PP	39	22	32	17	4.692
		CL-ISO	8	22	0	17	6.815
		CL-PAR/HYPO	42	34	5	17	11.365*
Motion event – Phase II	Motion-Cat	V (+endpoint)	0	13	0	0	11.356*
		V (-endpoint)	19	17	0	22	4.760
		V (-motion)	89	66	63	54	8.701
	Goal (at endpoint)	PP	77	48	16	22	26.230**
		Noun – Spatial anchor	23	17	5	2	8.548
		CL-ISO	42	31	26	37	1.637
		CL-PAR/HYPO	42	45	0	2	33.290**

p ⩽ .01, ** p ⩽ .001.

Eighteen narratives, three from each group of participants (12%), were chosen randomly and recoded after a six-month interval for intra-rater reliability by the first author who developed the analytical framework used in this article (see Kupersmitt, 2016; Kupersmitt et al., 2014). Intra-class correlation coefficients (ICC) for the ratio of causal relations yielded an ICC score of .921. For the binomial data (linguistic forms) Cohen’s κ was used for 17 different linguistic categories. There was a strong agreement for 10 categories (1.00 > κ > .824, p < .001) and moderate agreement for 6 categories, (.766 > κ > .526, p < .01). One category yielded surprisingly low, yet positive agreement (k = .438, p = .063), but this reflected the very low use of this category by the children (which made it inappropriate to use Cohen’s κ). Descriptively, there was an agreement on 16 of the 18 narratives (89%).

The ratio data were analyzed using parametric tests. Equality of variance was tested with Levene’s test, and Tamhane’s T2 post hoc analyses were used when this was violated. In all other cases the data met the expected assumptions for parametric statistics. Non-parametric tests were used for comparing the number of participants who were making use of specific functions or forms based on detailed analyses of the data. These data are binomial comparing children who made use of a particular category with those who did not. Chi-square tests of independence are used to report differences between the four groups followed by three-way cross-tab chi-square tests with language background and language ability as independent variables to explore the separate effects of each variable. For the chi-square analyses, Holm–Bonferroni correction was used to compensate for Type 1 errors due to multiple comparisons and reduce the risk of artificially increasing Type 2 errors due to a lack of statistical power. The p-values in the manuscript are the actual p-values rather than the corrected ones, to provide the fullest information.

Regression analyses were conducted to address the impact of non-dichotomous variables (e.g., raw scores from the Hebrew screening test; age of onset for bilingualism) on causal relations and correlational analyses checked for possible relations between linguistic forms and the ability to produce causal relations.

Results

As the basis for a comparative analysis of the dependent variables and as a general measure of language production, the stories were analyzed by number of clauses. One-way ANOVAs showed significant group differences (BiTLD, BiDLD, MoTLD and MoDLD) on number of clauses (F_{(3, 146)} = 5.518, p = .001, η² = .102). Equality of variance across the groups could not be assumed (Levene’s, F_(3,146) = 3.135, p = .027). A two-way ANOVA with language ability and language background as independent variables and number of clauses as the dependent variable revealed a significant interaction of the former (F_(1,144) = 10.539, p < .001). Tamhane’s T2 post hoc analyses showed that number of clauses differentiated between monolingual children with TLD (M = 16.35) and their peers with DLD (M = 10.42) (p = .003), but not between bilingual children with TLD (M = 13.08) and their peers with DLD (M = 12.98), with no bilingual effect. No difference in number of clauses was found among the bilinguals by HL as an independent variable (p = .335). In the sections below, results are presented for occurrence of causal relations and the factors related to their verbalization and for the array of language forms used to express causal relations in the four groups, followed by a more qualitative form–function analysis.

Expression of causal relations

One-way ANOVAs showed significant group differences (BiTLD, BiDLD, MoTLD and MoDLD) on ratio of causal relations (F_{(3, 146)} = 13.249, p < .001, η² = .214). Equality of variance across the groups could not be assumed (Levene’s, F_(3,146) = 5.765, p = .001). As expected, Tamhane’s T2 post hoc analyses showed no significant differences between monolingual and bilingual children with DLD, nor between monolingual and bilingual children with TLD, while both groups with DLD demonstrate significantly lower performance than the groups with TLD, as shown in Table 4.

Table 5 presents the percentage of narrators who expressed inferable motivational, psychological, enabling and physical relations in each group, revealing the source of the quantitative differences.

Up to 60% of the TLD monolinguals and bilinguals express some type of causality, but rates are much lower among their DLD counterparts. A chi-square test of independence on the ratio of children using each causal relation showed a significant difference between the groups for all but three relations (psy₂, psy₃ and enb₃).

A three-way cross-tab chi-square with causal relations as dependent variables and language ability (TLD, DLD) and background (Mo, Bi) as between-subjects variables showed that monolinguals were similar to bilinguals in all causal relations. By contrast, using a Holm–Bonferroni correction a significant effect of language ability (TLD, DLD) was found for psy₁ in both monolingual and bilingual groups (χ² = 8.935, p = .003 and χ² = 7.658, p = .006, respectively), for enb₁ (χ² = 8.759, p = .003) and phys (χ² = 7.929, p = .005) in bilinguals only, and for enb₂ in monolinguals (χ² = 9.396, p = .002). Nearly significant differences were found for enb₃ in the bilingual group only, for psy₂ and psy₃ in monolinguals only, and for mot in monolinguals and bilinguals.

