Narrative Assessments with First Grade Spanish-English Emergent Bilinguals: Spontaneous versus Retell Conditions

Abstract

This study used qualitative analyses to investigate similarities and differences in narrative production across two task conditions for four first grade Spanish-English emergent bilingual children. Task conditions were spontaneous story generation and retelling using the same story. Spanish stories from two children were compared on the basis of similarity in vocabulary, while English stories from two children were compared on the basis of similarity in overall discourse skills. Results show that when the total number of words used was similar across English narratives, the retell included more different words and higher quality story structure than the spontaneous story. When overall discourse scores in the Spanish examples were similar, the spontaneous story required more words than the retell, but also included more central events and greater detail. Yet, the retell included more advanced narrative components. This study contributes to our understanding of narrative skills in young Spanish-English bilinguals across task conditions.

Children who speak a language other than English at home are a growing population in U.S. public schools (Goldenberg, 2008; Suárez-Orozco, Suárez-Orozco, & Todorova, 2008). The emergent bilingual population has grown by 24 percent since 2000, and children younger than 8 with at least one parent who speaks a language other than English at home make up nearly one-third of all young children in the U.S. (Park, O’Toole, & Katsiaficas, 2017). Of this population, 62% identify themselves as Hispanic (Park et al., 2017). Spanish-English emergent bilinguals tend to come from low-income families, and their parents tend to have low levels of education, resulting in the well-documented discrepancies between the academic outcomes of these emergent bilinguals and their monolingual English-speaking peers by the time they are in late elementary school.

The literacy development of these emergent bilinguals (García & Kleifgen, 2010) has been studied extensively in recent years. Vocabulary, in particular, has been the focus of much research (August, Carlo, Dressler, & Snow, 2005; Carlo, August, & Snow, 2005). However, less is known about the broader oral language skills that emergent bilinguals exhibit in each language (Gutiérrez-Clellen, 2002; Uccelli & Paéz, 2007). Therefore, researchers have begun to advocate the systematic assessment of oral language, especially in story production and retelling (Fiestas & Peña, 2004; Miller et al., 2006; Muñoz, Gillam, Peña, & Gulley-Faehnle, 2003). Narrative ability is an important precursor to literacy for both monolingual and bilingual children (August & Shanahan, 2006; Dickinson & Sprague, 2001) and is therefore a fruitful area for further investigation (Gutiérrez-Clellen, Restrepo, Bedore, Peña, & Anderson, 2000). Although two methods of narrative elicitation – spontaneous production and retelling – are commonly used to assess emergent bilinguals, little is known about how the conditions differ in terms of performance. In particular, the conditions under which narratives are elicited in the children’s two languages may affect outcomes in ways that have the potential to inform both research and instruction.

Oral Narrative Assessment for Emergent Bilingual Children

Oral narrative assessment has long been used to identify children with language disabilities, but recently has also been used to examine language proficiency among typically developing emergent bilingual children (Bedore, Peña, Gillam, & Tsung-Han, 2010; Gutiérrez-Clellen, 2002; Lucero, 2015; Simon-Cereijido & Gutiérrez-Clellen, 2009). Narrative assessment may be particularly appropriate for such children because it provides a systematic way to examine oral language proficiency in both the first (L1) and second language (L2), and is considered less biased because it taps into both language-specific and language-universal skills (Paradis, Genesee, & Crago, 2011). It has also been found to be developmentally sensitive (Heilmann, Miller, & Nockerts, 2010; Roch, Florit, & Levorato, 2016; Squires et al., 2014; Suggate, Schaughency, & Reese, 2011), and it may be especially informative in the first few years of formal schooling when children’s language abilities are evolving most rapidly (Bohnacker, 2016).

Two types of oral narrative tasks are common: spontaneous story production and retelling (Boudreau, 2008). In the spontaneous condition, children see a sequence of pictures or wordless picture book, then generate the story using their own words. In the retell condition, children listen to a story while viewing a sequence of pictures or wordless picture book, then retell the story (Strong, 1998). Both tasks require the integration of various domains of language alongside broader cognitive literacy skills (Curenton & Justice, 2004).

Narrative tasks can be analyzed at two levels: macrostructure and microstructure. Macrostructure measures overall coherence, as well as the inclusion of story grammar elements (Gagarina et al., 2015; Schachter & Craig, 2013; Terry, Mills, Bingham, Mansour, & Marencin, 2013), providing insight into the child’s understanding of story structure, relationships among events, and character motivations (Bedore et al., 2010; Fiestas & Peña, 2004; Reese, Sparks, & Suggate, 2012).

