Background

As Toth (Reference Toth2008) documented, the effective management of a whole-class communicative task can provide models of the target grammar for the students, along with procedural assistance in generating and reformulating utterances. In addition, although teacher-fronted activities may provide fewer opportunities than pair/group work for individual student turns, it does provide input and rehearsal opportunities for non-participating students (e.g. Ohta's, Reference Ohta2000, Reference Ohta, Hall and Verplaeste2001 observations of private speech). Moreover, when the teacher goes beyond simply managing activities to actually participating in whole-class conversations with students, this can increase the richness of the input for learning grammar features (Collins, White, Trofimovich, Cardoso & Horst, Reference Collins, White, Trofimovich, Cardoso, Horst and Muñoz2012).

We also know from decades of L2 classroom research that feedback can be provided both by peers and by the teacher, and both are effective for learning (e.g. Sato & Lyster, Reference Sato and Lyster2012; Sippel & Jackson, Reference Sippel and Jackson2015). Peers can also construct grammatical knowledge cooperatively (e.g. Donato, Reference Donato, Lantolf and Appel1994; Swain, Lapkin, Knouzi, Suzuki & Brooks, Reference Swain, Lapkin, Knouzi, Suzuki and Brooks2009). However, peer-initiated attention to language may often be more narrowly focused on language needed for accomplishing task objectives (e.g. solving a puzzle, guessing a partner's picture); in contrast, teacher assistance may be more focused on the language objectives of the task (Davin & Donato, Reference Davin and Donato2013). In addition, learner-generated attention to form in pair/group work tends to be much more focused on vocabulary than grammar (Philp, Walter & Basturkmen, Reference Philp, Walter and Basturkmen2010; Kim & McDonough, Reference Kim and McDonough2011; Collins & White, Reference Collins, White, DeKeyser and Botana2019) and the grammar students focus on may not necessarily be the language objectives of the task anticipated by the teacher (Springer & Collins, Reference Springer and Collins2008; Nassaji, Reference Nassaji2013). While learner agency and control over learning is a desirable feature of L2 pedagogy, the dynamics of pair/group interaction may not maximize the learning potential of tasks that were chosen and designed to provide practice with specific grammar features (the contrast between task as workplan versus task as process, Breen, Reference Breen and Johnson1989; Coughlin & Duff, Reference Coughlin, Duff, Lantolf and Appel1994).

Research building on Toth's (Reference Toth2008) study has focused on ways in which teacher-managed oral practice can facilitate interlanguage development (e.g. Lantolf & Poehner, Reference Lantolf and Poehner2011; Toth, Reference Toth2011; Davin & Donato, Reference Davin and Donato2013; Ko, Reference Ko2014) and has documented ways in which teachers can create conditions for attention to language forms in peer interaction (e.g. Chavez, Reference Chavez2018; Collins & White, Reference Collins, White, DeKeyser and Botana2019). However, as Toth (Reference Toth2008) himself noted, the qualitative data, though informative, was limited to the recording of one task by one group only, which limited insights on the quantity and quality of the input/output opportunities that students had in their peer groups and how this compared with the teacher-led condition.

Research design

We begin by describing each of the four targeted variables, and then suggest how the study could be carried out.

1. (1) Grammar feature. The target feature in Toth's study was the Spanish anticausative se. It is simple in form, but semantically complex, which may explain the advantage to learners of receiving guidance from the teacher during oral practice. This raises the issue of whether forms that are more transparent in meaning and/or that require more repetitive practice to master their formal properties (e.g. compound verb tenses) may lend themselves better to communicative tasks in pairs/groups. An additional factor is learner familiarity with the target form: teacher guidance may be particularly beneficial in the early stages of learning a new grammar feature.

2. (2) Task. The tasks chosen for the Toth (Reference Toth2008) study for the two conditions (whole class/pair and small group interactions) were information exchanges that are more typically done in pairs/groups, such as comparing and describing pictures. Adapting the tasks to whole-class interaction resulted in uneven time spent doing the tasks, in favour of the whole-class condition. While this in itself is an interesting finding in terms of the teacher's management of on-task behaviour, it begs the question as to whether different types of communicative tasks would unfold in more comparable amounts of time in the two conditions (pair/group and whole class), facilitating more valid comparisons between TLD and LLD. In addition, the usefulness of the forms for completing the task could influence how much attention students give to the target form in LLD – for example, a spot the difference task involving various objects in different places may generate contexts where se could be used, but prepositions of place may actually be useful for solving the task.

