Beyond paradigm: The ‘what’ and the ‘how’ of classroom research

Abstract

This article reviews studies in second language classroom research from a cross-theoretic perspective, arguing that the classroom holds the potential for bringing together researchers from opposing theoretical orientations. It shows how generative and general cognitive approaches share a view of language that implicates both implicit and explicit knowledge, and that holds a bias towards implicit knowledge. Arguing that it is implicit knowledge that should be the object of research, it proposes that classroom research would benefit from incorporating insights from a generative understanding of language. Specifically, there is a need for a more nuanced view of the complexity of language in terms of linguistic domain, and the interaction between those domains. Generative second language acquisition research that shows developmental differences in terms of both linguistic domain and interface is reviewed. The core argument is a call for more attention to the ‘what’ of language development in classroom research and, by implication, teaching practice. As such, the language classroom is seen to offer potential for research that goes beyond paradigm to address both the ‘what’ and the ‘how’ of language development.

Keywords

acquisition classroom research cognitive explicit generative implicit knowledge language domain learning second language acquisition

I Introduction

Broadly speaking, contemporary mainstream research on language and language acquisition from a mentalist point of view can be divided into two approaches: that of generative orientation and that of general cognitive orientation. Within second language acquisition (SLA) these paradigms do not readily interact or intersect, with much scholarly work written or presented within paradigm for a within-paradigm audience. The broad aim of this article is to show that (1) there is in fact more agreement than the current polarization of paradigm suggests, and (2) the needs of the classroom can provide a unifying raison d’être for a collaborative approach to SLA capable of overcoming paradigm differences. The suggestion of cross-paradigm research does not mean that fundamental differences between schools of thought are denied. In fundamental ways, generative and general cognitive theories stand in direct opposition. After all, it is a contradiction in terms to say that language is a distinct or unique type of knowledge (the generative claim) while also being indistinct from other types of knowledge (the cognitive stance). Similarly, it would be difficult to posit grammatical principles to be both innately specified (generative) and deduced from the input (cognitive) at the same time. Despite these real differences, we look to (adult) second language (L2) development in the classroom as an area that holds potential for cross-paradigm research, with a more theory-neutral aim of informing classroom practice.

Before exploring this, however, we need to be more explicit about the areas of SLA research we will be addressing. While there is much research investigating questions of relevance to second language development and the language classroom, the concern in this article is limited to research from a mentalist perspective. Thus, we are not engaging with research exploring questions of social context (e.g. Lantolf and Thorne, 2006), identity (e.g. Block, 2007) or motivation (e.g. Dörnyei, 2001), for example. Instead we will be considering work that explores psycholinguistic development, with particular focus on those working within a generative framework and those assuming a cognitive or connectionist mental architecture. The thrust of the discussion is to argue that generative SLA (GenSLA) research has a contribution to make to classroom research, despite the fact that this approach has not been seen to be of much relevance to the classroom to date.

Indeed, while there is sometimes reference to or acknowledegment of the generative agenda in non-generative research,¹ there tends to be very little acknowledgement of non-generative concerns in GenSLA research. The reality is that the basis of GenSLA has been the very clear, if narrow, aim of exploring questions of unconscious ‘acquisition’ (as opposed to metalinguistic ‘learning’), with a concomitant focus on linguistic phenomena that are considered to be included within core competence, as defined by Chomsky (1965).² Moreover, one consistent aim of this research has been to validate the generative paradigm itself, showing evidence for innate language universals. By contrast, there is a wider range of researchers interested in language development in the classroom who tend to assume a cognitivist paradigm, though, unlike GenSLA, the primary aim of this research has not usually been to validate that paradigm necessarily, but instead to explore questions of classroom instruction. Some of the core strands in this tradition include research on different types of explicit and/or implicit instruction (e.g. Benati, 2005; Ellis et al., 2006; Izumi and Lakshmanan, 1998), corrective feedback (e.g. Carroll and Swain, 1993; Mackey, 2006; Philp, 2003) and interaction (e.g. Ellis et al., 1994; Long, 1996; Pica et al., 1987).

Considering the differing aims of these sets of research, it is perhaps not surprising that most have chosen to engage within paradigm. Even research methodology tends to vary by paradigm, making cross-paradigm debate more difficult. Yet, it is also not surprising, given that second language development is the shared object of study, that there are also areas where parallels can be drawn. One such area is to do with differing types of knowledge. In the next section we argue that there is a degree of agreement across frameworks about assumptions to do with explicit and implicit knowledge, bringing us to the conclusion that SLA research should measure implicit knowledge. We then explore the property-theoretic nature of GenSLA more fully and summarize the findings of research that shows differences in development by type of linguistic property, as well as effects of interaction between different domains of language. This is followed by consideration of two studies in terms of the desiderata for classroom research identified in this article. Our conclusion is that classroom research on how language develops should also incorporate insights from the generative focus on what develops.

II Types of second language knowledge

Whether overtly or not, all second language researchers assume a distinction between a deliberate, consciously controlled type of linguistic knowledge (termed ‘learned’ or ‘explicit’ knowledge) and an intuitive, automatic type of linguistic knowledge (‘acquired’ or ‘implicit’ knowledge), a point noted elsewhere (e.g. R Ellis, 2005). Agreeing with Ellis that there is no broad consensus in any paradigm on what the relationship is between these two types of knowledge, this section explores the concepts of acquired/implicit and learned/explicit knowledge, arguing that this is where the two paradigms might fruitfully engage with each other.

Focus on acquired knowledge is fundamental to the generative approach because of its concern with UG-constrained development of linguistic competence. One constant methodological challenge has been to minimize the effects of performance, and to avoid researching learned knowledge. The qualitative difference between acquired and learned knowledge, as articulated by Schwartz (1993), sees learned linguistic knowledge and acquired knowledge (‘competence’) as forever distinct, positing what has become known as the ‘no interface’ position between the two types of knowledge. Since first proposed, this position has not been refuted within the generative framework; and the issue no longer receives much discussion. Within cognitive research, there is also a distinction between implicit and explicit knowledge, with both types usually seen as a product of learning, and thus with no meaningful role for a process known as ‘acquisition’ that is distinct from ‘learning’, as understood by generativists.

