The Development of Discourse and Morphological Features in L2 Narratives: A Study with Classroom Spanish-Speaking Learners of French

: Previous research has examined the organization of second-language French narratives through discourse or morphological analyses. At the discourse level, the analyses have investigated the foreground/background relationship. Conversely, at the morphological level, the analyses have examined the role played by verbal morphology and verbal predicates. Different methodological caveats have limited the interpretation of ﬁndings in both types of analyses. In order to provide new data, this cross-sectional study examined the evolution of discourse and morphological resources in the written narratives of Mexican Spanish-speaking learners of French whose language learning experience is limited to the classroom. The learners in the cross-sectional sample ( n = 11) were selected from a population of 88 participants who completed lexical, past-tense and general proﬁciency tests. They also generated two written narratives during silent-ﬁlm retelling tasks. The cross-sectional sample selection was based on the learners’ test scores and the results of parametric statistical analyses. The narratives were analyzed for the identiﬁcation of foreground/background clauses, verbal morphology and verbal predicates. The results reveal that, as learners’ past-tense knowledge increases, the organization of narratives consolidates through a developmental path that involves the interrelated growth of discourse and morphological features.


Introduction
The production of a narrative is a linguistically complex task. Various researchers have examined how second-language (L2) learners develop the ability to concatenate discourse and morphological elements to create coherent narratives. In the case of the acquisition of L2 temporality, at the discourse level, some studies have examined the organization of narrative events through analyses of the foreground/background opposition (Klein and Perdue 1997;Noyau et al. 2005;Comajoan 2013). Conversely, morphological studies have examined the development of verbal morphology with predicates that carry distinctive inherent semantic properties (Bardovi-Harlig and Comajoan 2020). Despite the increasing number of discourse and morphological studies, which have independently examined the organization of L2 narratives, methodological caveats hinder the conclusiveness of the results (Bardovi-Harlig and Comajoan 2020). Different from previous studies, this study examines the interrelated development of discourse and morphological features in the L2 narratives of instructed learners of French using a functionalist (meaning-oriented) approach. This approach, used in the European Science Foundation project (Klein and Perdue 1997;Bardovi-Harlig 1998), allows for the examination of L2 learners' narratives through "function-to-form" analyses. During these analyses, all the linguistic devices that learners use in order to express a particular concept, such as time, are examined. Moreover, the data were collected through participant-selection and dataelicitation procedures which aim to overcome methodological caveats that have been identified in discourse and morphological research. Furthermore, through a cross-sectional design, the study provides evidence on the stages that may characterize the development of L2 narratives among instructed learners of French.