3. (3) Teacher experience. The teacher in Toth's (Reference Toth2008) study was quite skilled at managing the discourse to allow for rich and more sustained student participation than has often been seen in research of teacher-fronted classes. How much experience and training do teachers need to be effective at leading discourse that yields richer turns/exchanges than the limited use of ‘initiation, response, feedback’ discourse frequently observed in teacher-fronted L2 classes?

4. (4) Feedback and target form use during task. As noted above, the amount and quality of feedback and of the use of the target form across all of the tasks during the two conditions were not measured in Toth (Reference Toth2008). This is clearly an important source of information for understanding any observed learning differences in the TLD and LLD conditions.

One approach to researching this issue would be to begin with a descriptive study examining discourse features in the two conditions on different grammar points, accompanied by focus-group interviews or stimulated recall to probe learners’ and teachers’ perspectives on the different practice conditions. This could be followed by experimental research targeting one of the features examined in the first study, comparing learning gains in the two conditions.

Alternatively, if time and resources permit, a full experimental study measuring both learning outcomes and the process of learning (including student/teacher perspectives) would yield rich data that could be analyzed in stages in separate but linked studies from the same data set. The design should include pre-, post-, and delayed post-treatment measures of learning; audio/video recordings of several pairs/groups doing the same task (LLD), and of whole class interaction with the same task (TLD) ideally at different points in the intervention; and interviews with teachers and (subsets of) students, making use of stimulated recalls where possible.

An alternative to comparing two groups of students in TLD and LLD conditions, as Toth (Reference Toth2008) did, would be to use a time-series design (see Mackey & Gass, Reference Mackey and Gass2016 for information on time series designs in SLA). This design could vary the same students’ experience with the two types of practice in succession: TLD with one feature; LLD with another. Another possible design would be to vary the order in which students experience TLD and LLD with the same form: TLD followed by LLD; LLD followed by TLD. From a practical perspective, time-series designs require a smaller sample size; from a validity perspective they have the advantage of holding the participant variable constant across the different conditions, allowing for the exploration of the impact of LLD vs TLD on learning/use of a grammar feature within the same group of students.

The three main components of this pedagogical approach were: (i) the task/teacher's creation of a meaning context for a target form that students were not yet using productively; (ii) the brief interruption of the task in progress to provide explicit instruction on the form needed for the context, using and modifying the students’ original language; (iii) subsequent opportunities during the task for students to use the target language to express their intended meanings. An extension study would aim to explore the immediate and long-term benefits of the approach for learning the target grammar; whether it is equally effective for developing new knowledge of language or consolidating partially acquired knowledge (what Samuda refers to as ‘knowledge construction’ and ‘knowledge activation’ respectively); and how much experience or training is needed for a teacher to create and manage a meaning/form/meaning sequence effectively.

Background

Samuda's (Reference Samuda, Bygate, Skehan and Swain2001) study documented students’ increased use of target grammar (modals of prediction) during the completion of the task following the teacher's intervention – a finding that was corroborated by Little & Fieldsend's (Reference Little and Fieldsend2018) small-scale study using the approach to encourage students’ use of passive voice during task performance. However, neither study was designed to address the questions raised above. It has nevertheless received considerable attention in support of the ‘leading from behind’ approach in the theoretical literature. For example, it has been used as a model to illustrate how the given-to-new and awareness principles can be used to guide the oral production of selected grammar features (e.g. Batstone & Ellis, Reference Batstone and Ellis2009; Nava & Pedrazzini, Reference Nava and Pedrazzini2018); it has served as an example of how grammar can be integrated into communicative activities (e.g. Spada & Lightbown, Reference Spada and Lightbown2008), consistent with principles of interactionist and transfer-appropriate learning; and it has been chosen to illustrate the role of the teacher in supporting students’ use/learning of grammar in task-based teaching (e.g. Bygate, Reference Bygate2016; Van den Branden, Reference Van den Branden2016; Han, Reference Han2018).