Yet it is far from clear that these two approaches are as opposed as the use of different terminology might suggest. Consider the following quotes, the first from a leading proponent of the cognitive perspective, and the second from the original generative argument for the ‘no interface’ position:

(1) Explicit and implicit knowledge are distinct and dissociated; they involve different types of representation and are substantiated in separate parts of the brain … explicit knowledge does not become implicit knowledge nor can it be converted to it. (N Ellis, 2005: 307)

(2) … explicit data and ND [negative data] help to create another type of knowledge, one that should be seen as distinct from competence; whereas explicit data and negative data effect LLK [learned linguistic knowledge], they do not effect competence. (Schwartz, 1993: 160)

Notably, the cognitive quote (1) is even more explicit in its distinction between two types of knowledge than the no interface generative quote (2). The difference, of course, is that for N Ellis, the distinction between implicit and explicit systems applies to all types of knowledge, while Schwartz is speaking exclusively about language knowledge. Notably, however, it is not the case that within each paradigm a no interface position is held uniformly. While Ellis’s position is clear in (1), other cognitivists hold the view that explicit knowledge can become implicit over time and with practice (e.g. DeKeyser, 2007). Equally, although a strict generative no interface position assumes no interaction between types of knowledge there is a growing interest among GenSLA researchers in investigating learning in addition to acquisition (Whong et al., 2013a). While there has been no stated change in the generative position with regard to this epistemological interface, research exploring the effects of language teaching on acquisition begs the question of the relationship between learned and acquired knowledge.

Whether there is interaction between types of knowledge or not, we echo the observation of R Ellis (2005) that both paradigms posit a qualitative difference between implicit or acquired knowledge, on the one hand, and explicit or learned knowledge, on the other. Descriptions of the two types of knowledge are strikingly similar. In describing implicit knowledge N Ellis writes, ‘In fluency, both language processing and language tallying are typically unconscious; our implicit systems automatically process the input, allowing our conscious selves to concentrate on the meaning rather than the form’ (2005: 308). Even more striking, he goes on to say, ‘Implicit learning occurs largely within modality and involves the priming or chunking of representations or routines within a module’ (2005: 312). Though Ellis’s appeal to a module presumably does not refer to the Fodorian modular view of language (Fodor, 1983), which has been the mainstay of the generative approach, both paradigms assume function-specific modality that operates outside of conscious introspection as something separate from more deliberate, consciously controlled functionality. Interestingly, the passage in (2) from Schwartz (1993) continues in a way that resonates with our interest in both learned and acquired knowledge. She writes, ‘[the two types of knowledge] may, nevertheless, affect linguistic behavior, and sometimes that may be all we are seeking.’

Our emphasis on the existence of two types of knowledge within the two opposing mentalist paradigms should not be mistaken, however, for an argument that implicit and acquired knowledge are the same thing. Acquired knowledge is understood to be knowledge that is specific to the domain of language, while implicit knowledge includes a wide range of knowledge beyond the purely linguistic. Thus acquired knowledge can be seen as a subset of implicit knowledge, with the difference depending on one’s view of mental architecture. That generativists posit acquired knowledge to be a delineated subset of implicit knowledge is irrelevant to non-generativists because of the domain-general nature of the connectionist premise. Despite these fundamental differences, there is another shared point which, curiously, does not tend to be stated overtly: that implicit knowledge is in fact ‘better’ than explicit knowledge (Whong et al., 2013b). This assumption is inherent to theory-laden poverty-of-the-stimulus research that is only interested in evidence of acquisition. But in more cross-theoretic terms, the biasing of implicit knowledge is reasonable as well. After all, implicit knowledge is automatic and fast: the most basic components of fluency however defined. Additionally, implicit knowledge is assumed to be more lasting. Within a cognitive orientation, this is a result of deeper entrenchment that comes from repeated activation, while, from a generative approach, such knowledge has a privileged status as UG-constrained modular knowledge. Moreover, as explained by N Ellis above, the more that any speech act draws from implicit knowledge, the less the mechanics of language need to rely on explicit knowledge. This allows for a freeing up of deliberate processes for other things, such as consideration of meaning and context. In short, the more that knowledge of language is implicit, the more a speaker can exploit explicit mental resources for optimal delivery of thoughts and intentions.

The starting point of this discussion, then, is agreement on two points between the two mentalist approaches to SLA:

that there is a meaningful distinction between the two types of knowledge implicated in L2 development; and

that implicit knowledge is valued over explicit knowledge.

We keep these points about epistemology in mind as we turn to classroom research.

III Research on instruction: What to measure, and when

There is a large body of research that frames the question of learning in terms of the effectiveness of implicit versus explicit instruction, since the locus of interest is on what teachers can do in order to better facilitate language development among their learners or, in other words, on the ‘how’ of language development. The overall finding of the seminal meta-analysis by Norris and Ortega (2000, 2001), as well as subsequent meta-analyses (e.g. Mackey and Goo, 2007; Spada and Tomita, 2010), is clear support for the effectiveness of instruction, with indication that explicit instruction leads to larger effect sizes in aggregate than implicit.³ It is well known, however, that the nature of classroom research entails numerous methodological challenges (Doughty, 2003). In terms of the explicit/implicit distinction, there is a challenge that we will refer to as the Instructed SLA bind: the only way to show effectiveness of input is to test learner knowledge (Hulstijn, 2005). This bind can lead to a conflation of claims about input with claims about mental states because the effectiveness of different types of instruction can only be shown based on data that reflects changes in learner knowledge. If carefully made, claims about differences of instruction can be valid. But, crucially, research on type of instruction does not, in itself, allow for claims about the type of knowledge that develops as a result. In other words, explicit instruction does not necessarily lead to explicit knowledge. This might be unproblematic if there were no differences between types of knowledge. But, as has been established, SLA researchers of all persuasions accept a qualitative difference between implicit/acquired and explicit/learned knowledge.