L2 Discourse and Morphological Research
In the study of L2 narratives from a discourse perspective, some authors consider that the foreground/background distinction is a universal feature of narratives (Dahl 1984;Hopper 1979). For others, the foreground/background opposition is rather a continuum with an interplay of semantic and grammatical oppositions where some events are more salient than others (Fleischmann 1990). Despite these divergent viewpoints, researchers agree that the sequentiality of events constitutes a key element of the foreground. To identify this sequentiality, Klein and Stutterheim (2002) put forward a narrative analysis model that centers upon the Quaestio. This model allows for the identification of foreground clauses, which express the referential movement of time between events. The background includes propositions where the advancement of time is momentarily ceased. Background clauses create a rich context to support the plot through comments, evaluations and descriptions of the story (Hopper 1979). Based on the Quaestio model, researchers in the meaning-oriented approach examine the linguistic devices, beyond the verbal predicate, that are used by the learners to express temporal relations in discourse (Klein and Perdue 1997;Noyau et al. 2005). To this end, the number of participants remains low, as the analyses of a small number of narratives allow for an in-depth identification, for instance, of the lexical, pragmatic, and morphological features which enable the learner to create coherent narratives.
Researchers within the discourse perspective acknowledge that the foreground and background are materialized differently across languages (Klein and Perdue 1997;Noyau 2002). In French and Spanish, the L2 and first language (L1), respectively, of the participants in this study, the foreground/background opposition, can be rendered through verbal morphology.
Perfective morphology typically appears in the foreground and imperfective morphology in the background. In French, the passé composé and the passé simple, and in Spanish, the preterit, encode the perfective past. These forms typically indicate completed events that make the narration move forward and are thus in the foreground. French and Spanish both have an imperfective past form which indicates the momentary ceasing of temporal movement in narratives. The imperfective form typically appears in the background. While these two languages are typologically close, there are substantial form-meaning mapping differences between them with respect to their past-tense morphology, as the following sections illustrate. Therefore, the facilitative effects of L1-L2 typological proximity are not clear-cut for Spanish-speaking learners of French (Izquierdo and Collins 2008;Izquierdo 2009Izquierdo , 2014Izquierdo and Kihlstedt 2019). At the form-meaning level, Spanish and French express past characterization, habituality and progressivity through imperfective morphology (Izquierdo and Kihlstedt 2019). Additionally, Spanish has past progressive morphology, which is a compound form that requires an auxiliar verb (e.g., estar 'be', seguir 'continue') in the past and the -ando/-iendo GERUND ending attached to the main verb. Although progressivity is the central function of the Spanish progressive, e.g., cuando estaba checando su correo (while he be-IMPERFECTIVE check-PROG his mail, while he was checking his mail'), L1 Spanish speakers tend to opt for the imperfect to express progressivity in written narratives 'cuando checaba su correo' (while he check-IMPERFECTIVE his mail) (Izquierdo and Kihlstedt 2019). In French, progressivity can only be rendered through the imperfect. It can optionally be conveyed by the periphrasis en train de + infinitive verb (in the process of): 'il était en train de travailler' (He be-IMPERFECTIVE in the process of working). In contrast to French, Spanish may express habituality using a periphrastic construction with the defective verb soler (to usu-ally do), which is inflected with imperfective morphology + verb in infinitive, as in Julián solía-IMPERFECTIVE comer manzanas (Julian use-IMPERFECTIVE to eat apples).
Various studies highlight other form-meaning mapping differences between the French and Spanish imperfective (De Lorenzo 2002;Noyau et al. 2005). These studies reveal that the imperfective shows a more systematic use in the background of a narrative in Spanish than in French. In addition, the imperfective forms occur more frequently in L1 Spanish than in French written narratives; the Spanish imperfect has a higher lexical diversity of verbs than its French counterpart, which is strongly restricted to state verbs (Izquierdo and Kihlstedt 2019). This robust use of the Spanish imperfective has also been documented by Daidone (2019), who found that the preterit dominates past-tense use in L2 teacher discourse, whereas the imperfect dominates past-tense use in L1 Spanish corpus data.
The perfective form also differs between French and Spanish at the functional and morphological level. In French, perfectivity is rendered through the passé composé, a compound form (auxiliary + past participle), which has taken over the perfective function of the passé simple in colloquial discourse. The passé simple is used in literary texts and in tales. In Spanish, however, the perfective aspect is rendered through a set of verbal inflections. The indefinite preterit expresses the perfective function in both colloquial and literary texts. Spanish also has a compound past form which is somewhat equivalent to the passé composé at the morphological level. This compound form renders the perfect function that expresses events which take place in an unidentified point in time between the past and the present. It should be noted that the Spanish compound present perfect is becoming the default form of past perfective in temporally indeterminate past contexts in Peninsular Spanish (Schwenter and Cacoullos 2008). Nonetheless, in Latin-American Spanish, this use of the compound present perfect is less frequent, as in the case of the participants of this study who were Mexican-Spanish native speakers (see also Izquierdo and Kihlstedt 2019). In French, the passé composé is used for both the present perfect and the perfective function. To avoid confusion at the form-meaning level, the French term passé composé will be used, as it is not always clear which function is used by the learners.
In morphological research, based on the Lexical Aspect Hypothesis (LAH), various studies have examined whether the inherent meaning of verbs, or rather the verbal predicate, (cf. the semantic difference between build and build a house) influences the use of grammatical aspect during the production of L2 narratives (Bardovi-Harlig and Comajoan 2020). The predictions of the LAH (Andersen and Shirai 1994), built upon consideration of the four aspectual verb classes of Vendler, state that low proficiency learners prefer the use of past perfective morphology with telic verbs, such as accomplishments and achievements; these verbs hold an inherent terminal point, e.g., arrive, leave, build a house. Imperfective morphology (the imperfect and the progressive) emerges later and is initially associated with atelic predicates (states and activities) which do not carry an inherent temporal boundary, such as love, like, sing and run. According to the LAH, the endpoint of acquisition is reached when the learner is able to produce non-prototypical combinations; that is, when learners mark atelic verbs with perfective forms and telic verbs with imperfective morphology and have thus overridden the influence of verb semantics.
Some LAH studies have relied on the use of narratives to test the LAH predictions, but they do not consider the foreground/background opposition. Instead, their analyses center upon the use of verbal morphology across verb types. However, the potential interplay between verb semantics and the foreground/background opposition has led some researchers to develop the Discourse Hypothesis (DH). This hypothesis predicts that "learners use emerging verbal morphology to distinguish foreground from background in narratives" rather than combining semantically compatible verbs and grammatical forms (Bardovi-Harlig 2000, p. 279). This implies that, initially, the perfective is limited to telic verbs in the foreground and the imperfective appears with atelic verbs in the background. At first sight, there is a clear resemblance between the two hypotheses. Nonetheless, Bardovi-Harlig (1998) indicates that the two hypotheses can be distinguished theoretically and empirically. For instance, when telic verbs are inflected for past perfective, regardless of grounding, this use would support the LAH. However, if foreground clauses display past perfective morphology with atelic verbs, this finding would support the DH.
Only a few studies have analyzed L2 narratives considering the interplay between the foreground/background, perfective/imperfective and telic/atelic oppositions. According to Bardovi-Harlig (1998) and Salaberry (2011), the reason for this is that the criteria that distinguish foreground from background overlap with the inherent semantics of verb types. Thus, foreground events typically attract telic verbs, whose inherent semantics imply completeness and boundness (e.g., achievements and accomplishments). Due to their discourse function, background events are unbounded and typically attract activities and states. While some studies that have considered these discourse and morphological features reveal that grounding may override telicity, others indicate that grounding may operate with telicity even among advanced L2 learners (Comajoan 2005). Furthermore, Salaberry (2011) found that the prototypical association between lexical aspect and verbal morphology in L2 narratives tends to increase with learner proficiency. These findings opposed the claims of the LAH, where the prototypical past-tense use should be overridden at advanced stages of acquisition. Salaberry explains this fact indicating that a very broad level of contextualization is needed to create the need of non-prototypical associations, particularly in L2 narratives.