In addition, the approach addresses an attested long-term learning challenge that many features of grammar present; that is, the choice of the appropriate form for a given meaning/situation (Larsen-Freeman, Reference Larsen-Freeman, Celce-Murcia, Brinton and Snow2014) – particularly for features that are not semantically transparent, such as modality in English. It also engages students in ‘pushed output’ (Swain, Reference Swain, Doughty and Williams1998), a key benefit attributed to guided production, as it helps students notice gaps in their L2 knowledge for expressing intended meanings (Swain, Reference Swain, Cook and Seidlhofer1995). Moreover, there is evidence that learner-generated attention to form (including the resolution of grammatical problems) is greater and more successful when the teacher has identified key language useful for doing the task, and when the task requires repetition of the task-useful forms (Collins & White, Reference Collins, White, DeKeyser and Botana2019).

Research design

The research we suggest would follow the pedagogical principles outlined in the study, but would require innovations to the experimental design to address the research questions. Researchers may want to consider using a version of Samuda's (Reference Samuda, Bygate, Skehan and Swain2001) or Little and Fieldsend's (Reference Little and Fieldsend2018) tasks targeting epistemic modality or the passive voice, respectively, in English or any other language in which the forms exist. For other grammatical features, it will be important to identify those in which the learning challenge of the grammar is semantic (see above). Some examples include other types of modality (e.g. modals of obligation), the polysemous le particle in Chinese, and conditional forms in various languages. Of course, any new tasks/features will need to be piloted to ensure that the task generates natural contexts for the target features, and that the features are also useful for completing the task (Loschky & Bley-Vroman, Reference Loschky, Bley-Vroman, Crookes and Gass1993).

Researchers interested in developing a program of research on this topic would ideally begin with rich case studies and follow up with experimental designs with control/comparison groups.

We recommend that initial research control for prior knowledge of the grammar feature, since evidence (Samuda, Reference Samuda, Bygate, Skehan and Swain2001; Little & Fieldsend, Reference Little and Fieldsend2018) shows an increased use of the feature during the task by students who are creating new knowledge (as opposed to students who already have some knowledge of the form). Follow-up research could then make knowledge of the target feature a variable to be manipulated. For example, practice using a new feature could be compared with practice activating and consolidating existing knowledge of a feature, among the same participants. The comparisons would include both measures of production during the task and overall learning outcomes, in addition to learners’ perspectives on the relative benefits of the approach for the different features they practiced.

For the overall impact on the effectiveness of the approach, researchers could compare the relative benefits of semantically enhancing a feature in the ways in which Samuda (Reference Samuda, Bygate, Skehan and Swain2001) describes versus the more commonly used linguistic enhancement of the feature. The latter involves techniques such as increasing the frequency of the target form in aural or written texts students are working with; or increasing the salience of the form through emphasis or textual enhancement. Another contrast could be the timing of the explicit information on the same grammar point: pre-task, similar to a presentation-practice-production approach, versus mid-task, as advocated in Samuda's meaning-form-meaning sequence. Student perspectives would also be an integral part of the design.

Data should include recordings of student interactions during group tasks and whole class interactions with the teacher; counter-balanced pre-, post-, and delayed post-instruction measures of learning (with distractors); and post-instruction questionnaires to teacher and students on their experiences with the instruction and their perceptions of any learning benefits, probed in greater depth during interviews (ideally involving stimulated recalls). Researchers may also want to consider the advantages of a time-series design (explained above for Research Task 1).

The goals of the research are: (i) to document and analyze young L2 learners’ conscious focus on grammar during their interactions with each other and (ii) to provide insights on the communicative tasks conducive to generating children's attention to grammar and on the guidance children may need from the teacher to do so.

Background

Most classroom research supports the benefits of interaction for language learning. From the interactionist view, it draws attention to the context of meaning and allows learners to give and receive feedback, and adjust their language accordingly (Long, Reference Long, Ritchie and Bhatia1996). From a sociocultural perspective, it provides opportunities for peers to co-construct knowledge during meaningful interaction (Lantolf & Thorne, Reference Lantolf and Thorne2006) and to strengthen form-meaning connections as learners use and retrieve forms needed for communication (Mackey, Reference Mackey2012). This has often been measured in terms of language-related episodes (LREs) (Swain & Lapkin, Reference Swain and Lapkin1995) and their characteristics, such as resolution and language focus (Storch, Reference Storch2008; Payant & Reagan, Reference Payant and Reagan2018; García Mayo & Imaz Agirre, Reference García Mayo and Imaz Agirre2019). LREs have also been linked to learning outcomes, typically measured in post- and delayed post-tests (e.g. Swain & Lapkin, Reference Swain and Lapkin1998; Williams, Reference Williams2001).