The Instructed SLA bind is obvious from a perspective that prioritizes questions of epistemology and not instruction. Thus one advantage that could come from a generative contribution to classroom research is in controlling for this problematic bind. Moreover, we would argue that the best way to demonstrate the effectiveness of instruction is if learners can be shown to have implicit knowledge of the linguistic phenomena under study, and not just explicit knowledge. This is because of the cross-paradigm acceptance of the superiority of implicit knowledge. Of course, we would not advocate attempts to ignore explicit knowledge altogether. However, the distinction needs to be made clear and researchers should do what they can to control for the type of knowledge underlying their data through careful manipulation of research methods.

R Ellis (2002, 2005) and Housen et al. (2005) also note that studies often do not make sufficient allowance for the potential effects of explicit knowledge on measures assumed to tap implicit knowledge. Ellis (2005) sets out to validate measures of implicit and explicit knowledge through triangulation using five different tests, two identified as measuring explicit knowledge and three – two production tasks and a timed grammaticality judgement task – identified as measuring implicit knowledge. However, while any usage-based perspective would naturally place high value on production data, spontaneous production data is not a sufficient method for showing implicit knowledge, since learners can potentially draw from all sources of knowledge concurrently. Moreover, production data is, by nature, limited; regardless of the amount of the output, any set of production data, even if captured by very large corpora, will only ever be a subset of what learners, whether individually or in aggregate, can produce. In addition, production data cannot show the limits of knowledge as it can never show what is ruled out by the learner’s underlying L2 knowledge, whether in terms of grammaticality or interpretation.

Yet there seems to be a growing assumption within classroom research that production is the most valid way to study language development. Thirty-three of the 49 studies included in the original Norris and Ortega (2000, 2001) meta-analysis relied on methods that cannot show the limits of the learner’s knowledge, whether because they relied exclusively on production data or production data combined with being asked to choose the correct response from a set of options. In a more recent analysis, Spada and Tomita (2010) comment that the reliance on production data seems to be growing, and they cite Mackey and Goo (2007) who report that 52% of the measures used in their meta-analysis were ‘open-ended production measures (e.g. oral production tasks and writing tests)’. Our concern is that Spada and Tomita go on to conclude, ‘[t]hus, it would appear that SLA researchers are responding to the call for more measures of spontaneous, unanalyzed (i.e. implicit) knowledge and use’ (Spada and Tomita, 2010: 288). Given the limits of production data, we question this practice. It cannot be assumed that production necessarily captures implicit knowledge. While spontaneous production data may draw from implicit knowledge, equally, it may be scaffolded by, or be fully derived from, explicit knowledge.⁴

One existing way to increase the validity of claims about the effectiveness of instruction is the inclusion of a delayed post-test. After all, durability of knowledge, say, a year after instruction, indicates stable knowledge. Yet any researcher knows how difficult this is in practical terms. Another option is the use of on-line psychometric measures such as reaction time or eye-tracking methods (for an overview, see Roberts, 2012), or techniques in neurolinguistics such as fMRI, though access to this technology is a challenge. In sum, while classroom research shows that explicit instruction is effective, as measured by immediate post-tests, whether instruction results in increased levels of implicit (and hence, durable, ingrained, automatic) knowledge is not as clear. In sum, we argue that the distinction between implicit and explicit knowledge needs to be more robustly recognized in research design, and suggest that implicit knowledge should be the target of research, regardless of theoretical premise.

IV Linguistics and classroom research

In considering the distinction between explicit and implicit knowledge, Hulstijn notes that ‘definitions of learning – whether implicit or explicit – as a process (how) can easily become contaminated with the object of learning (what)’ (2005: 133). In addition to differences between implicit and explicit knowledge, there is the need for considerations of ‘what’ develops in terms of linguistic phenomenon.

The question of ‘what’ was the intuition underlying the meta-analysis of Spada and Tomita (2010), which asks whether the effectiveness of type of instruction depends on type of language feature, making a distinction between ‘complex’ and ‘simple’ linguistic phenomena. Re-analysing data from 41 studies, they are careful to distinguish between implicit and explicit knowledge, and to not conflate type of input with resulting knowledge; and they consider durability by factoring in short- versus long-term findings. They are also laudably upfront about the degree to which the notions of ‘complex’ and ‘simple’ are problematic. They note three different ways to define these concepts, citing psycholinguistic, linguistic and pedagogic definitions; and they point out problems with each of these definitions (2010: 268). In the end, they follow Hulstijn and de Graaff (1994) and Celce-Murcia and Larsen-Freeman (1999), opting for a linguistic definition. For them, this means defining complexity in terms of number of derivations or transformations involved in arriving at the form in question. However, this notion of ‘complexity’ is problematic. By this definition, articles are categorized as ‘simple’, for example, in contrast with the dative alternation, which is considered ‘complex’ (Spada and Tomita, 2010: 273, Table 3). It is well accepted in linguistic analyses that articles are far from simple. Knowledge of articles in English requires knowledge of definiteness, specificity and genericity (Ionin et al., 2004, 2011; Snape and Yusa, 2013), and how those properties interface with both the discourse and with other properties of grammar, such as the mass/count distinction within noun phrases. From a linguistic point of view, articles are certainly not simple.

This problem of the definition may be one reason why Spada and Tomita do not find any connection between type of linguistic phenomenon and effectiveness of teaching. Instead, their overall conclusion is to support the already known finding that there are larger effect sizes for explicit instruction over implicit. The complexity factor reveals the largest effect sizes for explicit instruction of complex language features, with complex forms enjoying larger effect sizes than simple forms – but within both the categories of explicit and implicit instruction (Spada and Tomita, 2010: 281). In other words, they are not able to show an interaction between type of instruction and type of linguistic feature, based on their definition. They suggest that their definition of complexity might have undermined their research. We concur. Though expressed in different terms, questions about type of linguistic property have driven GenSLA in the last couple of decades, in an attempt to determine which aspects of language are readily acquired, and which are not. Crucially, this research has led to an understanding of second language development in terms of the properties of language itself. In the remainder of this section we explore how the property-theoretic nature of the generative paradigm can provide the basis for exploring what should be taught.