L2 Narrative Research: Methodological Challenges
Narratives allow for the examination of comparable L2 discursive and morphological features between learners (Bardovi-Harlig 2013; Izquierdo and Kihlstedt 2019), as they could elicit homogeneous content across all the participants. Nonetheless, questions arise with respect to previous findings related to the development of L2 discursive and morphological features in learners' narratives. This is because narrative studies often exhibit some methodological caveats that hinder the conclusiveness of the results beyond the student sample (Bardovi-Harlig 2000; Bardovi-Harlig and Comajoan 2020).
One methodological caveat in discourse and morphological research relates to the narrative production modality itself. In L2 French, most studies have elicited narrative data through oral production tasks (Dietrich et al. 1995;Bergström 1997;De Lorenzo 2002;Noyau 2002). Sometimes, oral narratives are inserted in interviews and focus on learners' personal experiences, where the open-ended nature of personal narratives hinders the comparability between learners (Bardovi-Harlig 2013). Another type of oral narrative elicitation task relies on film-retelling where the learner orally narrates the content of a short film. In this impersonal narrative task, spontaneous oral production can put students in cognitive overload, as they need to remember the story while they mobilize linguistic resources without interlocutor support (Izquierdo and Kihlstedt 2019). Furthermore, the personal/impersonal narrative distinction brings about another challenge. Personal narratives typically show greater background elaboration than impersonal narratives. The latter increases comparability but "tends to sacrifice background information and hence eliminates the environment for imperfective" (Bardovi-Harlig and Comajoan 2020, p. 19; see also Salaberry 1999Salaberry , 2011Noyau 2002;Comajoan 2005;Camps 2005). In impersonal narratives, background events can be elicited through prompts, such as "every morning . . . " in order to set an imperfective habitual frame. The foreground can be prompted through hints such as "until one day", to create obligatory contexts for a sequence of events (Bardovi-Harlig and Comajoan 2020). However, these prompts interfere with the free use of the foreground/background opposition in L2 narratives.
A second caveat relates to the objective identification of the L2 abilities of the learners in discourse-oriented L2 research. For instance, previous studies have considered the number of instructional hours or prerequisite L2 levels of an academic program. Other studies have established the learners' L2 level through qualitative parameters, such as self-assessment, biographical data about the L2 learning experience and the learners' curriculum. Some studies have looked at the lexical threshold of the students. Lexical measurements set the basis for the establishment of different proficiency levels among participants who are then required to produce a narrative. As Meara and Miralpeix (2017) argue, the number of words a learner knows is a reliable indicator of overall level. However, this criterion alone does not necessarily correlate with narrative discourse. Learners may have a personal style to construct a narrative or may not know the cohesion and grammatical devices of narratives. Therefore, these criteria can make the conclusiveness of results difficult.
In L2 French research, a third consideration relates to the learning conditions of the participants. In previous discourse-oriented research, the participants are often foreign students and migrants in a francophone country, who come from various L1s and different L2 learning conditions. Their exposure to the L2, thus, takes place in instructed and uninstructed contexts. In addition to variation in the amount of input, questions arise with respect to the type of L2 input they have received. While L2 migrants may be exposed to colloquial input at the workplace or within the L2 community, learners in school settings receive extensive exposure to academic aspects of the L2. This variation in the L2 learning conditions of the learners could hamper the reliability of the narrative data.
Based on the aforementioned issues, through a cross-sectional design, this study addressed this research question: What are the discourse and morphological features that characterize the production of L2 written narratives among Spanish-speaking classroom learners of French at different levels of proficiency?

