Second language structural priming: A critical review and directions for future research

I Introduction

Structural priming – also referred to as syntactic priming, structural persistence, adaptation, or alignment – is the phenomenon whereby speakers are more likely to repeat a structure they have recently heard or produced (Bock, 1986). For example, upon hearing a passive sentence, like the church was struck by lightning, a speaker is more likely to use a passive construction, like the boy was stung by a bee, in a subsequent utterance than its active alternative, as in the bee stung the boy. Experimental research shows that priming occurs across a variety of structures and languages, in both written and oral modalities, in comprehension and production, and among child and adult first-language (L1) and second-language (L2) speakers (for reviews, see Mahowald et al., 2016; Pickering and Ferreira, 2008). Structural priming also occurs in natural discourse, as demonstrated by L1 and L2 corpus analyses (e.g. Collentine and Collentine, 2013; Reitter and Moore, 2014; Thomas, 2016; Travis et al., 2017; for review, see Gries and Kootstra, 2017). Further, such priming can be long-lasting, supporting claims that structural priming represents a form of implicit learning (e.g. Branigan and Messenger, 2016; Chang et al., 2006; Dell and Chang, 2014; Kaschak, et al., 2014). This connection between priming and learning has important implications for L2 acquisition research, as the investigation of how and when priming occurs in L2 production can provide critical insight into the underlying mechanisms that aid L2 acquisition and use (e.g. Hartsuiker and Bernolet, 2017; Hartsuiker and Pickering, 2008; McDonough and Trofimovich, 2009). In what follows, I present a critical review of experimental research on within-language priming among adult L2 speakers and its implications for L2 learning (for a review of cross-language priming research, see Kootstra and Muysken, 2017; van Gompel and Arai, 2017).

II Methodological tools to investigate L1 and L2 structural priming

A variety of methods exist to investigate structural priming and the factors that influence the magnitude and longevity of its effects. In one commonly-used experimental paradigm (e.g. Bock, 1986), participants complete a picture description task, in which they see a picture and hear a sentence presented via computer. They then complete some sort of distractor task, such as determining whether the sentence and picture match in meaning. This initial sentence serves as the prime sentence. Then participants describe a second picture, the target picture/sentence, and researchers measure how often the syntactic structure of participants’ target sentence production matches the structure of the prime sentence. In a second popular paradigm, confederate scripting (e.g. Branigan et al., 2000), participants complete a similar task, but do so with another person rather than a computer. Unbeknownst to the participant, the second person is not an actual participant and her sentences are scripted, serving as prime sentences for the actual participant in the study.

In experimental research on structural priming, one generally distinguishes between short-term (or immediate) priming and long-term priming. Short-term priming refers to instances when a speaker re-uses the syntactic structure of the immediately preceding sentence that she encountered. As explained in greater detail in the next section, short-term priming is often attributed to either residual activation of the syntactic node for the preceding sentence or explicit memory of the preceding sentence. Long-term priming refers to instances when priming extends over a longer period of time and is usually attributed to implicit learning mechanisms (for details, see the following section). Long-term priming occurs when speakers encounter multiple filler sentences with a variety of syntactic structures between the prime and target sentence (e.g. Bernolet et al., 2016; Bock and Griffin, 2000; McDonough and Kim, 2016; Shin and Christianson, 2012). The magnitude of long-term priming can also be measured by comparing the proportion of target utterances a speaker produces in a baseline phase to the proportion of target utterances a speaker produces in a posttest phase, with the baseline phase occurring prior to and the posttest phase occurring immediately after a priming phase containing the actual prime sentences (e.g. Hartsuiker and Westenberg, 2000; Jackson and Ruf, 2017, accepted; Kaschak et al., 2011; Kutta et al., 2017; McDonough, 2006; Ruf, 2011). Some studies also show that long-term priming can extend to increased production of the target structure on posttests conducted a day or even several weeks later, again compared to baseline performance prior to the priming task (e.g. Branigan and Messenger, 2016; Kaschak et al., 2014; Kim and McDonough, 2016; McDonough and Chaikitmongkol. 2010; McDonough and Mackey, 2008; Shin and Christianson, 2012). Finally, researchers also measure cumulative priming, which is whether the magnitude of priming increases over the course of a task due to repeated encounters with the target structure, either through prime sentences or a speaker’s own production (e.g. Jackson and Ruf, 2017; Kaan and Chun, 2017; Kutta et al., 2017).