To address the heterogeneity of the groups a stepwise regression was conducted exploring whether non-dichotomous measures (e.g., language screening test scores) are better predictors of the use of causal relations than the dichotomous division presented above (Mo, Bi, TLD and DLD). For all children together, language screening scores, number of clauses, chronological age, language ability and language background were introduced in this consistent order across all regression analyses.

TLD significantly predicted the ratio of causal relations (ß = .450, t (144) = 6.122, p < .001) and also explained 20% of the variance in the ratio of causal relations (F_{(1, 149)} = 37.481, p < .001). When number of clauses is added to the model, both TLD and number of clauses significantly predicted the ratio of causal relations (ß = .387, t (144) = 5.617, p < .001 for TLD, ß = .355, t (144) = 5.156, p < .001 for number of clauses), explaining together 32.4% of the variance (F_{(2, 148)} = 35.270, p < .001). For the bilingual children only, a model with these two variables (number of clauses and TLD) significantly predicted the ratio of causal relations (ß = .392, t (96) = 4.687, p < .001 for TLD, ß = .365, t (96) = 4.371, p < .001 for number of clauses), explaining together 30.4% of the variance (F_{(2, 100)} = 21.875, p < .001), while age of onset and length of exposure were excluded.

Looking at each of the four types of causal relations as separate variables, the analyses offered different models for different causal relations. Language screening scores significantly predicted ratio of enabling causal relations in all groups (ß = .357, t (144) = 4.646, p < .001). They also explained 12.7% of the variance in ratio of enabling causal relations in all groups (F_{(1, 148)} = 21.584, p < .001) and 11.4% of the variance taking only bilingual children (F_{(1, 101)} = 12.999, p < .001). By contrast, number of clauses and TLD significantly predicted psychological causal relations in all groups together (ß = .526, t (144) = 7.850, p < .001 for number of clauses and ß = .213, t (144) = 3.175, p = .002 for TLD), explaining 36% of their variance in all groups together (F_{(2, 147)} = 41.533, p < .001) and 34.7% when taking the bilingual group only (F_{(1, 101)} = 26.555, p < .001). No relations were found between the remaining causal relations (mot and phys) and any of the independent variables.

Language forms

Table 6 presents the language forms used by the narrators at phase I and II of the motion event in the first episode, including verbs, prepositional phrases to encode the goals of motion, nouns as spatial anchors and types of clause connectivity.

A chi-square test of independence was performed to examine the relation between group and the ratio of children using each form. Holm–Bonferroni correction was used for each phase of the motion separately (Phase I and Phase II, as shown in Table 6). In the first phase a nearly significant difference was found between the groups in use of parataxis (coordination) or hypotaxis (subordination) to connect between the attempt and the goal in path (bee), for path verbs when referring to the incoming bee, and for isolated clauses to mark the goal in path. In the second phase of the motion, the difference between the groups was statistically significant in the use of a specific motion verb that marks the endpoint of the motion (e.g., land) and in the use of prepositional phrases that mark path and ground. A nearly significant difference was observed for non-motion verbs to refer to the end of the motion (e.g., bumped, broke) and for nouns that serve as spatial anchors.

A three-way cross-tab chi-square with linguistic forms as dependent variables and language ability (TLD, DLD) and background (Mo, Bi) as between-subjects variables were used to tease apart the impact of these independent variables. As above, Holm–Bonferroni correction was used for each phase of the motion separately for each group of participants. Where the number of children using the different means is very small, only descriptive information is provided with no further analysis.

Verbs

Reference to the incoming bee by a specific verb of path was more frequent in children with TLD (around 60%) than with DLD (37%), yielding a significant difference between BiDLD and BiTLD (χ² = 6.728, p = .009), with no difference for monolinguals. Less specific verbs were used by 30–50% of the children in all the groups, including copular (e.g., ‘and then (there) was a fly’ [BiTLD, Rus, 5;04]), or perception verbs to present the bee from the boy’s or cat’s perspective (e.g., ‘and suddenly the cat woke-up and saw the bee’ [BiDLD, Eng, 6;03]).

In order to mention the first phase of the motion children in all groups preferred the specific verb kafac ‘jumped’, with a nearly significant advantage for MoTLD (69%) over MoDLD children (42%) (χ² = 3.311, p = .069). Around 20% used other non-motion verbs (e.g., ‘the cat got-rid of the bee’ [MoTLD, 5;00]), less specific motion verbs (‘and then the cat came to-the-fly’ [BiTLD, Eng, 6;00]), or semantically inappropriate motion verbs, particularly DLD children (e.g., ‘and the cat chased after the fly’ [BiDLD, Eng, 7;03]. No significant differences were found between the groups in the use of these forms.