1 Domains of language

Though debate remains about points of detail, generative researchers assume a privileged status for language-specific knowledge. Specifically, there are processes that are unique to syntax, phonology, semantics and the lexicon, respectively. Of much interest to GenSLA since the 1990s has been the difference between the domains of functional morphology and syntax. As is well known by every teacher of language, L2 learners seem to have great difficulty with production of functional morphology, whether that is inflection or free morphemes such as articles and other particles. Nonetheless, GenSLA findings have shown that, although target-like morphology may not be consistently produced, the syntax that the particular morphology is associated with, is, in many cases, intact (Lardiere, 1998a, 1998b; White, 2003). For example, syntactic theory argues that an abstract syntactic category of Tense plays a role in a range of linguistic properties, including subject–verb agreement, tense marking, nominative case on subject pronouns (e.g. She smiled vs. *Her smiled), whether or not null subjects are permitted, whether or not a copula is required, and the position of main verbs (which can be seen by whether or not they ‘raise’ in front of negation: He did not smile vs. *He smiled not).

If an L2 learner’s grammar contained no Tense category, or if the syntactic mechanisms associated with Tense were impaired in L2 acquisition, then we would expect to see random, non-target-like use in all of these areas. In fact, it has been found that for learners of different L2s, and with a variety of first languages (L1s), errors are not random, and, moreover, different patterns emerge according to the domain. Within morphology, for example, learners tend to exhibit high rates of omission (but not misuse) of inflectional morphemes that indicate subject–verb agreement or tense. At the same time, however, they are accurate on syntactic properties such as subjects, case-marking of pronouns and verb position (e.g. L2 English: Haznedar, 2001; Ionin and Wexler, 2002; and Lardiere, 1998a, 1998b; 2008; L2 French and German: Prévost and White, 2000a, 2000b). Similar results are found in the nominal domain: experimental data show that English-speaking learners of L2 Spanish are able to make use of the syntactic gender features in comprehension tasks, yet they make errors in the oral production of gender agreement morphology on determiners and adjectives (see, among others, Alarcón, 2011). Such findings suggest a disconnect between production of morphology and tacit knowledge of the underlying syntax, the source of which researchers have attributed to a problem with the process of mapping from the syntactic representation to the appropriate morphological form (e.g. Lardiere, 2008; Prévost and White, 2000a, 2000b).⁵

Work by Slabakova (2008) asks whether there is also a disconnect between morphology and meaning in L2 acquisition. Slabakova surveys a wide range of GenSLA research and concludes that there is consistent evidence that functional morphology acts as a ‘bottleneck’ in the domain of semantics, in the sense that it inhibits, but does not ultimately block, development. One example of this comes from work by Montrul and Slabakova (2002) (among others) on L2 knowledge of aspect in Spanish by English-speaking learners. English uses the simple past both for one-time past events (3a) and for habitual actions in the past (3b). Spanish, on the other hand, uses the perfective form for the former (4a) and the imperfective for the latter (4b) (examples from Slabakova, 2008: 170).

(3) a. Felix robbed a person in the street. (one-time event)

b. Felix robbed (people) in the street. (habitual)

(4) a. Guillermo rob-ó en la calle. (one-time event)

Guillermo rob-perf in the street

‘Guillermo robbed (someone) in the street.’

b. Guillermo rob-aba en la calle. (habitual)

Guillermo rob-imperf in the street

‘Guillermo habitually robbed (people) in the street.’

The participants in Montrul and Slabakova’s study completed two tasks: an inflection task that involved selecting between the imperfective and perfective forms of Spanish verbs in a set of sentences; and a judgement task, which tested knowledge of the semantic implications of the two different verb forms. In this second task, participants had to rate the acceptability of coordinated clauses, in which the first clause contained a verb in either the imperfective (5a) or perfective (5b), and the second clause was either a logical (i.e. acceptable) or contradictory (i.e. unacceptable) continuation, depending on the boundedness of the event (i.e. whether or not the event has an endpoint) in the first clause. The imperfective form yields an unbounded interpretation and the perfective a bounded interpretation, in native Spanish (Montrul and Slabakova, 2002: 131).

(5) a. La clase era a las 10 pero empezó a las 10:30. (logical)

The class was.imperf at 10 but started at 10.30

b. *La clase fue a las 10 pero empezó a las 10:30. (contradictory)

The class was.perf at 10 but started at 10.30

The key finding was that among those learners who were at least 80% accurate on the inflection task and could therefore be considered to have acquired the perfective/imperfective morphology, around half were also accurate on the judgement task and so had also acquired the semantics. However, among those who scored lower than 80% on the inflection task, almost none were accurate on the judgement task. Findings such as this support Slabakova’s (2008) conclusion that when there is a mismatch between the L1 and the L2 with regard to how meaning maps onto form, acquisition of meaning will be delayed until learners have acquired comprehension (but not necessarily production) of the form.

From this, Slabakova suggests that functional morphology is a key area in which language learners could benefit from time spent in contextualized practice. Particularly when there is a form–meaning mismatch between the L1 and the L2, the location of particular meanings in the L2 (such as the relationship between boundedness and perfective/imperfective verbal morphology exemplified for Spanish, above) may not be salient for acquisition. Practising these forms in relevant contexts may facilitate their acquisition. This suggestion brings Slabakova’s position close to non-generative voices. Slabakova, herself, notes (2008: 281) that DeKeyser (1997, 2001, 2007) has consistently argued for the role of practice in L2 development. Additionally, this research resonates with the impetus for Spada and Tomita’s (2010) meta-analysis, that the effectiveness of a particular type of instruction may vary with the type of linguistic property. If the Bottleneck Hypothesis is correct, we might expect that it is difficult to develop implicit knowledge of functional morphology. However, more reassuringly, explicit instruction could facilitate both explicit and implicit knowledge of syntactic and semantic properties of language. In sum, a more robust view of ‘what’ develops, with a more principled delineation into the domains of syntax, morphology, or semantics (leaving other domains aside for the moment), may reveal differences supporting Spada and Tomita’s original intuition that different types of instruction may have variable effects, depending on different properties. But even this more nuanced view of language does not tell the whole story.