Macrostructure performance has been found to correlate with, and even predict, reading proficiency among monolingual English speakers. For example, Reese, Suggate, Long, and Schaughency (2010) analyzed the narrative quality of children’s retells – based on the presence or absence of orienting and evaluating elements – in first grade and found that it uniquely predicted oral reading fluency both concurrently and one year later, even after accounting for receptive vocabulary and decoding. Similarly, a large-scale study of Spanish-English bilinguals found a robust relationship between oral narrative retell performance and reading proficiency (Miller et al., 2006). A longitudinal study with Spanish-English and Cantonese-English emergent bilinguals also found that children’s ability to produce a coherent oral narrative in first grade – in either the home language or English – explained some variance in their English reading comprehension one year later (Uchikoshi, Yang, & Liu, 2018; Uchikoshi, Yang, Lohr, & Leung, 2016). In a related study, macrostructure in both English and Spanish was found to improve during the first three years of school, from kindergarten through second grade, but English showed a steeper growth than Spanish for Spanish-English dual language learners in Northern California (Uchikoshi & Yang, 2016).

Microstructure includes linguistic elements like vocabulary and grammar, which have been shown to play a role in the development of reading proficiency (Carlisle, Beeman, Davis, & Spharim, 1999; Lindsey, Manis, & Bailey, 2003; Proctor, August, Snow, & Barr, 2010). Vocabulary, in particular, is an area of concern for bilinguals, and is highly predictive of reading comprehension for such students (Droop & Verhoeven, 2003; Gottardo & Mueller, 2009; Manis, Lindsey, & Bailey, 2004). Past research has shown positive growth in narrative vocabulary breadth and depth among bilinguals, and that shapes of growth are similar across languages from kindergarten through second grade (Rojas & Iglesias, 2012). In terms of grammar, Muñoz and colleagues (2003) found that older Latino preschool children’s narratives were made up of longer, more grammatically correct sentences than those produced by younger children.

Numerous researchers have reported significant relations between the two languages of young bilinguals at the macrostructure level (Gagarina, Klop, Tsimpli, & Walters, 2016; Gámez, Lesaux, & Rizzo, 2015; Iluz-Cohen & Walters, 2012; Laurent, Nicoladis, & Marentette, 2015; Pearson, 2002; Rodina, 2016; Simon-Cereijido & Gutiérrez-Clellen, 2009), suggesting that the ability to tell coherent, complex stories relies on cognitive processes that are common across languages. In contrast, there have been inconsistent findings with regard to cross-linguistic microlevel relations in the early elementary years. Some studies have shown that children exhibit comparable performance in vocabulary across languages (Gámez & González, 2017), with significant relations at first grade, but not at kindergarten (Uccelli & Paéz, 2007). Some narrative research has theorized that microlevel skills are less likely to transfer across languages than macrostructure because of their dependence on two specific linguistic systems (Bedore et al., 2010; Kang, 2012).

Spontaneous Story Generation versus Retelling

Of particular interest in the present study were differences in the demands of two narrative tasks: spontaneous story generation and retelling. Both tasks provide information about children’s understanding of holistic narrative elements alongside linguistic development (Muñoz et al., 2003; Roth, Speece, Cooper, & De La Paz, 1996), and both likely activate story schema (Merritt & Liles, 1989). Nonetheless, the tasks require different cognitive and linguistic skills (Gutiérrez-Clellen, 2002; Lever & Sénéchal, 2011; Roch et al., 2016), and children may perform differently on macrostructure and microstructure depending on the task (Spinillo & Pinto, 1994).

On one side, spontaneous story generation may be more difficult because producing a story based on picture cues alone is cognitively demanding (Liles, 1993; Merritt & Liles, 1987). Children have to conceptualize all possible elements and systematically organize them into a coherent narrative without the benefit of a model to draw on (Ripich & Griffith, 1988; Roch et al., 2016). Furthermore, while children likely understand what makes a good narrative at an early age, they may not apply this knowledge to spontaneously-produced stories (Lever & Sénéchal, 2011). Moreover, they may not have enough vocabulary to communicate their intended message. It seems likely that children may produce less complex stories in spontaneous story generation if they have not reached a certain threshold in vocabulary knowledge. Alternatively, the child may try to explain a word or situation using his or her limited vocabulary knowledge (e.g. saying “the house where bees live” for “beehive”), resulting in a longer story with less complex vocabulary.