At the second phase of the motion, very few children (only 13% of the TLD bilinguals) mentioned the end of the path by a specific verb (e.g., ‘at the end landed the cat onto the two dishes’ [BiTLD, Eng, 5;07]). Around 20% of the children in all groups except for MoDLD opted for the manner of motion verb kafac ‘jump’, sometimes repeating the verb twice to express the attempt and outcome (e.g., ‘and the cat jumped and the children got-scared and then the cat jumped onto the table’ [MoTLD, 5;11]). Most of the children (e.g., 89% among MoTLD) used a non-motion verb to express the outcome of the motion (e.g., ‘and the cat spilled the boy’s meal’ [BiTLD, Eng, 5;11]), showing a nearly significant advantage for MoTLD over BiTLD (χ² = 4.806, p = .028) and MoDLD children (χ² = 4.067, p = .044).

Prepositional and noun phrases

Beyond the verbal nucleus, monolinguals performed better than bilinguals when referring to the bee as goal in path by means of a prepositional phrase (e.g., ‘jumped onto the-bee’ [BiDLD, Rus, 5;10]) but this difference was not significant. However, TLD children significantly outperformed their DLD peers in use of a PP to mention the endpoint (e.g., table or dishes) (χ² = 16.419, p < .001 and χ² = 7.434, p = .006, in monolinguals and bilinguals respectively). Besides, BiTLD performed almost significantly better than BiDLD children when providing a noun phrase (e.g., soup, dishes) as spatial anchor (χ² = 5.370, p = .020).

Clause connectivity

MoTLD children significantly outperformed MoDLD children in use of paratactic/hypotactic strategies of clause connectivity to mention the goal in path (χ² = 7.704, p = .006) and a nearly significant advantage was found for BiTLD in the bilingual group (χ² = 3.742, p = .053). Also, an almost significant advantage was found for BiTLD over BiDLD in reference to the goal in path by means of isolated clauses (χ² = 3.672, p = .055). For the purpose of connecting the motion events in the first and second phase of the motion, TLD children used significantly more complex syntactic strategies than their peers with DLD in both monolingual and bilingual groups (χ² = 10.639, p = .001 and χ² = 22.509, p < .001, respectively). Comparative examples of these strategies are presented in Table 7.

Table 7.

Examples of TLD and DLD ability to arrange the clauses.

	Hebrew (Original)	English translation
(4)	ve az ha-xatula kama ve racta litfos et ha-dvora aval az hi kafca al ha-šulxan litfos oto ve hi nafla al ha-marak šel ha-yeled ve ha-yalda*.[BiTLD, Eng, 6;5]	‘and then the cat woke-up [PAR] and (she) wanted to catch the bee [PAR] but then she jumped onto the table [HYPO] to-catch *him [PAR] and she fell onto the boy’s and the girl’s soup.’
(5)	ha- xatul kam ve raa. raca litfos. kafac ve- nafal al ha-šulxan ve lo tafas et ha- dvora.[BiTLD, Rus, 5;09]	‘the cat woke-up [PAR] and (it) saw. [ISO] (it) wanted to catch [ISO]. (It) jumped [PAR] and fell onto the table [PAR] and didn’t catch the bee.’
(6)	ve ha-xatul kafac al zvuv ve ha-yeled harag et ha-zvuv. Ima ra’ata. Hi hayta acuva ve yalda gam hayta acuva ve herima et ha-yad. Az ha-xatul kafac al ha-oxel šel yeled ve yalda.* [BiDLD, Rus, 5;07]	‘and the cat jumped at bee [ISO] and the-boy killed the-bee. [ISO] Mother saw. [ISO] She was sad [ISO] and girl also was sad [ISO] and (she) raised the-hand. [ISO] Then the-cat jumped onto the food of boy and girl.’

show errors in gender marking and ommision of the definite determiner.

In example (4), the motion events occur in coordinated clauses with the same topic and the goal is expressed in a subordinated clause (i.e., parataxis and hypotaxis). Example (5) though less accurate in lexical terms shows tighter connectivity between the attempt and its negative outcome (‘and it didn’t catch the bee’), which also implies the goal of the motion. In (6), contrastively, all the clauses are isolated, and no cohesive links are found between the attempt, the goal and the outcome.

As expected, a correlational analysis shows very moderate yet significant correlation between use of paratactic/hypotactic strategies of clause connectivity at the first phase of the motion and the ratio of all four types of causal relations (r_s = .254, p = .002), that emerges from correlations with the enabling causal relations (r_s = .247, p = .002), and the motivational relation (r_s = .318, p < .001). Moderate correlations were also found between the former two types and the reference to grounds as spatial anchors (r_s = .265, p = .001 and r_s = .267, p = .001, respectively). No further correlations were found between other linguistic forms and causal relations.

Discussion and conclusion

This article focused on the linguistic forms used to encode a complex chain of motion events, as a particular domain that plays a central role in connecting the narrative units at the clausal, inter-clausal and episodic levels. The article investigates the sensitivity of these linguistic forms and causal relations to language background and language abilities.

Top-down abilities – the expression of causal relations at the episodic level

From a top-down perspective, the analysis of causal relations referred to an abstract level of representation considering how the episodic components of the story become causally and linguistically connected (Trabasso & Rodkin, 1994) rather than the presence or absence of the components themselves. Around a third of the children with TLD show an emergent ability to express inferable causal relations, as reported in previous studies with monolingual children of similar age (Sah, 2015; Trabasso & Nickels, 1992), while children with DLD showed poorer performance.