Materials and Methods
To answer the question, this cross-sectional study was conducted in compliance with the following methodological considerations. First, narrative production was prompted in writing through two impersonal film-retelling tasks administered in the L2 classrooms of the participants. This type of task allows learners to have access to their discourse and linguistic resources and provides them with time to retrieve content (Skehan and Foster 1999). Second, to control for L2 exposure, the French learning experience of our participants is limited to university classrooms in Mexico where their current and previous contact with French outside of the L2 program is almost non-existent. In this context, the amount and purpose of L2 exposure is determined by a pre-established school curriculum, institutionalized instructional materials and standardized pedagogical practices. Third, the L2 proficiency of the participants was examined through three tests which focused on different domains of the L2. Fourth, the L2 level of the learners in the cross-sectional sample was identified through systematic score differences that were established through parametric analyses. Finally, the developmental description of the discourse and morphological features of the L2 narratives emerged from comparisons of grounding, verbal morphology and telicity features of verbal predicates in the texts of the learners in the cross-sectional sample.

Context and Participants
The study was conducted with 88 students enrolled in a BA in Language Studies in a public university in the Southeast of Mexico. In this program, students complete nine levels of French. Each level is comprised of 16 study weeks in addition to a midterm and a final examination week. Each level includes 96 instructional hours. The L2 lessons can be classified as structure-based, as the syllabus is organized around L2 grammar and communicative functions, following the Common European Framework. The teachers implement form-and meaning-based activities to cover the L2 syllabus. To comply with ethical regulations, the researchers contacted the school dean and the teachers of the French department to discuss the study's objective, instruments and ethical principles. Only the teacher from level 8 did not consent to participation in the study. The dean and the other teachers granted the researchers access to the classrooms and assisted them with the organization of the research agenda.
During the first classroom visit, the researchers explained the study timeline to the students and informed them that their participation was voluntary and without remuneration. Initially, 273 participants consented to participate in the study. However, only 62 female and 26 male students were considered, as they completed all the tests, written narratives and the consent form. These students were enrolled in level four (n = 20), five (n = 18), six (n = 13), seven (n = 19) and nine (n = 18). Students below level four were excluded, as they had not received instruction on the perfective/imperfective contrast. Their background information was collected through a 10-question sociodemographic questionnaire which revealed that their ages ranged between 22 and 26 years old. The learners were native speakers of Mexican Spanish. They had all learned French in the BA program only and had never been to a French-speaking country. They were learning English in their BA and an additional foreign language (e.g., German, Italian, Japanese, Portuguese) at the university language center. From these learners, three male and eight female participants were selected for the cross-sectional sample following the statistical criteria indicated in Sections 4.2 and 4.3.

Proficiency Tests
To assess the knowledge of the learners in various L2 domains, they completed three independent tests that examined their general L2 proficiency, lexical, and past-tense knowledge.

General Proficiency Test
This instrument was developed at the participating university as a placement examination and includes 35 multiple-choice items which test the use of various grammatical features, such as tense, gender agreement and articles. For each correct answer, participants received one point; thus, their score could range from 0 to 35. This test was validated in a previous study using parametric analyses with intermediate university learners of French in Mexico and Canada, yielding reliable results, t(30) = 0.022, p = 0.98 (Izquierdo and Collins 2008). In addition to this validation procedure, the test was validated in the current study as indicated in Section 3.2.4.

Lexical Test
This instrument was based on an automated word-recognition test available via the Compleat Lexical Tutor (https://www.lextutor.ca/tests/yes_no_fr/FF101frame.html accessed on 26 November 2018). The test serves to calculate a learner's L2 vocabulary size using the 1000 most frequent words in French. The test presents 60 items that randomly include 40 real and 20 imaginary French words. For each item, learners select oui (yes) or non (no) to indicate whether they do or do not know the word (Meara and Buxton 1987). For the current study, the automated test was adapted into a paper version respecting the original organization of the lexical items. The learners circled oui or non. During the scoring procedures, the learners received one point for each real word that they acknowledged knowing and one point for each imaginary word they acknowledged not knowing. In the current study, the test was validated as indicated in Section 3.2.4.

Past-Tense Test
This test has been administered to learners of French from different L1 backgrounds in previous studies that explore the effects of verb semantics on the use of past-tense morphology (e.g., Izquierdo 2009Izquierdo , 2014. It includes a series of independent cloze passages that elicit perfective and imperfective morphology in an equal number of obligatory contexts (28 for each), in addition to distractor contexts for the use of present and future verbal morphology (n = 18). In this study, the perfective and imperfective contexts were scored. The participants received one point if they attempted the use of perfective or imperfective in the correct obligatory context, irrespective of morphological inaccuracies; thus, their score could range from 0 to 56. In a previous study, this test was piloted with two control groups of native speakers of Canadian and European French. The 56 obligatory past-tense contexts considered in this study yielded 100% agreement (Izquierdo and Collins 2008). In the current study, the test was also validated as indicated in Section 3.2.4.