2 Interfaces between domains

In addition to questions about the domains of language, such as functional morphology and syntax, are questions of interaction between domains. For example, phonology may interact with semantics, such that a sentence like You want this becomes a question if uttered with question intonation. Or, discourse may interact with syntax, rendering the non-canonical word order in Coffee, I like grammatical, if coffee is a discourse topic. An influential body of GenSLA research suggests that the syntax–discourse interface may be harder to acquire than other interfaces in L2 acquisition (Argyri and Sorace, 2007; Belletti et al., 2007; Sorace, 2005, 2006; Sorace and Filiaci, 2006; Sorace and Serratrice, 2009). This is interesting in the context of the L2 classroom, because discourse notions such as topic and focus do not commonly feature in classroom instruction. Yet if the syntax–discourse interface is problematic in L2 acquisition, this could be precisely where increased classroom attention could be beneficial.

Recent work by Valenzuela and McCormack (2013) aims to investigate the effect of classroom instruction on the acquisition of a complex set of interface properties found in Spanish topic–comment constructions such as (6a–b) (Valenzuela and McCormack, 2013: 103–04):

(6) a. [Context: I have a group of friends that I have known for many years.]

A Juan_i, lo_i conocí en París cuando era estudiante.

To Juan, CL I-met in Paris when was student

‘Juan, I met in Paris when I was a student.’

b. [Context: I eat fruit and vegetables in order to stay healthy.]

Manzanas, como todos los días.

Apples, I-eat all the days

‘Apples, I eat every day.’

In both examples the leftmost NP is a ‘topic’, and the subsequent clause a comment on the topic. In terms of syntax, the leftmost NPs (Juan and Manzanas) are assumed to have moved out of their canonical position in the subsequent clause (i.e. the object position of ‘meet’ in (6a) and ‘eat’ in (6b)). This movement is motivated by the context, whereby the moved NPs serve as a reintroduced topic that links to entities already introduced in the previous discourse, that is, ‘a group of friends’ in (6a) and ‘fruit and vegetables’ in (6b). Thus, topic–comment constructions involve the interface between syntax and discourse. In addition, (6a) includes a clitic pronoun lo that refers back to Juan, while such a clitic is absent in (6b). The clitic pronoun is required when the topic NP is specific. Thus, Juan in (6a) is specific (i.e. the speaker has a specific friend named Juan in mind) while manzanas in (6b) is non-specific (i.e. no specific set of apples is intended). This means that topic–comment constructions in Spanish involve the interface of syntax (movement), semantics (specificity of topic) and discourse (reintroducing given topics). For English learners of L2 Spanish, the topic–comment syntax could be transferred from their L1, but acquisition of when to use a clitic pronoun, which requires integration of the semantic and discourse properties that govern whether the given topic is specific or non-specific, represents a challenge.

Valenzuela and McCormack investigated what is taught in L2 Spanish classrooms with regard to topic–comment constructions and clitics, and then they examined some data from English-speaking learners of Spanish (from Valenzuela, 2005) in order to find out whether reflexes of the instruction could be identified in the learners’ performance. They found that topic–comment constructions are rarely taught, whereas instruction is commonly provided about clitic pronouns and their placement. However, the instruction about clitics does not include their interpretation in terms of the link between the topic and the wider context. The results of an oral sentence selection task (for comprehension) and a sentence completion task (for written production) showed that learners displayed grammatical accuracy by accepting and producing both topic–comment structures represented in (6a) and (6b). However, they tended to over-use the structure with clitic pronouns (6a) for non-specific topics and to over-accept the specific topic interpretation when the non-specific structure without clitic pronouns (6b) would have been target-like.

The results indicate that learners are aware that the structures in (6) are grammatical, but they are less successful in interpreting the structures appropriately in given contexts. In other words, classroom exposure, including explicit instruction, seems to have facilitated L2 acquisition of the structure of topic–comment constructions with and without clitics; but interpretation, which the learners have not been taught about, remains unacquired. This result resonates with other findings, mentioned above, that show persistent problems with linguistic properties in which syntax interfaces with discourse. Valenzuela and McCormack propose that classroom instruction could usefully include more focus on syntax–discourse phenomena. Specifically, they suggest that examples be provided of target constructions within a wider context, with explicit explanation of why one given construction (e.g. (6a)) is chosen over another (6b). Such instruction could facilitate understanding of required discourse notions such as topic and focus, which could lead to mastery of target properties that interface with discourse.

To summarize this section, we agree with Spada and Tomita (2010) that consideration of the type of linguistic property is important when evaluating or designing classroom L2 research. However, we suggest that insights from recent generative SLA research provide a more useful means of defining linguistic properties. With this in mind, we have explored GenSLA research findings showing that ease of acquisition of a given property appears to depend on the particular domain of language, or interface between domains. We have also showcased studies that aim to consider how such findings might be exploited in the language classroom.

V Putting it all together: Meeting the desiderata for classroom research

Earlier we argued that the key common ground between generative and non-generative approaches to SLA is the importance of acquired/implicit L2 knowledge as the target for research. We then argued that classroom research should (1) use methods that measure acquired/implicit knowledge rather than metalinguistic knowledge; and (2) pay more attention to the specific properties of language being taught and measured, with considerations of differences by linguistic domains and the interfaces between domains. A recent study by Toth and Guijarro-Fuentes (2013) shows attention to linguistic properties and recognizes the importance of measuring implicit knowledge. They ask whether L2 Spanish learners can develop implicit knowledge of the Spanish particle se if explicit instruction is given. Some transitive verbs in Spanish (7a) also allow a spontaneous, inchoative interpretation with no agent if the particle se is included (7b). This contrasts with the passive or impersonal reading that comes when se is used with accusative verbs (7c) (examples from Toth and Guijarro-Fuentes, 2013: 1168).