Establishing enabling and motivational relations was challenging overall. Children often failed to attribute intentionality to the motion event that triggered the whole episode (jumping to catch the bee) and to establish enabling relations between the motion, the goal in path and the goal at endpoint. This finding contradicts those of previous studies reporting that enabling relations were relatively frequent among 5-year-olds (Trabasso & Nickels, 1992; Trabasso, Stein, Rodkin, Munger, & Baughn, 1992). Since enablement is a weaker relation as compared to the other three types (Tapiero, Van den Broek, & Quintana, 2002; Trabasso et al., 1989) it seems important to investigate this subtle type of linkage in different narrative contexts and in scenes varying in event complexity. Here, making inferences about the relation between the goal, actions and outcomes in the complex motion chain was demanding and resulted in a poor verbalization of such relations, particularly in the DLD groups.

The expression of a physical relation between the attempt and the outcome related to the bee (flying away) was scarce overall and almost nonexistent in DLD narratives. In order to mention the outcome referring to the bee, narrators should maintain the mental representation of the hierarchical goal plan, without being distracted by the events in between, as depicted in complex scenes with multiple perspectives. It has been claimed that the distance of causal connections plays an important role in their inference and expression (Tapiero et al., 2002; Trabasso & Nickels, 1992), which may explain the absence of this type of relation.

Psychological relations between internal responses and their precipitating events were less affected by language background or ability when the responses were triggered by salient facial expressions and body gestures (Bamberg & Damrad-Frye, 1991; Kupersmitt, 2016; Kupersmitt et al., 2014). Nonetheless, a difference appeared in those relations linking the initiating event and the goal, which demanded a more abstract representation of goal-directed action. While most children in all groups were able to express ‘motion’ and ‘emotion’ concurrently, only children with TLD – and a few bilinguals with DLD – showed a further development towards the expression of ‘motivation’ and ‘intention-in-motion’, reflected in this case by a response that leads to goal-oriented motion (Kupersmitt, 2016; Nicolopoulou & Richner, 2007).

All in all, the findings on causal relations reflect the meticulous criteria adopted for the analysis that controlled the type of relations and their function in story coherence at the local and global levels, rather than considering causal relations indiscriminately. Longer texts were more likely to include causal relations, particularly psychological relations, but not necessarily more complex types of causal relations like enabling. This finding indicates that elaborations on the emotional states of the characters often yield longer utterances, while enabling relations were inferred from the expression of the attempt, the goal and the outcome, usually occurring in compact constructions.

Bottom-up abilities – the linguistic representation of the causal chain

Overall, children with TLD and DLD at the ages studied here only begin to elaborate on the expression of motion events in specific ways at the clausal and inter-clausal levels (Guo & Chen, 2009; Hickmann, 2003; Özçalişkan, 2009). From the first phase of the motion chain, children with DLD seemed particularly disadvantaged when narrating the ‘jumping scene’, which demanded complex language forms to encode multiple perspectives on motion, causality and reference. Struggling to access specific linguistic resources, they often provided a limited description of actions which failed to fit into a clear causal chain. Children in all groups were able to express the attempt by a specific verb like ‘jump’, but less likely to mention the outcome – ‘landing or falling on the table’ – which is also less salient in the pictorial stimuli. Thus, even when assuming that most of the children (including bilinguals) knew basic verbs such as nafal ‘fall’, almost none of them used this verb to encode the end of the path at the second phase of the motion, which was required for achieving coherence and advancing the plot.

Expressing information about grounds and goals of motion was less frequent altogether, replicating previous findings for children with TLD at similar ages (Berman & Slobin, 1994; Hickmann, 2003; Kupersmitt, Harr, Engemann, & Licandro, 2017; Ochsenbauer & Hickmann, 2010). Thus, only half of the children with TLD could mention the ‘bee’ as goal in the path of motion by means of a prepositional phrase or by a separate clause, as reported in Kupersmitt (2016), and children with DLD mentioned it far less frequently. Likewise, children with DLD were poorer than children with TLD when referring to the endpoint of the motion (‘table’ or ‘dishes’) and were less likely to make reference to spatial anchors, compared to their TLD peers.

The novel form–function approach undertaken in this study provided further insights about the interaction between bottom-up processes at the lexical and syntactic levels and top-down principles of discourse organization (Karmiloff-Smith, 1981). The analysis provided an encouraging picture with respect to bilingual children with TLD and some advantage to bilingual children with DLD, in the domain of causal relations and language forms. This may be attributed to methodological reasons.

First, unlike micro-level analyses in previous studies (Altman et al., 2016) the forms analyzed here were tightly related to the function they served – rather than counted indiscriminately – and were not language specific (e.g., morphemes). In typically developing bilinguals, top-down processes of discourse organization as well as typical knowledge of the HL could have paved the access to a range of language forms, even in Hebrew as the SL.