Test Validation Procedures
The three tests were subject to parallel-form validation. Hence, two versions of each test were created: Version A and Version B. Both versions included the same items but in reverse order. With the initial study population, equal numbers of versions (50%) were randomly distributed within each class. After the participant exclusion stage, an uneven version distribution was observed in the proficiency (A = 43; B = 45), lexical (A = 49; B = 39) and past-tense (A = 49; B = 39) tests. As the Kolmogorov-Smirnov tests indicated that the scores followed a normal distribution in the proficiency, D(88) = 0.075, p = 0.2, lexical, D(88) = 0.63, p = 0.2, and past-tense tests, D(88) = 0.092, p = 0.066, parametric analyses were run for the identification of significant differences between test versions. One-way ANOVAs revealed that the participants had equally performed in both versions of the proficiency, F(1) = 0.24, p = 0.88, lexical F(88) = 2.18, p = 0.14, and past-tense test F(1) = 0.54, p = 0.46. Thus, the version results were pooled.

Narratives
The use of discourse and morphological features in the L2 narratives was elicited through the production of two independent written texts using silent-film retelling tasks.

Narrative Elicitation Tasks
In class and under teacher guidance, the students watched the silent films Father and Daughter (F&D; Jennings et al. 2000) and Carl and Ellie (C&E; Rivera and Docter 2009) and generated their written narratives without dictionary use or teacher assistance. The teachers received a PowerPoint file with instructions and click-on features for the film projection, as a means to homogenize the task implementation across classrooms. These two five-minute silent films were selected, as they scaffold the production of narratives that include a clear story plot, beginning and end. In these narratives, L1 and L2 speakers produce present and past-tense morphology with a wide variety of telic and atelic predicates. Moreover, the use of both films leads to the generation of narratives that complement each other at the functional level, as F&D prompts the expression of habituality, whereas C&E motivates characterization and progressivity (Izquierdo and Kihlstedt 2019).
In Izquierdo and Kihlstedt (2019), these tasks were used with Spanish-speaking learners of French and native speakers of French and Spanish. The results revealed that the written narratives from both L1 groups diverged both at the morphological and functional level. The results also indicated that even the narratives of the most advanced Spanish-speaking learners of French showed L1 influence and, thus, systematically diverged from the L1 French narratives. These results, then, confirmed that L1 influence operates in written narratives and in cloze tests (see Izquierdo 2009Izquierdo , 2014Izquierdo and Collins 2008). Due to this evidence, the current study excluded the use of L1 data and focused on the evolution of the discourse and morphological elements of the written narratives of Spanish-speaking learners of French as their L2 proficiency increased.

Clause Identification
Following procedures in previous studies, the clauses in the narratives were first identified. The clauses consisted of "utterances or fragments of language that carry a single message, and therefore, constitute an independent semantic unit" (Izquierdo et al. 2016, p. 23). Each clause was then subject to discourse and morphological analyses.

Foreground/Background
To examine the foreground/background opposition, Klein and Stutterheim's (2002) model was considered, as it facilitates the analyses of referential movement in narrative discourse. The authors argue that the events in a narrative can be examined through an underlying question, the Quaestio which can be summarized as follows: what happened to p.
[protagonist] at t1 + 1 [just after the last referential time, i.e., next]? This question served to identify the way in which the referential movement of time was realized in the text.
All clauses that provided an answer to the Quaestio were classified as foreground clauses, whereas all the others were considered as background clauses. In this phase of the analyses, the clauses were examined independently of verbal forms to avoid circularity. This is because the foreground or background clauses can be concatenated and can carry different verbal forms as the two foreground clauses in Ils se sont mariés et ils décident d'avoir un bébé (They get married-PERFECTIVE and decide-PRESENT to have a child.)

Verbal Morphology
During the analyses, any attempt of an inflected form was identified and coded. The main identified forms were the imperfect, passé composé and the present. Other forms were the past participle without the auxiliary as in il tombé (he fall-PAST PARTICIPLE), which were categorized as a perfective form. All Spanish words, spelling and inflection errors were neglected if the form could still be identified. Incomprehensible forms, such as elle ne se reude pas avec le port (learner 243) 'She 'non-identified' not with the harbour,'were classified as "other".