However, retelling may be more difficult than spontaneous generation because it relies heavily on linguistic comprehension skills (Roch et al., 2016). In addition, some researchers have argued that children’s memory and attentional resources may be taxed by the need to focus on the linguistic form in which the story is presented (Gutiérrez-Clellen, 2002). Retells may be shorter than spontaneous stories, including only key events while still being better structured. In fact, when retelling, children can prioritize the most salient elements to incorporate into their own version of the story (Lever & Sénéchal, 2011; Roch et al., 2016).

Despite these theoretical arguments, the ways in which performance differs by task among bilinguals have not been well established. At the macrostructure level, some research has shown that retells tend to be better structured and include more sophisticated language. For example, Roch et al. (2016) found that kindergarten and first grade Italian-English bilinguals included more macrostructure elements in the retell task condition than in the spontaneous condition in both languages. Similarly, Kunnari, Välimaa, and Laukkanen-Nevala (2016) found that Finnish-Swedish bilingual 5 and 6 year-olds performed better on retelling in both languages, with the structure and complexity of retells in Finnish (L1) being significantly higher than spontaneously-generated stories, and retell complexity being higher in Swedish (L2). Maviş, Tunçer, and Gagarina (2016) found that Turkish-German bilingual 5 through 7-year-olds told stories that were better structured in the retell condition than the spontaneous condition in Turkish (L1), although differences were not significant.

In contrast, an influential study with Spanish-English bilingual second graders found that children who struggled to retell stories in a structured and coherent way in English (L2) were nonetheless able to spontaneously generate organized stories in the same language (Gutiérrez-Clellen, 2002). In Spanish (L1), there were no significant differences between the task conditions. This suggests that there may be differences within the languages of bilingual children.

As assessment and research tools, both conditions have advantages and disadvantages. The most notable advantage of retelling is that it allows for the assessment of listening comprehension – a key component of reading proficiency (Gough, Hoover, & Peterson, 1996; Hoover & Gough, 1990) – without the confound of decoding (Joshi & Aaron, 2011). In addition, the existence of a script facilitates comparability across transcripts, making them more efficient to transcribe and more reliably scored than spontaneous samples (Merritt & Liles, 1989).

An advantage of the spontaneous condition is that it does not tax children’s memory nor require listening comprehension. However, young children often get confused when spontaneously generating stories. They may inadvertently change the setting, drop characters without explanation, or lack a clear goal and resolution (Merritt & Liles, 1989). This lack of consistency in spontaneous stories makes comparability among children difficult.

The Current Study

The goal of this study was to qualitatively investigate differences in narrative performance across two tasks among first grade Spanish-English emergent bilinguals. This work is important because both tasks are commonly conducted with bilingual children, but little is known about how the conditions differ in terms of language production outcomes. We were specifically interested in relations between vocabulary and macrostructure across conditions given how little is known about this issue.

Two research questions guided this study:

1. What are the similarities and differences on vocabulary performance in English narratives between spontaneous and retold narratives given by two Spanish-English bilingual children who have comparable macrostructure scores?

2. What are the similarities and differences on macrostructure performance in Spanish narratives between spontaneous and retold narratives given by two Spanish-English bilingual children who have comparable vocabulary scores?

There has not been previous research investigating our first research question in English, although past research has shown a higher number of different words in retells than spontaneous stories in Spanish (Kunnari et al., 2016; Lever & Sénéchal, 2011). For the second research question, we expected children to tell better structured stories in the retell condition than in the spontaneous condition in their L1 when they had comparable vocabulary scores, supporting previous research (Kunnari et al., 2016; Lever & Sénéchal, 2011; Maviş et al., 2016; Roch et al., 2016). For both questions, we hypothesized that there may not be a linear relation between vocabulary and macrostructure across conditions. This may be so because in the spontaneous condition, children may not have sufficient vocabulary to tell a coherent story and thus compensate by using additional words to explain those they do not know. In the retell condition, in contrast, they have a language model and may therefore use fewer words to tell stories of similar complexity.