III Theoretical accounts of L1 and L2 structural priming

Researchers have posited several theories to account for findings from L1 priming research using the types of experimental paradigms outlined above. Lexicalist accounts propose that the residual activation of abstract syntactic representations, or combinatorial nodes, leads to repeated use of a particular syntactic representation (Pickering and Branigan, 1998). More specifically, nodes for individual lexical items are linked to combinatorial nodes representing the possible ways to combine these items in a sentence (e.g. ‘noun phrase’ + ‘noun phrase’ for double object dative constructions, like The man sends the woman a letter). During production, these lexical and combinatorial nodes are activated and remain active for several seconds, which can prompt a speaker to use the same construction over multiple utterances. Residual activation can successfully account for the lexical boost effect, whereby the magnitude of short-term priming increases in instances of lexical repetition between prime and target sentences, as such representations would strengthen activation of the overlapping lexical nodes and lead to greater residual activation of the corresponding combinatorial nodes from the prime sentence (for review, see Pickering and Ferreira, 2008).

However, long-term priming is more difficult to explain under a strict lexicalist account (but see Malhotra et al., 2008). Such longer-term effects are most often characterized as a type of error-based implicit learning (e.g. Branigan and Messenger, 2016; Chang et al., 2006; Dell and Chang, 2014; Jaeger and Snider, 2013; but for an alternative learning-based account, see Reitter et al., 2011). For example, if one hears a passive construction in a context where one had predicted to hear an active construction, the resulting prediction error leads to heightened expectations for encountering a passive sentence in later input. Such changes in expectations increase the likelihood of producing a passive sentence in subsequent production. Over time, such adjustments can result in cumulative and sustained changes within the language production system, i.e. learning.

Implicit learning accounts are supported by inverse frequency effects in priming, whereby the magnitude of priming is greater for less frequent vs. more frequent structural alternatives (e.g. Bock, 1986; Kaschak et al., 2011). Verb bias, namely the frequency with which a given structural alternative occurs with a particular verb, can similarly influence the magnitude of priming (e.g. Bernolet and Hartsuiker, 2010; Jaeger and Snider, 2013; Peter et al., 2015, but for counterevidence see Kaan and Chun, 2017). Additional evidence to support implicit learning accounts comes from cumulative priming effects, in that the magnitude of structural priming can strengthen over time, due to increased cumulative exposure to the target structure (e.g. Jackson and Ruf, 2017; Kaan and Chun, 2017; Kutta et al., 2017). In this regard, priming parallels learning in that greater learning generally occurs with something that is initially less well-known compared to something that is already well-known.

Bernolet et al. (2016) recently proposed a multifactorial account of structural priming (see also Reitter et al., 2011, for an alternative two-mechanism account). They argue that implicit learning mechanisms drive long-term priming, as outlined above. However, explicit memory of the prime sentence is the underlying mechanism that supports short-term priming, and the lexical-boost effect in particular, as explicit memory processes aid retrieval of the prime sentence’s structure in subsequent production (for similar discussion in the context of priming among L1-speaking children, see Branigan and McLean, 2016).

Explicit memory also plays a central role in the developmental model of shared syntax, recently proposed by Hartsuiker and Bernolet (2017). This model is designed to account for within- and across-language priming among L2 speakers, and how the strength of such priming effects varies with L2 proficiency. In the earliest stages of learning, L2 speakers’ representations are lexically based, with only weak links to abstract L2 syntactic representations (or combinatorial nodes) that constrain the possible ways to combine lexical items into a grammatical utterance. At this stage, L2 speakers may rely on existing L1 structures or imitate recently heard L2 structures when producing L2 utterances. As proficiency increases, links develop between syntactic representations and individual L2 lexical items, with these links developing earlier for more frequent L2 structures. In these earlier stages, priming stems from explicit memory of a previously encountered sentence, not the activation of abstract syntactic representations. Lexical boost effects (e.g. Jackson and Ruf, 2017; Kim and McDonough, 2008) are the result of greater explicit memory of prime sentences in instances of lexical repetition, although Hartsuiker and Bernolet posit that explicit memory processes may drive priming even in the absence of lexical repetition at this stage. Only at advanced proficiency levels are combinatorial nodes generalized across L2 lexical items and, eventually, between the L1 and the L2. Only then is within- and across language priming the result of residual activation of abstract combinatorial nodes, rather than explicit memory of recently encountered sentences.