Most of the above studies involve adults, with the youngest participants typically being adolescents (Swain & Lapkin, Reference Swain and Lapkin1998; Kim, Reference Kim2013) – García Mayo & Imaz Agirre (Reference García Mayo and Imaz Agirre2019) and Shintani's (Reference Shintani2014) work are notable exceptions. It has been posited that children learn their L2 like they learned their L1 – implicitly, relying more on phonological than grammatical cues (Karmiloff-Smith, Reference Karmiloff-Smith1979). This perspective may result from the prevalence of implicit instruction methods for young learners (Lichtman, Reference Lichtman2013). Yet there is evidence that children can acquire and use some metalinguistic knowledge to support their development (Harley, Reference Harley, Doughty and Williams1998; Bouffard & Sarkar, Reference Bouffard and Sarkar2008), and that they can identify and resolve some grammatical issues during communicative exchanges, especially when the teacher has cultivated a climate of cooperative learning (Gagné & Parks, Reference Gagné and Parks2013; Collins & White, Reference Collins, White, DeKeyser and Botana2019).

Research design

As noted in the introduction, examining this issue in school-based foreign language programs would make an important contribution to our understanding of children's focus on language in these common yet under-researched contexts. From a methodological perspective, it can also be easier for researchers in these contexts to control for and/or document some of the external factors that can have an impact on language learning behaviour, such as the amount of exposure and interaction in the target language outside of class time.

One approach would be to observe different aspects of children's classroom interaction using audio or video recordings of students during classroom communicative activities. Features examined could include interactional moves (e.g. clarification, correction), the extent and quality of collaboration, focus on language and/or task completion, and interactional roles assumed by the students. These variables could be compared with instances of focus on form identified in the recordings (perhaps operationalized as LREs) and to the type of grammar that children attend to. The latter could vary along several dimensions already documented in research among older learners, such as semantic salience (Bardovi-Harlig, Reference Bardovi-Harlig, Eckman, Highland, Lee, Mileham and Weber1995; Doughty & Williams, Reference Doughty, Williams, Doughty and Williams1998), perceptual salience (Goldschneider & DeKeyser, Reference Goldschneider and DeKeyser2001), and task-essentialness (Loschky & Bley-Vroman, Reference Loschky, Bley-Vroman, Crookes and Gass1993). Observations of activities typically used by teachers (as opposed to those designed by researchers for teachers to use) would enhance ecological validity (Collins & White, Reference Collins, White, DeKeyser and Botana2019).

The analysis could examine statistical associations between interactional features and type of language attended to. Depending on the interactional feature under study, coding may be categorical (e.g. interactional moves, student role, task versus linguistic focus), interval (e.g. number and length of turns), or ordinal (e.g. collaboration rating, Winke, Reference Winke, McDonough and Mackey2013). Grammatical items attended to can be classified according to semantic and perceptual salience, task-essentialness, or type of challenge presented. Researchers may choose to do in-depth analyses of individual learners or pairs, and report group results if the sample size allows. Group results can offer insights on what learners might be expected to do during specific tasks, which can help teachers choose or design tasks that promote desired behaviours (for example, which tasks can help them focus on a non-salient form). However, because the main goal of this research task is to examine learners’ on-task behaviour, focusing on specific learners or pairs might be more informative, documenting the inherent variability among individual learners.

Research questions targeting students’ and teachers’ perceptions of usefulness will also be informative. For example, do teachers choose communicative tasks that naturally promote focus on form or set up existing tasks to focus on language? Do young learners perceive differences between tasks with more communicative or more linguistic focus, and does this affect whether (and the extent to which) they focus on form? Exploring teachers’ perspectives will inform decisions such as whether tasks are best planned or adapted for target grammatical forms to minimize set-up time for teachers, and students’ perspectives can offer insights on how much guidance they need to focus on grammar during classroom activities.

The goals of the two sets of studies for this research task are: (i) to gain an understanding of spoken grammar as a cross-linguistic phenomenon and (ii) to design and evaluate pedagogical approaches to teaching spoken grammar, documenting teachers’ and students’ views on their experiences with such approaches.