Telicity
The clauses were further examined to identify the telic properties of their verbal predicates. Following DH studies (Bardovi-Harlig 2000; Salaberry 2011), the classification focused on the atelic/telic opposition. Based on the verbal typology of Vendler, it was considered that verbal predicates with an inherent end hold telic properties, whereas verbal predicates without an inherent end constitute atelic forms. Thereafter, each verbal predicate was subject to the question: does the event in the verbal predicate imply an indefinite duration? Predicates that elicited a positive answer were identified as atelic. Otherwise, they were coded as telic (see also Izquierdo and Kihlstedt 2019). It is well known that atelic verbs attract the imperfective and telic verbs the perfective as in this sentence from learner 99: Elle travaillait/IMPERFECTIVE dans un zoo "She work-IMPERFECTIVE in a zoo". Thereafter, the verbal predicates were analyzed considering the infinitive form of the verb in the predicate (e.g., to work in a zoo). This procedure enabled us to avoid influence from verbal morphology during the identification of temporal bounderies in the predicates.

Coding Validation
The coding was validated through inter-rater analysis procedures. Each author independently coded 126 clauses in the written narrative F&D and 215 clauses in the narrative C&E, which were produced by the students in the cross-sectional sample. They then crosschecked their codification results. Differences were found in 6 clauses in F&D and 7 clauses in C&E. These differences were sorted out through discussion and consensus.

Data Collection Timeline
As per the request of the teachers, the researchers visited each class on six different occasions during the first 30 min of the lesson. During session 1, the participants completed the background questionnaire and consent form. During sessions 2 through 4, a different proficiency test was administered each time. The narratives were produced in Sessions 5 and 6.

Results
This section first provides a description of the participants' results in the three tests and the procedures for the selection of the participants whose narratives would undergo further analyses. Then, the results of the discourse and morphological analyses of the L2 narratives are presented.

Proficiency Tests
As the data were normally distributed, parametric analyses were run on the past-tense, general proficiency, and lexical test results.
First, one-way ANOVAs were used to determine whether the three areas of L2 performance involved varying levels of difficulty. To this end, the 88 participants' scores were turned into percentages due to the differential number of items across the three tests. The ANOVA results revealed varying levels of difficulty across the three L2 areas under examination, F(2) = 56.7, p < 0.001. Pairwise comparisons with Bonferroni adjustments ('α = 0.05/3 = 0.016) showed that the learners were better in vocabulary recognition (M = 72; SD = 9.4) than in the use of the L2 features included in the general L2 proficiency (M = 59.7; SD = 17.1, p < 0.001) and past-tense test, p < 0.001. For the participants, the production of past-tense morphology (M = 45.8; SD = 20.3) proved to be more demanding than the production of L2 features in the general proficiency test (p < 0.001) and vocabulary recognition; therefore, it constituted the most challenging area of L2 performance.
To explore the possible relationship between the three areas of L2 performance, Pearson correlation analyses were run on the tests' results. The analyses revealed that learners' performance in the past-tense test strongly correlated with their general L2 proficiency (p < 0.001, r = 0.732). Their vocabulary recognition ability exhibited a moderate correlation with past-tense morphology production (p < 0.001, r = 0.511) and general L2 competency (p < 0.001, r = 0.507).

Cross-Sectional Groups' Test Results
In light of these results, the past-tense test scores were used to organize the participants into the five L2 levels in Table 1. As the score in this test ranged from 0 to 45, five L2 proficiency groups were created on the basis of a 9-point interval, with the exception of Group 5 where the participant with the highest score was also included. One-way ANOVAs were used to identify differences in the past-tense production ability, general proficiency, and lexical knowledge across the groups. The analyses confirmed a significant difference in past-tense production between the groups F(4) = 348.9, p < 0.001. The pairwise comparison results with Bonferroni adjustments ('α = 0.05/4 = 0.012) in Appendix A confirmed that their past-tense production significantly increased across groups: 1 < 2 < 3 < 4 < 5.
As for the general L2 proficiency of the learners, the analyses also confirmed a significant difference among the groups, F(4) = 24.89, p < 0.001. Nonetheless, the pairwise comparisons with Bonferroni adjustments ('α = 0.05/4 = 0.012) in Appendix A revealed that the general L2 proficiency of learners in Groups 1, 2 and 3 was similar, and significantly varied only in comparison to that of Groups 4 and 5.
With respect to L2 vocabulary recognition, the results pointed to a significant difference between the groups, F(4) = 8.81, p < 0.001. However, the pairwise comparisons with Bonferroni adjustments ('α = 0.05/4 = 0 0.012) in Appendix A indicated that learners in Groups 1, 2, 3 and 4 held a similar vocabulary recognition ability. Only Group 5 proved to be more lexically advanced.

Cross-Sectional Sample Test Results
From the five past-tense proficiency groups, the 11 participants in Table 2 were selected for narrative analysis. The selection of these participants on the basis of the following quantitative measures aims to provide reliable data on their L2 proficiency prior to the narrative analysis. From each group, the student with the lowest past-tense score and the student whose pasttense score was closest to the group mean were selected. When two students shared a similar past-tense score, the L2 general proficiency score was considered as the second selection criteria, as the results in this test strongly correlated with the past-tense test results. For Group 5, the student with the highest score was also retained. Based on the meaning-oriented approach that we adopted (Dietrich et al. 1995;Klein and Perdue 1997;Klein and Stutterheim 2002;Noyau 2002 and above Section 2.1), the selection of this small number of participants allows for in-depth analyses of the linguistic resources that characterize each learner's L2 narrative capacity. The cross-sectional design of the study facilitates comparisons across learners from various L2 proficiency levels; these comparisons can provide evidence of the evolution of the discourse and morphological resources in L2 written narratives.