(7) a. Pedro secó los platos.

Pedro dried the dishes

b. Los platos se secaron.

the plates PRT dried

c. Se lavaron los platos.

PRT washed the dishes

‘The dishes were washed.’ or ‘One washed the dishes.’

L2 Spanish speakers who were explicitly taught the impersonal/passive (7c) and spontaneous (7b) uses of se were compared with a control group who were not exposed to any focus on the form. To measure implicit knowledge they use a timed judgement task. Improvement by the instructed group in correctly accepting sentences like (7b,c) at both post and delayed post-tests lead them to conclude that explicit instruction can result in the development of implicit knowledge, a result which they take as evidence against a ‘no interface’ position between learned and acquired knowledge.

In presenting their study, Toth and Guijarro-Fuentes are critical of generative researchers who have a blind commitment to a no interface position saying that this ‘has led generative L2 theorists to circular, unfalsifiable interpretations of learner data that disregard alternative explanations’ (2013: 1165). While we agree that the unchallenged assumption within much GenSLA research needs discussion, we question whether Toth and Guijarro-Fuentes are really able to show evidence that explicit knowledge can become implicit. After all, the explicit focus on se, which came with much exemplification and practice, would also qualify as positive input and thus could be argued to be the source of the subsequent development of acquired knowledge. We would not insist that this alternative interpretation is the correct analysis for their results. However, we note that even a study carried out as carefully as that of Toth and Guijarro-Fuentes faces the difficult challenge of finding unequivocal evidence for the question of the relationship between implicit and explicit knowledge.

Another study asking about the effect of instruction on implicit knowledge explored the issue using neuroimaging technology. Yusa et al. (2011) investigate L1-Japanese learners of L2 English, focusing on never inversion, whereby subject–auxiliary inversion is required when the negative adverb never is fronted for stylistic purposes, as shown in (8a). Example (8b) shows the unmarked, non-inverted form of the sentence. Example (8c) shows that, if never is fronted without subject–auxiliary inversion, the result is ungrammatical. Finally, (8d) shows that with non-negative adverbs such as today, there is no subject–auxiliary inversion.

(8) a. Never will I eat sushi.

b. I will never eat sushi.

c. *Never I will eat sushi.

d. Today I will eat sushi. / *Today will I eat sushi.

While Yusa et al. explicitly taught the monoclausal structures in (8), they tested learners on biclausal never-inversion sentences, such as (9b), which is the stylistically inverted version of (9a).

(9) a. [Those students who will fail a test] are never hardworking in class.

b. Never are [those students who will fail a test] __ hardworking in class.

c. *Never will [those students __ fail a test] are hardworking in class.

Never inversion was chosen because it is rare in the input and not usually taught. Thus, Yusa et al. could be reasonably sure that their participants had had very little, if any, exposure to this structure prior to the study. The participants were divided into two proficiency-matched groups, one of which received explicit instruction about never inversion, the other which did not. Two test measures were used: responses to an on-screen paced grammaticality judgement task, and fMRI data that were measured while the participants completed the judgement task. Both measures were administered both before and after instruction.

The key finding in both the judgement task results and the neurolinguistic data was a significant change from pre-test to post-test in the instructed group, but not in the control group. In the judgement task, the instructed group became significantly more accurate in accepting grammatical never inversion – e.g. (8a) and (9b) – and rejecting ungrammatical tokens – e.g. (8c) and (9c) – even though no instruction about the biclausal structures had been provided. In the neurolinguistic data, Yusa et al. report significant changes in activation within Broca’s area (specifically, within the left inferior frontal gyrus), which is associated with the acquisition of syntax.

Both of these studies are exemplary in many respects. The explicit challenge to the question of the relationship between explicit and implicit knowledge is welcome. And both studies show gains in implicit knowledge that come after learners are given explicit instruction. However, the fact that explicit instruction is always going to include positive input means that it is difficult to demonstrate that explicit knowledge becomes implicit. There is nothing to rule out the parallel development of two types of knowledge. Rather than conclude that this means such research is not useful, we argue that this methodological conundrum is all the more reason why researchers interested in second language development should abstract away from their theory-dependent biases. After all, it is possible to make useful claims about the effectiveness of instruction on the development of both explicit and implicit knowledge which does not depend on questions of mental architecture. Arguably, this is the kind of research that is more likely to be useful to teachers. Thus, we agree with Toth and Guijarro-Fuentes that it is time to let go of blind commitments to paradigm; and we argue that we should, instead, focus on which aspects of language seem to benefit from instruction, in conjunction with how differing types of instruction compare.

Before concluding, we note the test design of Yusa et al., which intentionally tested a construction that was not taught. While the rationale is to do with confirming the generative adherence to UG-constrained development, we wonder whether one way forward in the difficult question of researching implicit knowledge might be to test constructions that are not taught, but that implicate the same linguistic principles of a cognate structure that is taught, as was done by Yusa et al. Admittedly, this kind of approach assumes a theory-dependent view of language, but it may still provide a useful way to investigate the effectiveness of instruction.

VI Conclusions

The motivation behind this article was to discuss SLA research from two different paradigms that are rarely represented in a single article, and to consider the current state of classroom research in order to explore what might be achieved if researchers from a generative orientation engaged with the language classroom. Accepting the conclusion that the most successful language teaching is likely to include focused explicit instruction, we have argued that beyond this very general claim, SLA researchers would do well to recognize what GenSLA research has shown about development by different linguistic domain, as well as development which requires interaction between language domains. In other words, more careful attention to the nature of the linguistic property under investigation is needed. In short, it is the nature of GenSLA as a property theory (Gregg, 2001) that may provide the most useful contribution to the existing agenda of classroom research. Moreover, there are implications for pedagogy. Knowing that some aspects of language – like functional morphology – seem impervious to mastery, while others – like core word order properties – seem to be acquirable, could help a teacher decide how to approach different aspects of grammar, and what to expect from instruction.