Background

There is evidence that students are interested in at least a receptive knowledge of spoken grammar (Timmis, Reference Timmis2002; Goh, Reference Goh2009), and that they could benefit from a more representative coverage of spoken grammar in course materials (Cullen & Kuo, Reference Cullen and Kuo2007; Jones & Waller, Reference Jones and Waller2011). In addition, recent studies point to challenges in teaching spoken grammar in classrooms: learners may hesitate to use certain spoken forms for fear of being seen as ‘inauthentic’ and appearing ‘too native-like’ (Soruç & Griffiths, Reference Soruç and Griffiths2015), and their use of these forms may be perceived as less correct or acceptable compared with syntactically equivalent forms produced by native speakers (Ruivivar & Collins, Reference Ruivivar and Collins2019).

These issues stem from limitations in current research, one of which is that much of it has been with English, with some findings also available for French (e.g. Armstrong, Reference Armstrong2001). Another is that many descriptive accounts of spoken grammar are based on discourse analyses, which depend on pre-selection of grammatical constructions that fulfill discourse functions in a particular language, language variety, or context (Dontcheva-Navratilova, Reference Dontcheva-Navratilova and Chapelle2012). The teaching of spoken grammar is therefore informed by a limited understanding of a few well-studied features, such as discourse markers (Jones & Carter, Reference Jones and Carter2014) and flexible word order (Timmis, Reference Timmis2005).

Many spoken grammar features appear to be shared across languages, or at least those that have been studied, i.e. English and French. These include subject ellipsis (outside of pro-drop languages), implying shared knowledge of the subject and facilitating real-time interaction; and historical present, which adds animacy and emotional intensity to spoken narratives. Identifying the types of grammatical variation shared across languages (‘sociolinguistic universals’), and the communicative or pragmatic functions they serve would be valuable in presenting students with grammatical choices (Larsen-Freeman, Reference Larsen-Freeman, Hinkel and Fotos2001).

Central to the issues we have described is the question of whether and how spoken grammar should be taught, given that it meets with resistance from L2 users. Important insights on this issue have come from Valdman's (Reference Valdman2000) well-cited pedagogical norm for teaching sociolinguistically variable features, which describes most features of spoken grammar. Among his recommendations are to teach features that are widely used and valued in the community. Our research tasks aim to address the latter, in particular how aspects of spoken grammar are ‘valued’ as means of conveying social information and facilitating discourse. From a functionalist perspective, one might argue that spoken grammar, in particular forms that deviate from prescriptive norms, should be taught as communicative resources in the target language community, consistent with what Larsen-Freeman (Reference Larsen-Freeman, Hinkel and Fotos2001) refers to as a ‘grammar of choice’; that is, the idea that grammar consists of choices between two or more forms, each of which fulfills a different social or pragmatic purpose. Such an approach can address learners’ reluctance to align with what they judge to be native-speaker norms, and the effect of native/nonnative speaker status on the perceived grammaticality of certain forms. This functional approach has been successfully implemented in a number of studies (e.g. Bardovi-Harlig, Reference Bardovi-Harlig, Overstreet, Rott, VanPatten and Williams2004; Izquierdo & Kihlstedt, Reference Izquierdo and Kihlstedt2019), but applications to spoken grammar (and the broader field of grammatical sociolinguistic variation) are scarce.

Research design

The two sets of studies we describe could be done independently, or in succession (with the first informing the second), as part of a larger research program.

The main question for the descriptive research task is which social functions are fulfilled by spoken grammar across languages. Spoken corpora are available for an increasing range of languages – a helpful list is maintained by the Extensible Markup Language for Discourse Annotation (EXMARaLDA, Schmidt & Wörner, Reference Schmidt and Wörner2009; https://exmaralda.org), a system for managing, transcribing, and analyzing spoken corpora. This study can be approached from three perspectives, which we outline below:

1. (1) Variationist: Friginal and Hardy (Reference Friginal and Hardy2014) suggest examining selected linguistic units (e.g. ellipsis, historical present) in relation to societal units (e.g. age, gender). Analysis can follow the variationist paradigm (Labov, Reference Labov1967) by making statistical associations between these units. In keeping with a grammar of choice, factors must also take into account dynamic social functions such as the power dynamics of speech acts (e.g., conceding, challenging), self-presentation (formal, informal, casual, careful) and perceived relationship with interlocutors (friendly, professional).

2. (2) Dialectology: Alternatively, one can borrow from the field of dialectology by correlating preselected societal units with spoken grammar features, then using principal and hierarchal component analysis (e.g. Grieve, Reference Grieve2009) to identify clusters. This can be particularly helpful for geographically constrained, vernacular features that may be productive in a language but tend to be overlooked in the L2 literature (e.g. the informal question particle -tu in Quebec French, or the second-person address form voseo common in parts of Central and South America).