Narrative Analysis Results
This section presents the discourse and morphological analysis results of the L2 narratives.

Grounding
At the discourse level, the foreground/background clauses were present in the production of the selected learners from all five groups, with the exception of learner 214 in Group 1. Table 3 shows less frequent use of background clauses in Groups 1-2 than Groups 3-5 in F&D. Across the participants, the use of background clauses exhibits a tendency towards growth. In the narrative C&E, however, the background clause increase is less evident, as the participants from the various groups exhibit frequent use of background clauses, even at lower levels.
The frequency counts in Table 3 indicate that what separates the production of the learners in lower proficiency groups from their peers in more proficient groups is that the former produce short narratives where the events are organized chronologically. These narratives give the impression of an uncompleted story, leaving out valuable background information and linguistic resources (e.g., connectors, auxiliary verbs, verbal morphology) as in Example 1 from learner 213. This example is different from Example 2, where learner 101 in Group 5 creates a coherent narrative with background clauses, where the use of past forms contributes to separate the foreground from the background. Moreover, in Example 2, the hierarchization of events (foreground/background) is distinct, thanks to the use of temporal connectors, such as puis (then), un jour (one day), and alors (so). The video tell-PRESENT the story of a fiancé who get-PERFECTIVE married with much love. With time, they go-PRESENT through a series of situations that reinforce-PRESENT their relation. For instance, they lose-PERFECTIVE their baby. He give-PRESENT a promise of travelling to Peru and they start-PRESENT saving money. Example 2. "C'est l'histoire de deux personnes qui tombent amourau. Ils se sont mariés, puis ils ont construit la maison de leur rêves. Ils tout les jours regardaient les nooiches et il les cherchaient des formes different. Ils étaient beaucoup d'imagination. Aussi, ils aiment lire ensamble. Un jour, ils ont décidé d'avoir des enfants mails ils n'ont pas pu. Alors ils ont décidé de faire leur rêve realité. Ils rêvent d'avoir une maison dans une belle lieu. Alors, ils ont essayé de economiser mais ils ont eu des problèmes. Un jour, elle tombe malade et elle s'est restée dans l'hospital et elle mortée à la fin.
This story is about two people who fall-PRESENT in love. They get-PERFECTIVE married, then they build-PERFECTIVE the house of their dreams. Every day, they watch-IMPERFECTIVE the clouds, looking for cloud figures. They have-IMPERFECTIVE lots of imagination. Also, they like-IMPERFECTIVE reading together. One day, they decide-PERFECTIVE to have children, but they be-IMPERFECTIVE not able. So, they decide-PERFECTIVE to make their dream reality. They dream-PRESENT of having a house in a beautiful place. So they try-PERFECTIVE to save money but they have-PERFECTIVE problems. One day, she become-PRESENT ill and stay-PERFECTIVE at the hospital, and she die-PERFECTIVE in the end.

Verbal Morphology
Verbal morphology contributes to narrative coherence through the perfective/imperfective distinction which separates the foreground from the background. In our data, it is only in the most proficient group where the two past forms, the imperfect and the passé composé, are systematically used to aspectualise the narrative. In Table 4a,b, the distribution of verbal forms follows a consistent pattern in the two narratives. The learners mainly use the present as the narrative form in the foreground until Group 4. In this group, the passé composé takes over this role. An evolution is tangible in Group 3 where the number of present, passé composé and past participle forms increases, as well as a few occurences of the imperfect. Example 3 (Student 41 from Group 3) from F&D illustrates the diversity of these forms. The example shows how the learner attempts the present form, est marié (gets married), the past participle retournés (returned) and then a series of target-like occurrences of passé composé forms. The use of these forms suggests that sensibility to verbal morphology is emerging through a trial-and-error strategy. Moreover, the increase in the number of passé composé in the foreground among the learners in Groups 4 and 5 (see Table 4a,b) points to a gradual reorganization of the learners' morphological system, where the prototypical foreground-passé composé combination consolidates. Time go-PAST PARTICIPLE and she get-PRESENT married. The family come-PASSÉ COMPOSÉ to the place. She return-PAST PARTICPLE when she be-PRESENT grown-up. The bicycle fall-PASSÉ COMPOSÉ and the river die-PASSÉ COMPOSÉ. She be-PASSÉ COMPOSÉ and she fall-PASSÉ COMPOSÉ and she die-PASSÉ COMPOSÉ when she meet-PASSÉ COMPOSÉ her father. Table 5a,b show that the distribution of verbal forms in the background follows a consistent pattern. First, among the low proficiency learners, the number of background clauses is small, such as 0 for learner 214 (Group 1) and 2 for learner 213 (Group 2). Second, these learners almost exclusively use the present tense. The imperfect emerges in the production of learners in Group 3. Third, the number of imperfect instances increases in the background of the narratives of the learners in Groups 4 and 5 (see Table 5a,b). This use of verbal morphology suggests that the most proficient learners are consolidating the prototypical background/imperfect combination. As for the most complex forms, only one learner (99) produces the gerund. Example 4 shows a series of four concomitant events (stay/cannot/die/take the hand) reinforced by the adverbial jusqu'au ('until'). This temporal relation of closeness between events, where the referential movement of time encompasses sequentiality and simultanousness (en prenant 'while taking the hand'), is infrequent in our data (see also Kihlstedt 2015). Example 4. Elle est restée à l'hopital jusqu'au elle n'a pas pu plus. Elle est morte en prenant la main de son copain.
She stay-PASSÉ COMPOSÉ at the hospital until she can-PASSÉ COMPOSÉ not anymore. She die-PASSÉ COMPOSÉ while take-GERUND the hand of her boyfriend.