Method and Data Sources

Participants

Four first grade children from Spanish-speaking homes in California and Oregon were the focus of this qualitative analysis. All four children were identified as English language learners based on state English proficiency tests and produced intelligible stories in both languages. The children were from Spanish-speaking homes and were learning English at school. Therefore, they received language exposure in both English and Spanish on a daily basis. These four transcripts were chosen to illustrate the patterns we were interested in. English narratives were compared across task conditions (spontaneous and retell) given comparable macrostructure scores, and Spanish narratives were compared across conditions on the basis of comparable Spanish vocabulary scores.

Comparable English macrostructure scores

Two children with the same macrostructure scores on the tasks were chosen. The spontaneous story was told by a 6 year 3 month old boy. The retell was by a 6 year 5 month old girl.

Comparable Spanish microlevel scores

Two children with comparable microlevel scores on the different tasks were chosen. The spontaneous story was told by a 6 year 8 month old boy. The retell was told by a 6 year 7 month old boy.

Assessment Procedures

Narratives in both conditions were conducted using Frog, Where are You? (Mayer, 1969). Utilizing this story is a popular method of obtaining narratives for school-age children (e.g., Heilmann et al., 2010), and these books have been widely used to assess bilingual heritage-Spanish speakers (Bedore et al., 2010; Miller, Andriacchi, & Nockerts, 2016; Simon-Cereijido & Gutiérrez-Clellen, 2009). Scripts for the books can be found in the Child Language Data Exchange System (CHILDES) (MacWhinney, 2000)) and the Systematic Assessment of Language Transcripts (SALT) (Miller et al., 2016).

Spontaneous procedure

Like previous studies (e.g. Uchikoshi, 2005), children were assessed individually and randomly selected to produce the narrative in L1 or L2 first. The same book was used to elicit L1 and L2 narratives. Narratives were collected on two separate days by trained native Spanish-speaking or English-speaking research assistants. Children were asked to look through the book on their own, looking at each page. Then, when they were ready, they were asked to narrate a story to the assessor while looking at the pictures. During the task, assessors were permitted to only provide backchannel responses or to restate the child’s last utterance. If the child began to describe the picture instead of providing a narrative, the assessor prompted the child with questions such as “What is happening in the story?” or “What happened next?” If the child appeared to be taking no definite course or ended the narrative abruptly, the assessors were instructed to ask, “Is that the end of the story?” or “How does the story end?”

Retell procedure

Retell assessments were administered individually either by the Spanish-English bilingual principal investigator or a trained English-speaking doctoral student (English assessments only). Children were assessed in Spanish first and English approximately a week later, using two frog books. A counterbalanced design was employed, in which half of the children were randomly selected to hear Frog Where are You? in Spanish and half in English. Only results for Frog, Where You? are reported in this article.

Children listened to the story using headphones while the assessor worked in another part of the room. This naïve listener condition typically leads to more detailed retellings because of the lack of shared knowledge between researcher and child (Strong, 1998). Children were given a brief introduction to the story and told they would be asked to retell the story without the aid of pictures. After listening, the child was given the option to review the pictures again before the assessor put the book away. The assessor remained silent throughout the retell, unless the child paused for longer than three seconds. At that point, the assessor gave a general prompt such as “Tell me more” (“Dime más”) or “Anything else?” (“¿Algo más?”) (Miller, Andriacchi, & Nockerts, 2015). Once the child had retold most of the story or paused for longer than five seconds, the researcher asked, “Is that all you remember?” (“¿Es todo lo que recuerdas?”) (Justice et al., 2006).

Coding and Analysis

Across task conditions, narratives were audiorecorded, transcribed, and coded using SALT (Miller et al., 2016) conventions. All narratives were transcribed and verified by two separate English-Spanish bilinguals. Then all narratives were professionally coded by trained SALT transcribers. Narratives were analyzed for story structure at the macrolevel and vocabulary at the microlevel. Story structure was assessed using the Narrative Scoring Scheme (NSS) (Heilmann, Miller, Nockerts, & Dunaway, 2010), which consists of 7 components, and is scored using a combination of discrete coding criteria and qualitative researcher judgment. The story components are: introduction, character development, mental states, referencing, conflict/resolution, cohesion, and conclusion. Each element is scored from 1 (minimal/immature) to 5 (proficient) for a total possible 35 points (for specific scoring criteria, see Heilmann et al. 2010). The NSS is not coded automatically, so transcripts were hand-coded.