3. (3) Conversation analysis: CA methods may be useful for identifying the social functions of certain grammatical features that have not previously been documented in the literature. Ford and Mori's (Reference Ford and Mori1994) study of the use of causal connectors (e.g. because) to manage disagreement in English and Japanese is a good example of a CA-informed cross-linguistic grammatical investigation. Studies could undertake an analysis of NS-NNS interaction, which might reveal gaps in learners’ (in)ability to use and adjust their grammar to express concepts such as uncertainty, authority, or social distance.

Findings from the studies above would inform future directions for pedagogical design and research. This being said, it is also possible to conduct pedagogical studies that build on what we already currently know about spoken grammar. Notably, van Compernolle (Reference van Compernolle2013) and van Compernolle and Henery (Reference van Compernolle and Henery2014) report positive outcomes for concept-based instruction in university foreign-language settings. This approach promotes the teaching of concepts underlying grammatical choices, such as formality and social distance in the use of second-person address systems (e.g. French tu and vous, German du and Sie). Van Compernolle's work would benefit from replication in other contexts and using different L1-L2 pairings to see to what extent the benefits of such an approach can be generalizable. In addition, it would be interesting to see whether the outcomes of spoken grammar instruction are mediated by factors such as type of feature (e.g. salience, social markedness), learner proficiency, and learning context. Such a study would also require: (i) theoretically justified operationalizations of relevant constructs, such as Yuan and Zhang's (Reference Yuan and Zhang2018) use of a three-point scale to determine social distance and power differences; and (ii) new materials for concepts that have not been the subject of previous pedagogical studies. We recommend study designs other than experimental (pre-post-delayed post), as the type of knowledge targeted here may not be apparent within the typically short instructional period of pedagogical intervention studies. Longitudinal designs would be ideal to document the emergence of the target forms throughout instruction, but alternative designs that would be easier to implement include a time-series design (capturing behavior at different points in a one-semester intervention) or a micro-genetic design using audio or video recordings of classroom interaction to document changes in students’ discourse during instruction (including their attention to and uptake of features).

There is also very little research on how teachers and learners respond to such instruction beyond small-scale studies (e.g. Jones & Carter, Reference Jones and Carter2014; Soruç & Griffiths, Reference Soruç and Griffiths2015). Consequently, there is much to be learned from research that examines the feasibility of a focus on spoken grammar. To accomplish this goal, we suggest investigating learners’ perceptions and teachers’ responses to the pedagogy with participants representing a wider range of ages and goals for learning the L2 than have been studied to date. These can be obtained through interviews or stimulated recall, subjected to qualitative analysis, and, if resources permit, compared with performance outcomes. The findings would inform and refine current pedagogical approaches to address issues of social identity, learner agency, and other sociocultural challenges to teaching spoken grammar, which have only been hinted at in the pedagogical research to date.

The goal of the research is to explore how grammatical information in a corpus could be presented to render it more appropriate for pedagogical purposes, thereby minimizing teacher involvement and promoting learner autonomy. In other words, how could grammatically tagged corpus tools be made more learner-friendly?

Background

Learners can independently explore linguistic data from corpora and concordancers, either by accessing corpora on their own or with items pre-selected and provided by the teacher in paper-based activities. Boulton and Cobb's (Reference Boulton and Cobb2017) meta-analysis showed overall positive learning outcomes for DDL in classrooms. Although much of the research they reviewed was on vocabulary acquisition, encouraging results have also been reported for grammar. In particular, DDL draws learners’ attention to less salient forms in the input, allowing them to correct persistent errors such as overuse of be (e.g. *I am go to work, Moon & Oh, Reference Moon and Oh2018) and incorrect verb-preposition collocations (e.g. *participate to, Ruivivar & Lapierre, Reference Ruivivar, Lapierre, Taalas, Jalkanen, Bradley and Thouësny2018). Others have shown that DDL affords greater cognitive involvement and meaningful engagement with target forms, thereby increasing retention (e.g. English passive voice in Smart, Reference Smart2014; and a range of lexicogrammar features in Boulton, Reference Boulton2010).