Telicity
The data in Table 6a,b support the DH hypothesis, which predicts that perfective morphology appears with telic predicates in the foreground and imperfective morphology with most atelic predicates in the background. The pattern is more evident in the narrative F&D (67 telic PC and 32 atelic IMP forms) than in C&E (50 telic PC and 41 atelic IMP).    The results in Table 6a,b point to a delicate interaction between telicity, verbal morphology and grounding. The systematic and regular opposition between telic past forms in the foreground and atelic past forms in the background is only evident in Group 5 and begins in Group 4. Moreover, the majority of "other" forms in the tables represent present tense morphology or past participles. For instance, participant 41 demonstrates an emergent grounding ability in F&D (Table 6b, see also Example 3) with eight telic PC predicates and three telic predicates with past participles (=Other) in the foreground and seven instances of atelic predicates in the background. Among less proficient learners (Groups 1-2), telic predicates, mostly in the present tense, allow students to maintain the story line. Table 6. Telic property distribution in background/foreground clauses. Table 6. Cont.

Telic Predicates Atelic Predicates Telic Predicates Atelic Predicates
late in the acquisition process. Moreover, they substantiate his finding that non-prototypical combinations are infrequent, even among the most proficient L2 learners. Our findings diverge from earlier studies that have used impersonal narratives where the use of background clauses is often scarce (Bardovi-Harlig and Comajoan 2020). In our data, the use of background clauses was scarce at the beginning but slowly consolidated and became frequent in the production of the most proficient learners. One possible explanation is that written narratives may enhance the use of grounding devices, as learners had time to retrieve and organize coherent narratives during task completion (Skehan and Foster 1999). Another difference concerns the use of lexical devices, such as temporal connectors. In studies of oral narratives in L2 French (e.g., Noyau et al. 2005;Benazzo and Starren 2007), using the meaningoriented approach, beginning and intermediate learners of French use connectors to create a coherent story in the absence of functional morphology. We did not observe this strategy in our data. It could be that the higher demands of oral narratives on the working memory creates a need to anchor the narrative on concrete lexical devices, such as sequential connectors.
The use of various tests for the selection of the participants in the cross-sectional sample constitutes a valuable aspect in our study of L2 narratives. These tests provided some insights into how different domains of the L2 profile correlate with learners' narrative abilities. The tests revealed that the learners held considerable general and lexical L2 knowledge, but their ability to identify obligatory contexts for the use of the perfective/imperfective distinction was still in progress. The narrative data indicated that the use of the aspectual opposition was still lagging behind but in evolution. This evolution was observable in the increasing use of imperfective morphology in background clauses. Nonetheless, the issue of how various L2 domains contribute to the learners' ability to create coherent narratives deserves further investigation. Due to the small size of our crosssectional sample, our findings call for future investigations that test the generalizability of the developmental path observed in our narrative data.
Author Contributions: Both authors equally contributed to the organization of the study and the manuscript. All authors have read and agreed to the published version of the manuscript.
Funding: This research received no external funding.

Institutional Review Board Statement:
The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the university department where the participants were enrolled.
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: Not applicable.

Conflicts of Interest:
The authors declare no conflict of interest. Table A1. Probabilities for between-group mean differences. Tables 4-6 present raw frequency counts, which are low. The tendencies of development appear clear in raw figures and are substantiated by the excerpt descriptions. The number of foreground/background clauses on the other hand, are numerous, and thus presented in percentages.