Second, most studies that reported different use of language forms in bilingual narratives examined the weaker language in non-dominant bilinguals (Kaufman, 2001; Kupersmitt, 2004). In the present study Hebrew, the societal language, is the dominant language for most participants, which can also explain the lack of difference between the bilingual group and the monolingual group. As well, correlations excluded age of onset and bilingual experience as possible predictors of the children’s performance, corroborating that the bilinguals included in this study were well beyond the first stages of L2 acquisition (Kupersmitt et al., 2014). Third, the socio-cultural, linguistic and literacy-related background of the bilingual population could affect the language performance. In this particular study, the bilingual population speaks English, a highly prestigious language, which is widely used in educational settings and in the media, and Russian, a heritage language strongly maintained and supported among the first, second and even third generation of Russian-speaking immigrants (Altman et al., 2014).

Taking monolingual and bilingual populations, with and without DLD, the study sheds further light on current theoretical assumptions related to bilingualism and language impairment. Bilingual children are expected to differ from monolinguals in linguistic experience and skills necessary for bottom-up processes, but to share some knowledge of narrative organization and structuring. Despite the heterogeneous nature of the bilingual and DLD groups, the criteria adopted for differentiating between TLD and DLD in monolinguals and bilinguals were most accurate in predicting the variance in causal relations. The finding that bilinguals and monolinguals performed similarly on causal relations, regardless of language proficiency or exposure to the language of the story, suggests that this top-down ability can cut across languages, at least in dominant bilinguals at the ages examined here.

As language ability (TLD vs. DLD) rather than linguistic proficiency predicted the use of causal relations, the regression models support the idea that the difficulties in discourse that children with DLD might present can be attributed to subtler linguistic and cognitive difficulties, including the ability to maintain the relations between two story elements and to plan a coherent narrative.

The study also corroborated previous findings about the particular difficulty of children with DLD to connect the narrative content into cohesive syntactic units (Marini, Tavano, & Fabbro, 2008; Norbury & Bishop, 2003; Reilly, Losh, Bellugi, & Wulfeck, 2004). Monolinguals and bilinguals with TLD showed an emergent ability to arrange the chain of motion events in unified syntactic packages by means of paratactic and hypotactic strategies of connectivity. Children with DLD, by contrast, could hardly portray a syntactically unified representation of the whole goal–attempt–outcome (GAO) unit. Interestingly, results showed a tendency for syntactic and referential strategies to be related to the expression of more complex types of causal relations. Though percentages were too small to reach firm conclusions, they can suggest that higher syntactic abilities are required to organize the story in hierarchically, causally related strings (Berman, 2008; Berman & Nir-Sagiv, 2009; Berman & Slobin, 1994).

To conclude, this study brought an original approach to the discriminatory analysis of narratives among bilingual and monolingual children with TLD and DLD, but extending this analysis to other narrative genres is necessary to validate its diagnostic potential. The lack of difference between the bilingual groups suggests that this method can be generalizable to other bilingual groups who have a literate background and are dominant in the language of testing. The detailed attention to causal relations of the four types as well as to the motion events made it possible to explore the group differences in top-down and bottom-up abilities. For children with DLD, the form–function analyses revealed their difficulty with language use in tasks with high processing demands as in narrative production (Berman, 1995).

The integrative perspective of this study makes it possible to move beyond analysis of micro-structure and macro-structure which limited the study of the impact of bilingualism and DLD on narrative abilities, and identify areas which are more influenced by DLD and by bilingualism both in top-down and in bottom-up analysis while showing possible bilingual advantage across these different processes and in the interaction between them. In sum, the study reinforced the differences between children with TLD and children with DLD, with no difference between monolinguals and bilinguals, as typical language development rather than the proficiency in a particular language was found to be responsible for the ability to use causal relation and the particular linguistic forms that yield together a coherent narrative.

Footnotes

Appendix

Appendix 1.

Episodic structure and content of the Goralnik story (Goralnik, 1995).

Picture number	Episode Number	Episodic component		Content
1	1	Setting (S)		Mother washing dishes, boy and girl eating soup, cat sleeping.
2	1	Initiating event (IE)		Bee flies into kitchen.
2	1	IR₁		Boy and girl look annoyed. Boy tries to get rid of the bee. Cat wakes up.
2	1	Goal 1		Catch the bee.
3	1	Attempt 1 (A₁)		Cat jumps onto bee. Bee flies down toward the soup on the table.
3	1	IR₂		Boy and girl stand up, shout, and look angry. Mother turns around and looks surprised.
4	1	Outcome 1 Bee (O_1-Bee)		Bee flies away.
4	1 and 2	Outcome 1 Cat*	(O_1a-b-Cat)	Cat lands on the table or dishes / spills food or soup.
		Outcome 1 Cat*	(O_1c-Cat)	Cat gets dirty.
		IR₃		Mother and children look sad/distressed.
4	2	Goal 2		To wash the cat.
5	2	Attempt 2 (A₂)		Mother washes the cat. Boy looks, girl holds towel.
6	2	Outcome 2 (O₂) + IR₄		Cat is clean and happy. Mother and children smile.

Outcome 1 Cat is the Initiating event of the second episode.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Akinci

Jisa

Kern

(2001). Influence of L1 Turkish on L2 French narratives. In Verhoeven

Strömqvist

(Eds.) Narrative development in a multilingual context (pp. 189–208). Amsterdam, The Netherlands: John Benjamins.