IV Cross-language interaction in L2 structural priming

Given that the frequency of different structural alternatives and verb-bias information can impact the strength of L1 structural priming, a critical question for L2 research is the extent to which the frequency of a given structure in the L1, L2 or some combination of both has the greatest impact on L2 structural priming. This question contributes to larger debates regarding the importance of cross-language influence in L2 acquisition and processing more generally (see, for example, amongst others, Clahsen and Felser, 2006; Jarvis and Pavlenko, 2008; MacWhinney, 2012). Initial studies with advanced L2 English speakers revealed that the presence or absence of a given L1 structural alternative had little impact on the magnitude of L2 priming (for parallel discussion of L1 frequency effects in L2 comprehension priming, see Nitschke et al., 2010, 2014). For instance, Shin and Christianson (2012) found that advanced L1-Korean–L2-English speakers exhibited significant immediate short-term priming of particle placement with phrasal verbs (e.g. The girl is turning the heater down vs. The girl is turning down the heater), and that the increased production of separated phrasal-verb constructions extended to a delayed posttest one day later, even though no equivalent structure exists in the participants’ L1 Korean. In contrast, short-term and long-term priming was more variable for double-object (DO) vs. prepositional-object (PO) dative constructions (e.g. The man is reading his grandson a story vs. The man is reading a story to his grandson), even though both DO and PO constructions exist in the participants’ L1 Korean. Similarly, Flett et al. (2013) found equivalent short-term priming for DO constructions in L2 English among advanced L1-Spanish–L2-English and L1-German–L2-English speakers, even though only PO dative constructions exist in Spanish, whereas both PO and DO dative constructions exist in German and English. Together, such findings suggests that at least among highly-proficient L2 speakers, L1 experience is not a significant factor for within-language L2 structural priming.

More recently, however, Jackson and Ruf (2017) argued that L1 frequency, combined with the frequency of a given structure in L2 input, can modulate the strength of short-term and longer-term priming among intermediate L2 learners. Specifically, intermediate L1-English–L2-German learners exhibited equivalent short-term priming of adverb-first word order (adverb–verb–subject vs. subject–verb–adverb order) in L2 German with both fronted temporal (TP) and fronted locative (LP) phrases (e.g. Im Winter / Auf dem Berg trägt der Schüler eine Jacke ‘In winter / on the mountain the pupil wears a jacket’), with greater priming in the presence of lexical repetition of the fronted TP or fronted LP between prime and target sentences. However, long-term priming, measured via sustained production of adverb-first (i.e. XVS) sentences in a posttest phase that immediately followed the priming phase of the task, was limited to sentences containing TPs. In contrast, the participants produced no fronted-LP sentences in a baseline phase at the beginning of the task, and their production of fronted-LP sentences was not sustained in the posttest phase. This differs from a previous study, in which more-proficient L1-English–L2-German speakers, as well as L1 German speakers, exhibited significant long-term priming of XVS word order with similar LP constructions (Ruf, 2011).

Situating their findings within Hartsuiker and Bernolet’s (2017) developmental model of shared syntax, Jackson and Ruf (2017) concluded that these intermediate L2 learners had developed abstract syntactic representations for fronted-TP constructions, leading to significant long-term priming. For fronted-LP constructions, however, such representations were still lexically-based, accounting for the presence of significant short-term, but no long-term, priming. These differences across constructions likely stemmed from the frequency of these two constructions in the learners’ L1 English, whereby fronted TPs are more common than fronted LPs (Jackson, 2012), and the fact that most beginning German textbooks include fronted TPs but fronted LPs are rare (e.g. Lovik et al., 2013). Critically, the target structure (i.e. XVS word order) is identical for fronted-TP and fronted-LP constructions, suggesting that among less-proficient L2 learners, the robustness of long-term priming – but not short-term priming – may hinge on the strength of specific semantically-constrained constructions in memory, rather than the strength of generalized syntactic representations of a given structure per se. Further, the nature of such semantically-constrained representations may be modulated by L1 experience. Such an account is bolstered by Gerwien and Flecken (2015), who reported no significant conceptual priming of progressive aspect in L2 Dutch among L1-German–L2-Dutch speakers, which they attributed to the fact that the primed L2 structure required participants to encode L2 conceptual information that is not part of their L1 German grammar.