We have also argued that it is important for SLA researchers to avoid the instructed SLA bind so that type of instruction is not erroneously conflated with type of resulting knowledge. Moreover, we prioritize the testing of implicit knowledge rather than explicit knowledge as the former is understood across frameworks to be ‘better’. It is acknowledged, however, that isolating implicit knowledge is very difficult in terms of research methodology, especially for the vast majority of researchers who do not have access to brain scanning equipment.⁶ We have also noted the difficulty of controlling for positive evidence when giving explicit instruction. However, given that neither theoretical stance has yet to determine the relationship between implicit and explicit knowledge, we find this question an area that warrants a collaborative approach to classroom research that goes beyond paradigm. In sum, we argue for an approach to classroom research in which ‘what’ is added to the question of ‘how’ language develops. The hope is that by placing the needs of the classroom first, we may find a shared focus that allows researchers to set aside paradigm differences and pursue a research agenda with the potential of providing some useful guidance for teachers facing the onerous task of teaching language.

Footnotes

Acknowledgements

Melinda Whong is the first author of this paper, as the initiator of the idea for, and structure of, the article. All three authors shared in the development and writing of the article. We are grateful to an anonymous reviewer for helpful comments, and to Margaret Thomas for constructive feedback and excellent author care.

Declaration of Conflicting Interest

The author declares that there is no conflict of interest.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Notes

References

Alarcón

(2011) Spanish gender agreement under complete and incomplete acquisition: Early and late bilinguals’ linguistic behavior within the noun phrase. Bilingualism: Language and Cognition 14: 332–50.

Argyri

Sorace

(2007) Crosslinguistic influence and language dominance in older bilingual children. Bilingualism: Language and Cognition 10: 79–99.

Belletti

Bennati

Sorace

(2007) Theoretical and developmental issues in the syntax of subjects: Evidence from near-native Italian. Natural Language and Linguistic Theory 25: 657–89.

Benati

(2005) The effects of processing instruction, traditional instruction and meaning–output instruction on the acquisition of the English past simple tense. Language Teaching Research 9: 67–93.

Block

. (2007) Second language identities. London: Continuum.

Carroll

Swain

(1993) Explicit and implicit negative feedback. Studies in Second Language Acquisition 15: 357–86.

Celce-Murcia

Larsen-Freeman

(1999) The grammar book. Boston, MA: Heinle and Heinle.

Chomsky

(1965) Aspects of the theory of syntax. Cambridge, MA: MIT Press.

DeKeyser

(1997) Beyond explicit rule learning. Studies in Second Language Acquisition 19: 195–221.

10.

DeKeyser

(2001) Automaticity and automatization. In Robinson

(ed.) Cognition and second language instruction. Cambridge: Cambridge University Press, 125–51.

11.

DeKeyser

(2007) Skill acquisition theory. In VanPatten

Williams

(eds) Theories in second language acquisition. New York/Abingdon: Routledge, 97–113.

12.

Dörnyei

(2001) Teaching and researching motivation. Harlow: Longman.

13.

Doughty

(2003) Instructed SLA: Constraints, compensation, and enhancement. In Doughty

Long

(eds) The handbook of second language acquisition. Oxford: Blackwell, 256–310.

14.

Ellis

(2005) At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in Second Language Acquisition 27: 305–52.

15.

Ellis

(2002) Does form-focused instruction affect the acquisition of implicit knowledge? A review of the research. Studies in Second Language Acquisition 24: 223–36.

16.

Ellis

(2005) Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition 27: 141–72.

17.

Ellis

Loewen

(2007) Confirming the operational definitions of explicit and implicit knowledge in Ellis (2005): Responding to Isemonger. Studies in Second Language Acquisition 29: 119–26.

18.

Ellis

Loewen

Erlam

(2006) Implicit and explicit corrective feedback and the acquisition of L2 grammar. Studies in Second Language Acquisition 28: 339.

19.

Ellis

Tanaka

Yamazaki

(1994) Classroom interaction, comprehension, and the acquisition of L2 word meanings. Language Learning 44: 449–91.

20.

Fodor

(1983) The modularity of mind. Cambridge, MA: MIT Press.

21.

Gregg

(2001) Learnability and second language acquisition theory. In Robinson

(ed.) Cognition and second language instruction. Cambridge: Cambridge University Press, 152–80.

22.

Gutiérrez

(2013) The construct validity of grammaticality judgment tests as measures of implicit and explicit knowledge. Studies in Second Language Acquisition 35: 423–49.

23.

Hawkins

(2005) Revisiting wh-movement: The availability of an uninterpretable [wh] feature in interlanguage grammars. In Dekydtspotter

Sprouse

Liljestrand

(eds) Proceedings of the 7th Generative Approaches to Second Language Acquisition Conference (GASLA 2004). Somerville, MA: Cascadilla Proceedings Project, 124–37.

24.

Hawkins

Chan

(1997) The partial availability of Universal Grammar in second language acquisition: the ‘failed functional features hypothesis’. Second Language Research 13: 187–226.

25.

Haznedar

(2001) The acquisition of the IP system in child L2 English. Studies in Second Language Acquisition 23: 1–39.

26.

Housen

Pierrard

Van Daele

(2005) Rule complexity and the efficacy of explicit grammar instruction. In Housen

Pierrard

(eds) Investigations in instructed second language acquisition. Amsterdam: Mouton de Gruyter, 235–69.

27.

Hulstijn

(2005) Theoretical and empirical issues in the study of implicit and explicit second language learning. Studies in Second Language Acquisition 27: 129–40.

28.

Hulstijn

de Graaff

(1994) Under what conditions does explicit knowledge of a second language facilitate the acquisition of implicit knowledge? A research proposal. AILA Review 11: 97–112.

29.

Ionin

Wexler

(2002) Why is ‘is’ easier than ‘-s’?: Acquisition of tense/agreement morphology by child second language learners of English. Second Language Research 18: 95–136.

30.

Ionin

H-J

Wexler

(2004) Article semantics in L2 acquisition: The role of Specificity, Language Acquisition 12: 3–69.