Most DDL researchers claim that the strength of such an approach lies in ‘cutting out the middleman’ (Johns, Reference Johns1991, p. 30); in other words, students develop independent language learning strategies with minimal guidance from the teacher. This is particularly relevant for computer-based DDL, as learners are expected to explore language beyond the scope or the time frame of their language course. However, with regard to grammar, it is precisely in promoting independent use that DDL appears to pose challenges. One issue is that learners can be overwhelmed not only by the large numbers of concordance lines generated by the search, but also by the difficulty understanding the context in which a target feature is used (Geluso & Yamaguchi, Reference Geluso and Yamaguchi2014). Another issue is that many grammar-focused DDL studies make use of grammatically tagged corpora which may be better designed for research than for pedagogy. As such, DDL often requires a fair amount of metalinguistic knowledge, computer literacy, search skills, and teacher guidance, particularly for younger and lower-proficiency learners. In a study tracking learners’ real-time behaviour during DDL activities using the British National Corpus (BNC), Pérez-Paredes, Sánchez-Tornel, Alcaraz Calero and Aguado Jiménez (Reference Pérez-Paredes, Sánchez-Tornel, Alcaraz Calero and Aguado Jiménez2011) found that learners using the tool without teacher guidance made less effective use of corpus data than the guided learners. However, in both guided and non-guided conditions, learners made no use of the part-of-speech tags they had been taught. Geluso and Yamaguchi (Reference Geluso and Yamaguchi2014) had students use a different tagged corpus, the Corpus of Contemporary American English (COCA), and while the students found it useful overall, students felt discouraged by the time it took to learn the concordancing functions. To be sure, developing students’ skills for autonomous corpus exploration is expected to take time and scaffolding (Cheng, Warren & Xun-feng, Reference Cheng, Warren and Xun-feng2003), and teacher guidance is necessary for managing the size and number of concordance lines presented and providing questions to focus students’ attention (Lin & Lee, Reference Lin and Lee2015). However, minimizing teacher involvement would be more consistent with the goal of promoting independent learning. It is therefore worth exploring how corpora could be designed to address what appear to be the main obstacles: managing the amount of data that students encounter, and making search functions more intuitive and accessible.

Research design

Specific research questions would include what problems learners typically run into when using grammatical information in a corpus (for example, understanding codes for parts of speech, translating a grammar question into a search string, or processing too many concordance lines), how these problems might be addressed (e.g. more transparent codes or intuitive search commands, selecting the most useful concordances), and whether implementing these changes can address some of the problems documented in grammar-focused DDL research. Boulton (Reference Boulton2010) has also noted that DDL studies tend to focus on learner preferences and impressions of DDL, with less attention paid to learning outcomes. Whether and how learners’ grammar improves as a result of DDL would therefore also be worth investigating. A useful starting point might be the issues commonly encountered in automatic tagging, such as polysemous words (e.g. set, let) and distinguishing between gerunds and infinitives; and how learners might deal with such challenges in the absence of teacher intervention. Researchers may also target specific learner populations that typically require more teacher-guided forms of DDL, such as younger and lower-proficiency learners (e.g. Boulton, Reference Boulton2010, see also Allan's (Reference Allan2009) work with a graded reader corpus), or start by targeting one or two forms that are known to be challenging to learners in the L2 of interest (e.g. article use, particles, prepositions). These studies would initially be exploratory and might employ a user experience perspective, focusing on learners’ usage behaviours to paint a clearer picture of the obstacles to independent usage, similar to Pérez-Paredes et al. (Reference Pérez-Paredes, Sánchez-Tornel, Alcaraz Calero and Aguado Jiménez2011). Findings from such studies could then inform potential avenues for improving accessibility and minimizing the need for guidance.

An exploratory study could be followed by experimental or quasi-experimental work deploying more accessible corpus tools in classrooms. In keeping with Boulton and Cobb's (Reference Boulton and Cobb2017) recommendations, the pedagogical goal – in addition to specific grammatical knowledge – would ideally include transferrable skills that are not necessarily limited to grammar, such as inferring meaning from context and online search strategies. Further down the line, and in response to Flowerdew's (Reference Flowerdew, Leńko-Szymańska and Boulton2015) call for large-scale, longitudinal studies on the benefits of DDL, these studies can inform future work on the specific affordances of DDL for grammar. For example, are there benefits for students struggling with different lexicogrammatical forms (e.g. avoir and être participles in French)? Are the benefits/challenges similar when the grammatical query involves form only (e.g. regular/irregular past tense marking in English) versus choosing the appropriate form for an intended meaning (e.g. tense/aspect contrasts)? A longitudinal perspective could also document whether any attested benefits extend to students’ more autonomous use of DDL over time.