Aksu-Koç

A. A.

Tekdemir

(2004). Interplay between narrativity and mind reading. In Strömqvist

Verhoeven

(Eds.), Relating events in narrative: Typological and contextual perspectives (pp. 307–327). Mahwah, NJ: Lawrence Erlbaum.

Aksu-Koç

A. A.

von Stutterheim

(1994). Temporal relations in narrative: Simultaneity. In Berman

R. A.

Slobin

D. I.

(Eds.), Relating events in narrative: A cross-linguistic developmental study (pp. 393–455). Hillsdale, NJ: Lawrence Erlbaum.

Altman

Armon-Lotem

Fichman

Walters

(2016). Macrostructure, microstructure, and mental state terms in the narratives of English–Hebrew bilingual preschool children with and without specific language impairment. Applied Psycholinguistics, 37, 165–193.

Altman

Burstein-Feldman

Yitzhaki

Armon-Lotem

Walters

(2014). Family language policies, reported language use and proficiency in Russian–Hebrew bilingual children in Israel. Journal of Multilingual and Multicultural Development, 35, 216–234.

Armon-Lotem

Altman

Joffe

Oz-Abutbul

Walters

(2015). Ethno-linguistic identity, language exposure and language acquisition in bilingual preschool children from English and Russian-speaking backgrounds. In Grüter

Paradis

(Eds.), Input and experience in bilingual development (TILAR Series, pp. 77–98). Amsterdam, The Netherlands: John Benjamins.

Armon-Lotem

Meir

(2016). Diagnostic accuracy of repetition tasks for the identification of specific language impairment (DLD) in bilingual children: Evidence from Russian and Hebrew. International Journal of Language & Communication Disorders, 51, 715–731.

Bamberg

Damrad-Frye

(1991). On the ability to provide evaluative comments: Further explorations of children’s narrative competencies. Journal of Child Language, 18, 689–710.

Bedore

L. M.

Peña

E. D.

Gillam

R. B.

T. H.

(2010). Language sample measures and language ability in Spanish–English bilingual kindergarteners. Journal of Communication Disorders, 43, 498–510.

10.

Berman

R. A.

(1988). On the ability to relate events in narrative. Discourse Processes, 11, 468–497.

11.

Berman

R. A.

(1995). Narrative competence and storytelling performance: How children tell stories in different contexts. Journal of Narrative and Life History, 5, 285–313.

12.

Berman

R. A.

(1996). Form and function in developing narrative abilities. In Slobin

D. I.

Gerhardt

Kyratzis

Guo

(Eds.), Social interaction, social context, and language (pp. 343–367). Mahwah, NJ: Lawrence Erlbaum.

13.

Berman

R. A.

(2008). The psycholinguistics of developing text construction. Journal of Child Language, 35, 735–771.

14.

Berman

R. A.

(2014). Acquiring and expressing temporality in Hebrew: A T(M/A) Language. SKASE Journal of Theoretical Linguistics, 11, 2–29.

15.

Berman

R. A.

Katzenberger

(2004). Form and functions in introducing narrative and expository texts: A developmental perspective. Discourse Processes, 38, 57–94.

16.

Berman

R. A.

Nir-Sagiv

(2009). Clause-packaging in narratives: A cross linguistic developmental study. In Guo

Lieven

Budwig

Ervin-Tripp

Nakamura

Özҫalişkan

Ş.

(Eds.), Cross-linguistic approaches to the psychology of language: Research in the tradition of Dan I. Slobin (pp. 149–162). Mahwah, NJ: Lawrence Erlbaum.

17.

Berman

R. A.

Slobin

D. I.

(1994). Relating events in narrative: A crosslinguistic developmental study. Hillsdale, NJ: Lawrence Erlbaum.

18.

Bishop

D. V.

Snowling

M. J.

Thompson

P. A.

Greenhalgh

, & the CATALISE-2 Consortium. (2017). Phase 2 of CATALISE: A multinational and multidisciplinary Delphi consensus study of problems with language development: Terminology. Journal of Child Psychology and Psychiatry, 58, 1068–1080.

19.

Boerma

Leseman

Timmermeister

Wijnen

Blom

(2016). Narrative abilities of monolingual and bilingual children with and without language impairment: Implications for clinical practice. International Journal of Language & Communication Disorders, 51, 626–638.

20.

Brown

D. D.

Lile

Burns

B. M.

(2011). Basic language skills and young children’s understanding of causal connections during storytelling. Reading Psychology, 32, 372–394.

21.

Chen

Pan

(2009). Development of English referring expressions in the narratives of Chinese–English bilinguals. Bilingualism: Language and Cognition, 12, 429–445.

22.

COST Action IS0804. (2011). Parents of Bilingual Children Questionnaire (PaBiQ): A part of the LITMUS Battery (COST IS0804). Available from http://www.bi-sli.org

23.

Fichman

Altman

Voloskovich

Armon-Lotem

Walters

(2017). Story grammar elements and causal relations in the narratives of Russian-Hebrew bilingual children with DLD and typical language development. Journal of Communication Disorders, 69, 72–93.