V Structural priming and L2 learning

In light of the differential long-term priming of fronted-TP vs. fronted-LP constructions in Jackson and Ruf (2017), the question arises as to what are the necessary prerequisites for priming to take place, especially among less-proficient L2 learners. McDonough (2006) found that among intermediate to advanced L2 English speakers from a variety of L1 backgrounds, only those participants who produced at least one DO dative construction in a baseline phase at the beginning of the study subsequently exhibited significant short-term priming of DO sentences during a priming task. Even then, the longevity of DO priming did not extend to speakers’ productions during a postpriming phase. In contrast, participants exhibited significant short-term and long-term priming with PO dative constructions: precisely the construction they were already more likely to produce in the baseline phase.

More recently. Kaan and Chun (2017) found significant cumulative priming over the course of the priming task for DO constructions among L1-Korean–L2-English speakers – as measured by how often participants typed a DO or simple transitive construction, rather than a PO construction, after encountering a DO prime sentence – but no comparable short-term priming of DO constructions immediately after encountering a DO prime. Further, the magnitude of cumulative priming for DO and transitive constructions was greater than the magnitude of cumulative priming for PO constructions. Kaan and Chun argued that because DO constructions are the less-frequent alternative in L2 Korean speakers’ output, cumulative priming was greater for this structural alternative, in line with the inverse frequency effect found in L1 structural priming research (e.g. Branigan and Messenger, 2016; Kaschak et al., 2011; Peter et al., 2015). At the same time, the absence of immediate short-term priming of DO constructions, combined with the fact that the L2 speakers often typed simple transitive constructions rather than complete DO sentences that contained both a direct and an indirect object (e.g. The boy threw the ball vs. The boy threw his friend the ball), led Kaan and Chun to conclude that the L2 speakers’ syntactic representation for DO dative constructions was still underspecified.

The results from Kaan and Chun (2017) suggest that L2 speakers may exhibit inverse frequency effects – similar to findings from child and adult L1 priming research – if one calculates frequency as a function of which structural alternative L2 speakers are less likely to produce independently of priming (e.g. DO dative constructions). However, based on Jackson and Ruf (2017) and McDonough (2006), some level of baseline knowledge of the less-frequent structural alternative is required before significant priming, and long-term priming in particular, can occur. In terms of Hartsuiker and Bernolet’s (2017) developmental model of L2 priming, such findings reinforce the claim that abstract syntactic representations only develop over time, as cumulative exposure to target L2 structures increases, and that the depth of such representations significantly modulates the magnitude and longevity of priming in L2 production.

This hypothesis was tested more directly by McDonough and Fulga (2015), who investigated the initial stages of L2 learning using the transitive construction in Esperanto (see also McDonough and Trofimovich, 2015). Esperanto has flexible word order, and marks accusative objects with the morphological suffix -n. L1 Thai and L1 Farsi speakers (all of whom also reported proficiency in L2 English) first completed an auditory picture matching task that contained both SVO and OVS sentences involving simple transitive action verbs (e.g. kapro batas tauron; SVO ‘goat hits ball’), to expose the participants to meaningful L2 Esperanto input. Critical to the discussion here, only those participants who successfully detected both the syntactic and morphological components of the target structure, as measured by high comprehension accuracy for both SVO and OVS sentences on immediate posttests, were successfully primed to produce OVS and SVO sentences in a subsequent primed production task. In other words, they were more likely to produce a correct OVS sentence after hearing an OVS prime, and more likely to produce a correct SVO sentence after hearing an SVO prime, although the strength of SVO priming was modulated by participants’ L1 (Thai, an SVO language, vs. Farsi, an SOV language). Participants who only comprehended SVO sentences in the posttest measures – who presumably had adopted an incorrect word order strategy to identify agent/patient roles in Esperanto – were only primed to produce SVO sentences. Participants who did not comprehend either SVO or OVS sentences in the posttest measures exhibited no significant priming for either SVO or OVS sentences. Such results highlight that priming can occur even after only limited exposure to a novel structure, and that the strength of such priming may be stronger for more novel structures (i.e. OVS sentences), paralleling inverse frequency effects reported elsewhere. At the same time, detection of the structure was a necessary prerequisite for priming to occur, suggesting that priming can help reinforce form-meaning mappings, but its impact is minimal in the absence of sufficient knowledge of the target structure.