Vocabulary was measured in two ways: 1) total number of words (TNW) and 2) number of different words (NDW) (Miller et al., 2006). TNW is a count of all the words used to tell the story (barring maze behavior and repetitions), and is considered a measure of fluency. NDW, in contrast, is a count of only unique, uninflected root words; it is developmentally-sensitive and considered a robust indicator of the breadth of a child’s vocabulary (Heilmann, Miller, Nockerts, et al., 2010). Both measures were calculated using SALT.

Findings

Narratives with comparable macrostructure scores

The first pair of narratives highlights possible differences in TNW scores between two English stories with the same NSS score of 16. The spontaneous had an EngTNW of 449, while the retell had an EngTNW of just 78. Therefore, these two children used vastly different numbers of words to tell stories of comparable coherence and complexity.

The spontaneous story included more central events, and used more details when describing those events. While the retell included only three main events, the spontaneous story included most of the events shown in the pictures, including intermediate events (see Table 1). For example, the retell mentioned the dog and boy looking “into the hole in the ground. But a gopher hit his nose when it came back up.” This minor event was not mentioned in the retell. These differences were reflected in the scores for conflict resolution – 1 for the retell and 3 for the spontaneous story.

Table 1.

Comparison of English transcripts with similar NSS scores

	Spontaneous (Child #1137)	Retell (Child #10)
TNW/NDW	449/129	78/37
NSS total	16	16
Introduction	1 “The dog and the kid were watching the frog in the jar.” (inappropriate use of definite articles)	3 “First they was a boy a dog and a frog.” (correct use of indefinite articles)
Character Development	2 introduced main & supporting characters, but no use of first-person voice	2 introduced main but not supporting characters, did use first-person voice
Mental States	2 “the dog thought the hive was a piñata.” (some use of mental state verbs)	4 “the boy was mad.” (some use of mental state verbs)
Referencing	3 “And then the dog stick his head into the jar. But he had trouble taking it off. But the children was yelling to find him.” (inconsistent referencing)	3 “The dog did go in the jar. Then the dog fell down. Then the jar broked.” (some avoidance of pronouns)
Conflict Resolution	3 “So the children looked into the hole into the ground. But a gopher hit his nose when it came back up. And then he made the hive drop.” (most events included, described in some detail)	1 “Then the boy was mad. Then he saw a family *of frogs. The boy said that if he can have the frog.” (few events included, described in limited detail)
Cohesion	2 “But the jar was made out of glass so it broke. They were looking. They were went to the forest.” (lack of clear connections among events)	2 “Then the boy was mad. Then the dog, he saw a family *of frogs.” (lack of clear connections among events)
Conclusion	3 “And then it was time to the frog to go back home. But there was one frog that was sad to leave his frog friend, get back home. Then end.” (specific event concluded, but no general conclusion)	1 “He saw a family *of frogs. The boy said that if he can have the frog, and the frog said yes.” (somewhat abrupt & incomplete ending)

Performance on other NSS elements was also different. For example, the spontaneous narrative scored higher on conclusion by providing a specific event to end the story, by saying “it was time (for) the frog to go back home.” The retell ended abruptly with the boy asking if he could have the frog.

In contrast, the retold narrative had higher scores for introduction and mental states, both critical for the development of narrative comprehension (Barnes, Kim, & Phillips, 2014; Reese et al., 2010). The retell appropriately used indefinite articles (e.g., a boy, a dog) to introduce characters, whereas the spontaneous narrative inappropriately used definite articles (e.g., the dog, the kid) as if the listener already had knowledge of the characters. Research has suggested that the quality of introductions in narratives can play an important role in comprehension; Barnes et al. (2014) found that among four- to seven-year-old English monolinguals, proper character introductions added significant variance to listening comprehension. Furthermore, well-formed narratives include evaluation devices, such as use of mental state words, to signal that the story is being told from the narrator’s perspective (Chang, 2004; Labov, 1972). Older children use a larger variety of evaluation devices than younger children, showing a developmental trend (Peterson & McCabe, 1983).

In terms of what was similar, the stories had comparable levels of referential and lexical cohesion. Both showed inconsistent referencing that sometimes made it difficult to identify the subject of the clause. In addition, both children struggled to provide clear connections among events, similar to the narrative structures of 4 to 5-year-old monolingual English-speaking children when providing a simple recounting of successive events (Peterson & McCabe, 1983).