24.

Gagarina

Klop

Kunnari

Tantele

Välimaa

Balčiūnienė

Walters

(2012). MAIN: Multilingual assessment instrument for narratives. ZAS Papers in Linguistics, 56, 1–155.

25.

Gagarina

N. V.

Klassert

Topaj

(2010). Sprachstandstest Russisch für mehrsprachige Kinder [Russian language proficiency test for multilingual children]. ZAS Papers in Linguistics, 54.

26.

Goralnik

(1995). Language screening test for preschool Hebrew language speakers (in Hebrew). Even Yehuda, Israel: Matan.

27.

Guo

Chen

(2009). Learning to express motion in narratives by Mandarin speaking children. In Guo

Lieven

Budwig

Ervin-Tripp

Nakamura

Özҫalişkan

Ş.

(Eds.), Cross-linguistic approaches to the psychology of language: Research in the tradition of Dan I. Slobin (pp. 193–208). Mahwah, NJ: Lawrence Erlbaum.

28.

Håkansson

Salameh

E. K.

Nettelbladt

(2003). Measuring language proficiency in bilingual children: Swedish–Arabic bilingual children with and without language impairment. Linguistics, 41, 255–288.

29.

Hickmann

(2003). Children’s discourse: Person, space and time across languages. Cambridge, UK: Cambridge University Press.

30.

Hickmann

(2004). Coherence, cohesion, and context: Some comparative perspectives in narrative development. In Strömqvist

Verhoeven

(Eds.), Relatingevents in narrative: Typological and contextual perspectives (pp. 281–306). Mahwah, NJ: Lawrence Erlbaum.

31.

Hickmann

(2006). The relativity of motion in first language acquisition. In Hickmann

Robert

(Eds.), Space across languages: Linguistic systems and cognitive categories (pp. 281–308). Amsterdam, The Netherlands: John Benjamins.

32.

Iluz-Cohen

Walters

(2012). Telling stories in two languages: Narratives of preschool children with typical and impaired language. Bilingualism: Language and Cognition, 15, 58–74.

33.

Jisa

(1984). French preschoolers’ use of et pis (‘and then’). First Language, 5, 169–184.

34.

Karmiloff-Smith

(1979). A functional approach to child language: A study of determiners and reference. Cambridge, UK: Cambridge University Press.

35.

Karmiloff-Smith

(1981). The grammatical marking of thematic structure in the development of language production. In Deutsch

(Ed.), The child’s construction of language (pp. 121–147). London, England: Academic Press.

36.

Kaufman

(2001). Narrative development in Hebrew and English. In Verhoeven

Strömqvist

(Eds.), Narrative development in a multilingual context (pp. 319–340). Amsterdam, The Netherlands: John Benjamins.

37.

Kunnari

Välimaa

Laukkanen-Nevala

(2016). Macrostructure in the narratives of monolingual Finnish and bilingual Finnish–Swedish children. Applied Psycholinguistics, 37, 123–144.

38.

Kupersmitt

(2004). Form-function relations in Spanish narratives: A comparative study of bilingual and monolingual speakers. In Strömqvist

Verhoeven

(Eds.), Relating events in narrative: Typological and contextual perspectives (pp. 395–434). Mahwah, NJ: Lawrence Erlbaum.

39.

Kupersmitt

J. R.

(2015). Moving on and off the timeline in narratives of personal experience written in three languages: A form-function developmental study. Linguistics, 53, 1303–1339.

40.

Kupersmitt

J. R.

(2016). When causation meets motion: Motion events at the interface between coherence and cohesion in narrative development. In Perera

Aparici

Rosado

Salas

(Eds.), Written and spoken language development across the lifespan (pp. 227–249). Cham, Switzerland: Springer.

41.

Kupersmitt

J. R.

Harr

A. K.

Engemann

Licandro

(2017). Narrative-embedded motion events: A developmental cross-linguistic study of German and Hebrew. Paper presented at the 14th International Congress for the Study of Child Language (IASCL), July, 17–21, Lyon, France.

42.

Kupersmitt

J. R.

Yifat

Blum-Kulka

(2014). The development of coherence and cohesion in monolingual and sequential bilingual children’s narratives: Same or different? Narrative Inquiry, 24, 40–76.

43.

Mandler

J. M.

(1978). A code in the node: The use of a story schema in retrieval. Discourse Processes, 1, 14–35.

44.

Marini

Tavano

Fabbro

(2008). Assessment of linguistic abilities in Italian children with specific language impairment. Neuropsychologia, 46, 2816–2823.

45.

Minami

(2008). Telling good stories in different languages: Bilingual children’s styles of story construction and their linguistic and educational implications. Narrative Inquiry, 18, 83–110.

46.

Montanari

(2004). The development of narrative competence in the L1 and L2 of Spanish-English bilingual children. International Journal of Bilingualism, 8, 449–497.

47.

Muentener

Lakusta

(2011). The language of causal events: Exploring the link between representations of intentional action and causal events. Cognition, 119, 341–355.

48.

Nicolopoulou

Richner

E. S.

(2007). From actors to agents to persons: The development of character’s representation in young children’s narratives. Child Development, 78, 412–429.