Whether L2 speakers prime in the absence of lexical repetition between prime and target sentences has also been implicated in the potential for short-term priming to lead to long-term priming and L2 learning. Several studies have shown that only L2 speakers who exhibit significant short-term priming in the absence of lexical repetition – and not those who only exhibit short-term priming in lexical-boost conditions – continue to produce target structure utterances on immediate posttest measures (Jackson and Ruf, accepted; Ruf, 2011) or on delayed posttests up to five weeks later (McDonough, 2011). In a classroom-based study using communicative priming tasks that targeted the production of wh-questions, in which there was no lexical repetition between prime and target sentences, McDonough and Chaikitmongkol (2010) found a significant correlation between the number of target sentences L2 English speakers produced during priming tasks and the magnitude of L2 speakers’ sustained production of target wh-questions on immediate and delayed posttests (see also McDonough and Mackey, 2008). Critically, there was no significant correlation between participants’ production on the posttest measures and the number of prime sentences that they heard during the communicative priming tasks. This underscores the importance of actually producing accurate target sentences for supporting long-term priming, as simple exposure to such structures via prime sentences may not be sufficient. This finding is supported by other recent studies in which repeating prime sentences aloud increases the probability of significant long-term priming on posttest measures among less-proficient L2 speakers (Jackson and Ruf, accepted; Kim and McDonough, 2016; for similar results with child L2 speakers, see also Gámez and Vasilyeva, 2015). Further, Kim and McDonough (2008) reported greater short-term priming of passive constructions in the absence of lexical repetition among more-proficient L1-Korean–L2-English speakers, while less-proficient L2 speakers were primed only in lexical boost conditions (see also Ruf, 2011; for parallel findings in cross-language priming, see Bernolet et al., 2013; Hartsuiker and Bernolet, 2017). Together, such findings highlight that developing syntactic representations that generalize beyond individual lexical items is a critical factor in whether significant short-term priming can lead to long-term priming and the potential for longer-term L2 learning (see also McDonough and Mackey, 2006).

Several studies have also investigated the impact of type frequency on short-term and long-term priming of wh-questions among L1-Thai–L2-English speakers. McDonough and Kim (2009) found that activities that forced L2 speakers to produce target wh-questions with a wider variety of verbs led to greater long-term gains in the production of more advanced forms of wh-questions, as measured by a post-priming task that immediately followed a sequence of priming tasks. The degree of lexical variety in prime sentences had no significant impact on participants’ subsequent productions. However, in a follow-up study McDonough and de Vleeschauwer (2012) found that only those L2 speakers with high auditory pattern discrimination abilities benefited from such high prompt-type frequency. Participants with lower auditory pattern discrimination abilities exhibited greater gains when they were prompted to produce target sentences with only a limited set of target verbs. These differences were maintained on a delayed posttest administered two weeks after the priming activities were completed. From this, McDonough and de Vleeschauwer concluded that only L2 speakers who are good at detecting underlying L2 structural patterns benefit from producing utterances with greater lexical diversity. In contrast, L2 speakers with lower pattern discrimination skills may need repeated exposure and practice with the same lexical items in order to recognize critical L2 structural patterns. This conclusion underscores that the magnitude of L2 structural priming can vary significantly across individuals, and that not all L2 speakers are primed under the same conditions. Such differences highlight that the development of connections between lexical and syntactic representations may not proceed in identical ways for all learners, which has important implications for theoretical accounts of structural priming.

VI Conclusions and future directions

As demonstrated by the research reviewed here, even less proficient L2 learners can exhibit within-language structural priming during L2 production (e.g. Jackson and Ruf, 2017; Kim and McDonough, 2008, 2016; McDonough and Fulga, 2015). Additionally, the manipulation of L2 input in ways that encourage priming, and the incorporation of priming activities in the L2 classroom, show promise as a means to boost L2 grammatical learning (e.g. McDonough and Chaikitmongkol, 2010; McDonough et al., 2015; Trofimovich and McDonough, 2011; for evidence that the benefits of classroom-based priming activities can extend to L2 pronunciation, see also Trofimovich et al., 2013). However, more research is needed to understand the underlying mechanisms that support structural priming and the contexts in which priming is most successful for promoting longer-term learning. In particular, more research should focus on identifying at what stages of L2 proficiency short-term priming is most effective for facilitating longer-term changes in L2 production. Specifically, additional research is needed on whether priming can ever successfully facilitate the acquisition of a new structure in earlier stages of L2 learning, or whether it is most beneficial in later stages of learning, as a means to reinforce the development of abstract syntactic representations (McDonough and Fulga, 2015; McDonough and Trofimovich, 2015). Many researchers also work under the assumption that priming among L1 speakers is implicit in nature (e.g. Bernolet et al., 2016; Kaschak et al., 2011; Kutta et al., 2017), but this may not always be the case in L2 contexts (e.g. Jackson and Ruf, accepted; McDonough and Kim, 2016; Shin and Christianson, 2012; see also discussion in Kidd, 2012a, 2012b). Expanding research in these ways would not only advance our understanding of the pedagogical implications of structural priming, but would also serve to refine theoretical accounts of L1 and L2 structural priming, and provide insight on whether the mechanisms that support priming, and its potential to facilitate longer-term learning, work in similar ways across populations.