Narratives with comparable vocabulary scores

The second pair of narratives illustrates the differences in NSS scores between two Spanish narratives with similar TNW scores. The overall quantity of vocabulary used in both stories (spontaneous TNW = 207; retell TNW = 209) was almost identical. However, the NSS scores diverged (spontaneous SpNSS = 17, retell SpNSS = 21), suggesting that the two children used the same number of words to tell stories of different coherence and complexity (see Table 2).

Table 2.

Comparison of Spanish transcripts with similar TNW scores

	Spontaneous (Child #1013)	Retell (Child #58)
TNW/NDW	207/71	209/94
NSS total	17	21
Introduction	3 “Primero él estaba en la casa. Después estaban mirando la rana. Y después él *se durmió.” [First he was in the house. Then they were looking the frog. And then he slept.] (introduces general setting but no main characters)	3 “Que había un niño con una rana. Había un perro con un niño. Y tenían una rana.” [That there was a boy with a frog. There was a dog with a boy. And they had a frog] (introduces main characters but no general setting)
Character Development	2 mentioned main but not supporting characters, with limited use of first-person voice	3 both main and active supporting characters are mentioned, with limited use of first-person voice
Mental States	2 “el niño estaba muy enojada.” [the boy was very angry] (only one use of a mental state verb)	2 “El niño ni se dio cuenta *de que el perro había tirado donde viven las abejas.” [The boy didn’t realize that the dog had knocked down where the bees live.] (only one use of a mental state verb)
Referencing	3 “Después X se salió. Después él estaba buscando. Y no la encontró.” [Then he left. Then he was looking for *it. And he didn’t find it.] (inconsistent referencing)	4 “Y la ventana estaba abierta. Entonces estaban durmiendo el niño y el perro. Y después la rana se escapó.” [And the window was open. So they were sleeping the boy and the dog. And then the frog escaped.] (most references are clear throughout the story)
Conflict Resolution	2 “Y después el perro corrió. Después él dijo “shh”. Y después X allá arriba de allí.” [And then the dog ran. Then he said “shh”. And then X there above there.] (some confusion about how events happened)	3 “Y después miraron. Y había una rana mujer y su rana. Y tenían muchos bebés. Y después una rana bebé se brincó.” [And then they looked. And there was a mother frog and a father frog. And they had many babies. And then a baby frog jumped up.] (some detail in describing events)
Cohesion	3 “Y después iba corriendo a mirar en el árbol. Después miró en el hoyo y no estaba. Después miró allá arriba donde están las abejas y no estaba.” [And then he went running to look in the tree. Then he looked in the hole and it wasn’t there. Then he looked over there up where the bees were and it wasn’t there.] (events follow a logical order)	3 “Y después se subió a una piedra, y se agarró sobre unos palos para que no se cayera. Pero no eran palos; sino que era unos cuernos de un venado.” [And then he climbed on a rock, and he grabbed one of the branches so that he wouldn’t fall. But they weren’t branches; instead they were the antlers of a deer.] (events follow a logical order)
Conclusion	2 “Y después miró las dos ranas y sus bebés. <Examiner: y aquí?> Y después le dijo “bye, ranas”. The end.” [And then he saw the two frogs and their babies. <Examiner: And here?> And then he said “bye, frogs”. The end.] (specific event concluded only with examiner prompting)	3 “Y después una rana bebé se brincó. Y después se fue con el niño. Y les dijo “bye” a toda la familia.” [And then a baby frog jumped up. And then it went with the boy. And it said “bye” to the whole family.] (specific event concluded, but no general conclusion)

The retold story had higher scores on most elements than the spontaneous story did. However, the retell did not score more than one point higher on any individual element, and they scored the same on three elements. The retell earned more points for character development, referencing, conflict resolution, and conclusion. Thus, it included more events that were thoroughly described, along with more consistent, accurate antecedents and clarifiers. As in the set of English examples, the retell introduced the characters using appropriate referencing, while the spontaneous story started with él (he) at the outset, without introducing the main character first. This is another example of the presumption of shared knowledge that can disorient a listener at the beginning of a story (Barnes et al., 2014). In terms of conclusion, the retell provided a specific event to end the story, whereas the child telling the spontaneous story ended abruptly and only provided an ending after being prompted by the assessor.