31.

Ionin

Montrul

Santos

(2011) An experimental investigation of the expression of genericity in English, Spanish and Brazilian Portuguese. Lingua 121: 963–85.

32.

Isemonger

(2007) Operational definitions of explicit and implicit knowledge: Response to R Ellis (2005) and some recommendations for future research in this area. Studies in Second Language Acquisition 29: 101–18.

33.

Izumi

Lakshmanan

(1998) Learnability, negative evidence and the L2 acquisition of the English passive. Second Language Research 14: 62–101.

34.

Lantolf

Thorne

(2006) Sociocultural theory and the genesis of second language development. New York: Oxford University Press.

35.

Lardiere

(1998a) Case and tense in the ‘fossilized’ steady state. Second Language Research 14: 1–26.

36.

Lardiere

(1998b) Dissociating syntax from morphology in a divergent L2 end-state grammar. Second Language Research 14: 359–75.

37.

Lardiere

(2008) Feature assembly in second language acquisition. In Liceras

Zobl

Goodluck

(eds) The role of formal features in second language acquisition. New York: Lawrence Erlbaum, 106–40.

38.

Long

(1996) The role of the linguistic environment in second language acquisition. In Ritchie

Bhatia

(eds) Handbook of second language acquisition. New York: Academic Press, 413–48.

39.

Mackey

(2006) Feedback, noticing and second language development: an empirical study of L2 classroom interaction. Applied Linguistics 27: 405–30.

40.

Mackey

Goo

(2007) Interaction research in SLA: A meta-analysis and research synthesis. In Mackey

(ed.) Conversational interaction in second language acquisition: A series of empirical studies. Oxford: Oxford University Press, 407–52.

41.

Montrul

Slabakova

(2002) Acquiring morphosyntactic and semantic properties of aspectual tenses in L2 Spanish. In Pérez-Leroux

Liceras

(eds) The acquisition of Spanish morphosyntax: the L1–L2 connection. Dordrecht: Kluwer, 113–49.

42.

Morgan-Short

Steinhauer

Sanz

Ullman

(2012) Explicit and implicit second language training differentially affect the achievement of native-like brain activation patterns. Journal of cognitive neuroscience 24: 933–47.

43.

Norris

Ortega

(2000) Effectiveness of L2 instruction: A research synthesis and quantitative meta-analysis. Language Learning 50: 417–528.

44.

Norris

Ortega

(2001) Does type of instruction make a difference? Substantive findings from a meta-analytic review. Language Learning 51: 157–213.

45.

Philp

(2003) Constraints on ‘noticing the gap’. Studies in Second Language Acquisition 25: 99–126.

46.

Pica

Young

Doughty

(1987) The impact of interaction on comprehension. TESOL Quarterly 21: 737–58.

47.

Prévost

White

(2000a) Accounting for morphological variation in second language acquisition: truncation or missing inflection: The acquisition of syntax. London: Longman.

48.

Prévost

White

(2000b) Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research 16: 103–33.

49.

Roberts

(2012) Psycholinguistic techniques and resources in second language acquisition research. Second Language Research 28: 113–27.

50.

Sanz

Leow

(2011) Implicit and explicit language learning: Conditions, processes, and knowledge in SLA and bilingualism. Georgetown University Round Table on Language and Linguistics Series. Washington, DC: Georgetown University Press.

51.

Schwartz

(1993) On explicit and negative data effecting and affecting competence and linguistic behavior. Studies in Second Language Acquisition 15: 147–62.

52.

Slabakova

(2008) Meaning in the second language. Berlin: Mouton de Gruyter.

53.

Snape

Yusa

(2013) Explicit article instruction in definiteness, specificity, genericity and perception. In Whong

Gil

K-H

Marsden

(eds) Universal Grammar and the second language classroom. Dordrecht: Springer, 161–83.

54.

Sorace

(2005) Syntactic optionality at interfaces. In Cornips

Corrigan

(eds) Syntax and variation: Reconciling the biological and the social. Amsterdam: John Benjamins, 46–111.

55.

Sorace

(2006) Gradience and optionality in mature and developing grammars. In Fanselow

Fery

Schlesewsky

Vogel

(eds) Gradience in grammars: Generative perspectives. Oxford: Oxford University Press, 106–23.

56.

Sorace

Filiaci

(2006) Anaphora resolution in near-native speakers of Italian. Second Language Research 22: 339–68.

57.

Sorace

Serratrice

(2009) Internal and external interfaces in bilingual language development: beyond structural overlap. International Journal of Bilingualism 13: 195–210.

58.

Spada

Tomita

(2010) Interactions between type of instruction and type of language feature: A meta-analysis. Language Learning 60: 263–308.

59.

Toth

Guijarro-Fuentes

(2013) The impact of instruction on second-language implicit knowledge: Evidence against encapsulation. Applied Psycholinguistics 34: 1163–93.

60.

Valenzuela

McCormack

(2013) The syntax–discourse interface and the interface between generative theory and pedagogical approaches to SLA. In Whong

Gil

K-H

Marsden

(eds) Universal Grammar and the second language classroom. Dordrecht: Springer, 101–14.

61.

White

(2003) Fossilization in steady state L2 grammars: Persistent problems with inflectional morphology. Bilingualism: Language and Cognition 6: 129–41.

62.

Whong

Gil

Marsden

(2013a) Universal Grammar and the second language classroom. Dordrecht: Springer.

63.

Whong

Gil

Marsden

(2013b) How we can learn from acquisition: The acquisition-learning debate revisited. Proceedings of the 12th Generative Approaches to Second Language Acquisition Conference (GASLA 2013). Somerville, MA: Cascadilla Proceedings Project. In Cabrelli Amaro

Judy

Pascual y Cabo

(eds), 203–210.

64.

Yusa

Koizumi

Kim

Kimura

Uchida

Yokoyama

Miura

Kawashima

Hagiwara

(2011) Second-language instinct and instruction effects: Nature and nurture in second-language acquisition. Journal of Cognitive Neuroscience 23: 2716–30.