Additionally, researchers should more carefully consider the potential impact of individual learner variables, like statistical learning ability and openness to linguistic variability, on the magnitude and longevity of L2 priming, especially in light of studies that highlight the high level of variability across L2 speakers in their tendency to exhibit short-term and long-term priming (e.g. Jackson and Ruf, accepted; McDonough and Chaikitmongkol, 2010; McDonough and Kim, 2016; McDonough and de Vleeschauwer, 2012; for parallel work with L1 children, see Kidd, 2012a, 2012b; Serratrice et al., 2015). L2 priming research – much like L2 acquisition research more generally – would also benefit from expanding research to include a wider variety of L1–L2 pairings and grammatical structures, as a vast majority of L2 priming research has focused on a limited number of structures in L2 English (for exceptions, see Collentine and Collentine, 2013; Gerwien and Flecken, 2015; Jackson and Ruf, 2017, accepted; McDonough and Fulga, 2015; Ruf, 2011; Thomas, 2016). The systematic investigation of individual difference variables and additional languages and linguistic structures would significantly expand the scope of L2 priming research, and test for the generalizability of current findings to broader populations.

Also understudied in L2 priming research to date is how social factors may modulate L2 structural priming. For instance, Chun, Barrow and Kaan (2016) recently showed that a speaker’s accent in English (American, Korean, or Indian speaker of English), and how familiar someone is with that accent, can modulate the magnitude of short-term priming among L1 English speakers. Such findings support and expand the interactive alignment account proposed by Pickering and Garrod (2004, 2013; see also Garrod and Pickering, 2007), which posits that speakers align with – or prime – each other during dialogue, as a means to aid mutual understanding. However, no research to my knowledge has investigated how various social factors, including the L1 vs. L2 speaker status of an interlocutor, or the proficiency level of an L2 interlocutor, might modulate the strength and longevity of priming among L2 speakers. However, such research would speak to core issues in L2 research regarding the importance of interaction, and peer interaction in particular, for L2 learning (e.g. Mackey et al., 2012).

Similarly understudied in previous L2 structural priming research is the systematic investigation of short-term and long-term priming of sentences presented in a richer discourse context, as most studies have investigated priming using experimental paradigms involving isolated and decontextualized sentences. However, recent corpus and classroom-based research demonstrates that L2 priming also occurs in more natural contexts outside of the laboratory (e.g. Collentine and Collentine, 2013; McDonough and Chaikitmongkol, 2010; McDonough et al., 2015; Thomas, 2016; for relevant work with L1-speaking children, see also Hesketh et al., 2016; Serratrice et al., 2015; Vasilyeva et al., 2006). More work along these lines is needed, especially concerning the acquisition of discourse-appropriate use of various L2 linguistic forms. This is important because natural conversation does not consist of isolated sentences, but rather of connected discourse that is speaker and context dependent (for reviews, see Brown-Schmidt et al., 2015; Hoeks and Brouwer, 2014), and socially-situated interaction plays a central role in L2 acquisition (e.g. Mackey et al., 2012). Given that the ultimate goal of L2 acquisition is not just to acquire a particular structure, but to use that structure in a context-appropriate manner, such research into the role of discourse context on short-term and long-term priming, and its impact on longer-term learning has broader implications for L2 acquisition research.

In conclusion, researchers have made great strides in the last 15 years to understand how and under what circumstances structural priming can modulate L2 production and contribute to L2 learning. As highlighted in this review, the expansion of structural priming research to include within-language priming among different L2 populations (as well as cross-language priming) provides a critical means to advance our understanding of the linguistic and cognitive mechanisms that support L2 acquisition, and to inform theoretical models proposed in the L1 literature to account for structural priming more generally. However, important open questions remain, providing many avenues for future research.