An interesting intersection between story structure and vocabulary can also be illustrated using these two transcripts. Although the narratives had very similar numbers of total words, the number of different words used by each child diverged – 71 for the spontaneous story and 94 for the retell. Indeed, the child telling the spontaneous story struggled to produce the vocabulary needed to continue the story. Twice, he said “que yo no sé” for animal names he did not know. The child retelling the story could not remember the word buho (owl) either, but he paused rather than telling the assessor he did not know the word. This highlights two points: first, it presents the possibility that the different strategies children use to tell stories can be understood by a combination of quantitative and qualitative analysis; second, it underscores the importance of attending to multiple measures of vocabulary within oral narratives, as vocabulary likely influences overall narrative proficiency (Lucero, 2015).

Discussion

The present study was undertaken to investigate differences in performance on related narrative tasks. This investigation is worth conducting because these two methods of narrative elicitation are commonly used to assess emergent bilinguals, but little is known about how the conditions differ in terms of performance on macrolevel and microlevel skills. Our results add nuance to existing research because they show that the tasks may highlight different strengths and provide varying levels of scaffolding, which may or may not be needed in each language.

Results from the two stories with the same macrostructure score show that the spontaneous story had more than five times the number of words and contained more central events and details than the retell. Gutiérrez-Clellen (2002) had a similar finding: Spanish-English bilingual second graders produced more story elements in spontaneously-generated stories in English (L2). This may be because children telling the spontaneous narratives were allowed to look at the pictures as they gave their narratives. Had children in the retell condition been given pictures to remind them of the storyline, their narratives might have included a similar number of intermediate events. We did not have a measure of working memory, so memory effects are not knowable. However, such possibilities need to be examined in future research.

At the same time, we found that the retell included more developed introduction and mental states. These two components tend to be developmental, such that older children typically score higher than younger children (Peterson & McCabe, 1983). The introduction launches the narrative and is an important component, whereas the use of mental state words helps the listener understand character emotions and thought processes. This suggests that emergent bilinguals may benefit from more instruction and modeling of story introductions and discussion on character emotions and their thought processes. Especially with this age group, this could be tied into a socio-emotional development curriculum wherein children learn and discuss mental state words, both in English and Spanish.

In terms of similarities, both stories had comparable levels of inconsistent referencing. This finding is in line with other research, which has shown that sophisticated referencing skill remains challenging for children in the first two years of school (Barnes et al., 2014; Mäkinen, Loukusa, Nieminen, Leinonen, & Kunnari, 2014).

Second, results from the two stories with comparable vocabulary scores support our hypothesis that children would perform better on story structure in the retell condition when the total number of words used was similar. This finding supports previous research investigating this issue with bilingual children (Kunnari et al., 2016; Maviş et al., 2016; Roch et al., 2016). However, it should be noted that although the total number of words was similar, the retell used 94 different words and the spontaneous only 71 words, highlighting the importance of examining multiple measures of vocabulary within oral narratives.

A strength of this study was the use of the same elicitation material across conditions. However, a limitation is that we only analyzed the retells of four children. Our goal in doing such an in-depth analysis of a small number of narratives was as a first step in a larger qualitative examination of spontaneous narratives and retells. Future studies should examine larger samples, and consider other measures of language proficiency. Another limitation is that we compared performance on only one story. Future research should investigate the performance of emergent bilinguals on both tasks to compare performance at the individual student level. Future research might also consider controlling for vocabulary using standardized tools, and investigate the role of using pictures while the child is retelling the story. Different books should also be used to see if results change. Such research would undoubtedly enrich our understanding of the demands and affordances of both tasks.

This study was exploratory and therefore far from definitive. Nonetheless, the findings contribute to the growing field of oral narrative research, adding nuance to what is already known, and charting a course for continuing to explore what is not. Vocabulary is often the focus of instruction for bilingual children, who may be seen as having a deficit in this area. However, our findings highlight the critical importance of maintaining a focus on more holistic conceptions of language. They also provide evidence for the importance of read-alouds, storytelling, and explicit story structure instruction that goes beyond isolated vocabulary development (Boudreau, 2008; Consortium, 2015). Various oral language skills are important to overall literacy development, and instructional and assessment practices should reflect this. Decisions about which method to use under what circumstances should be based on an understanding of how performance is likely to be affected by the chosen task. The tasks can also be used in concert to assess a given child, with each condition providing information about strengths and needs related to oral narrative skills.