49.

Norbury

C. F.

Bishop

D. V. M.

(2003). Narrative skills of children with communication impairments. International Journal of Language & Communication Disorders, 38, 287–313.

50.

Ochsenbauer

A. K.

Hickmann

(2010). Children’s verbalization of motion events in German. Cognitive Linguistics, 21, 217–238.

51.

Özçalişkan

. (2009). Learning to talk about spatial motion in language-specific ways. In Guo

Lieven

Budwig

Ervin-Tripp

Nakamura

Özҫalişkan

Ş.

(Eds.), Cross-linguistic approaches to the psychology of language: Research in the tradition of Dan I. Slobin (pp. 263–276). Mahwah, NJ: Lawrence Erlbaum.

52.

Pearson

B. Z.

(2001). Logic and mind in Spanish-English children’s narratives. In Verhoeven

Strömqvist

(Eds.), Narrative development in a multilingual context (pp. 373–398). Amsterdam, The Netherlands: John Benjamins.

53.

Peterson

McCabe

(1983). Developmental psycholinguistics: Three ways of looking at a child’s narrative. New York, NY: Plenum.

54.

Peterson

McCabe

(1991). Linking children’s connective use and narrative macrostructure. In McCabe

Peterson

(Eds.), Developing narrative structure (pp. 29–54). Hillsdale, NJ: Lawrence Erlbaum.

55.

Reilly

Losh

Bellugi

Wulfeck

(2004). Narratives in children with specific language impairment, early focal brain injury, and Williams syndrome. Brain and Language, 88, 229–247.

56.

Sah

W. H.

(2015). The development of coherence in narratives: Causal relations. Taiwan Journal of Linguistics, 13, 25–51.

57.

Serratrice

(2007). Referential cohesion in the narratives of bilingual English–Italian children and monolingual peers. Journal of Pragmatics, 39, 1058–1087.

58.

Slobin

D. I.

(1996). From ‘thought and language’ to ‘thinking for speaking’. In Gumperz

J. J.

Levinson

S. C.

(Eds.), Rethinking linguistic relativity (pp. 70–96). Cambridge, UK: Cambridge University Press.

59.

Squires

K. E.

Lugo-Neris

M. J.

Peña

E. D.

Bedore

L. M.

Bohman

T. M.

Gillam

R. B.

(2014). Story retelling by bilingual children with language impairments and typically developing controls. International Journal of Language & Communication Disorders, 49, 60–74.

60.

Stein

N. L.

Glenn

C. G.

(1979). An analysis of story comprehension in elementary school children. In Freedle

R. O.

(Ed.), New directions in discourse processing (pp. 53–120). Norwood, NJ: Ablex.

61.

Talmy

(2000). Toward a cognitive semantics, Vol. II: Typology and process in concept structuring. Cambridge, MA: MIT Press.

62.

Tapiero

Van den Broek

Quintana

M. P.

(2002). The mental representation of narrative texts as networks: The role of necessity and sufficiency in the detection of different types of causal relations. Discourse Processes, 34, 237–258.

63.

Terrell

S. L.

Terrell

(1983). Distinguishing linguistic differences from disorders: The past, present, and future of nonbiased assessment. Topics in Language Disorders, 3(3), 1–7.

64.

Trabasso

Nickels

(1992). The development of goal plans of action in the narration of a picture story. Discourse Processes, 15, 249–275.

65.

Trabasso

Rodkin

(1994). Knowledge of goals/plans: A conceptual basis for narrating ‘Frog, where are you?’ In Berman

R. A.

Slobin

D. I.

(Eds.), Relating events in narrative: A cross-linguistic developmental study (pp. 85–107). Hillsdale, NJ: Lawrence Erlbaum.

66.

Trabasso

Stein

N. L.

Rodkin

P. C.

Munger

M. P.

Baughn

C. R.

(1992). Knowledge of goals and plans in the on-line narration of events. Cognitive Development, 7, 133–170.

67.

Trabasso

Van Den Broek

(1985). Causal thinking and the representation of narrative events. Journal of Memory and Language, 24, 612–630.

68.

Trabasso

van den Broek

Suh

(1989). Logical necessity and transitivity of causal relations in stories. Discourse Processes, 12, 1–25.

69.

Tsimpli

I. M.

Peristeri

Andreou

(2016). Narrative production in monolingual and bilingual children with specific language impairment. Applied Psycholinguistics, 37, 195–216.

70.

Viberg

. (1994). Bilingual development of school-age students in Sweden. In Extra

Verhoeven

(Eds.), The cross-linguistic study of bilingual development (pp. 181–198). Amsterdam, The Netherlands: North Holland.

71.

Viberg

. (2001). Age-related and L2-related features in bilingual narrative development in Sweden. In Verhoeven

Strömqvist

(Eds.), Narrative development in a multilingual context (pp. 87–128). Amsterdam, The Netherlands: John Benjamins.

72.

Wiig

E. H.

Secord

Semel

E. M.

(2004). CELF preschool 2: Clinical evaluation of language fundamentals preschool. San Antonio, TX: Pearson/PsychCorp.