1. Introduction
Research on non-primary language acquisition in the generative linguistic paradigm (in recent years dubbed GenSLA (generative second language acquisition)) addresses the nature of the implicit knowledge of the second language (L2). GenSLA examines the relationship between Universal Grammar (UG), the mother tongue (L1) and the input in shaping the acquisition, representation and use of languages acquired subsequent to the L1, be it second or third language acquisition, or more. I will adopt the term L2 to cover acquisition of any language other than the mother tongue, while recognizing that language acquisition beyond L2 is not the same as L2 acquisition and has developed its own sub-field, namely, third language (L3) acquisition (see, for example, Cabrelli et al., Reference Cabrelli, Flynn and Rothman2012).
The aim of GenSLA research is to establish the underlying linguistic competence of L2 learners and L2 users (henceforth, L2ers) and how this is acquired. It is not part of the aim to provide implications for language pedagogy, although this is sometimes misunderstood; see, for example, de Bot (Reference de Bot2015). The assumption that theories must have applications involves a fundamental misconception: linguistic theories explore the nature of grammar; theories of second language acquisition explore the nature of language learning. Understanding how languages are learned does not necessarily translate into insights as to how languages should be taught. No theory of second language acquisition, generative or not, entails that language must be taught in a particular way. In fact, the same distinction is true of most of theoretical and applied sciences – theoretical physicists and chemists, for example, do not set out to propose theories with necessary applications.
Nevertheless, applications may be discovered and there is currently a growing interest in considering whether and how findings from linguistic theory and GenSLA could be applied to language teaching. Indeed, there has recently been a surge in interest in this topic (see, for example, Trotzke & Kupisch, Reference Trotzke and Kupisch2020; Whong et al., Reference Whong, Gil and Marsden2013; Widdowson, Reference Widdowson2020), even to the extent of identifying a domain known as pedagogical linguistics (Trotzke & Rankin, Reference Trotzke and Rankin2020).
In this paper, I will present an overview of research, past and present, which explores potential insights for the language classroom. I will argue that linguistic theory and GenSLA theory have more to offer in terms of considering which aspects of language might be taught rather than how languages should be taught. Research that was not intended to have pedagogical implications may turn out to do so when considered from this perspective.
2. Grammars
In the context of pedagogical issues, it is worth considering the nature of grammars. Broadly speaking, grammars can be either explicit or implicit. Three types of grammars fall into the former category: (i) descriptive grammars developed by linguists to characterize the languages that they are interested in; (ii) prescriptive grammars that are typical of grammar books, which lay down which linguistic usages are or are not desirable; and (iii) pedagogical grammars developed for use by language teachers and language learners, giving instruction and information on grammar rules for the language being learned. There is some overlap between these types of grammars: a pedagogical grammar may draw on linguistic descriptions (although I will suggest perhaps not sufficiently); pedagogical grammars may also draw on prescriptive grammars, stating that certain linguistic properties of the L2 are or are not permitted. An essential difference between descriptive grammars and the other two types is that linguists try to describe language as it is used and not how it should be used. All three types are explicit in the sense that they draw properties of language to the attention of users and are amenable to metalinguistic awareness.
Implicit grammars, in contrast, are rather different. An implicit grammar is the unconscious knowledge of language, or linguistic competence, that native speakers and language learners attain. In other words, it is a mental representation of language. In generative grammar, it is the implicit grammar that is of concern; it is not enough for linguists to concern themselves with descriptive grammars – they must go beyond this to arrive at grammars that account for underlying linguistic competence. In the L2 domain, the implicit grammar is often referred to as the interlanguage grammar. This grammar changes during the course of language development. The implicit grammars of learners are typically not identical to those of native speakers but there are many commonalities.
The implicit grammar mediates between sound and meaning. It consists of a lexicon, a phonological system, a syntactic system, morphology and semantics, as shown in Figure 1. The grammar does not operate in isolation but intersects with the linguistic context, or discourse, which often determines which structures are appropriate for a given occasion.
Figure 1. The implicit grammar
2.1 Acquiring a grammar
Learners (L1 and L2) must come up with an (unconscious) grammar for the language being acquired. They are confronted with linguistic input, which must be organized and made sense of, allowing them to understand and produce language. In generative grammar, the argument is that this is not achievable on the basis of input alone. While input is clearly essential, it does not provide sufficient information to work out all the properties of grammar that learners eventually come to know; this is the so-called poverty-of-the-stimulus or underdetermination problem. Our unconscious knowledge of language is more complex, abstract and subtle than would be expected if language was acquired solely on the basis of input.
This underdetermination problem motivates the claim for innate linguistic knowledge (in the form of UG) to explain how L1 acquirers come to acquire aspects of language not directly inducible from the input. GenSLA researchers apply the same arguments to L2 acquisition, pointing out that learners also face a poverty of the stimulus problem, acquiring properties that go beyond the L2 input and that are not based solely on the L1; consequently, they must also have access to UG (see White, Reference White2003). According to the Full Transfer Full Access Hypothesis (Schwartz & Sprouse, Reference Schwartz and Sprouse1996), when faced with L2 input, the learner starts by trying to analyse it solely in terms of the L1 grammar (=full transfer). In addition, UG comes into play (=full access). The combination of L2 input, the L1 grammar and UG yields a series of interlanguage grammars, which get revised in the course of acquisition. This situation is illustrated in Figure 2 (adapted from White, Reference White and Archibald2000).
Figure 2. The Full Transfer Full Access Hypothesis
Note that input is crucial for establishing the interlanguage grammars, in interaction with the L1 grammar and UG. Thus, identifying cases where properties of the input are potentially problematic for learners may, indirectly, provide suggestions relating to pedagogical intervention. This is an issue that we will consider below.
3. Implications relating to what to teach
In considering potential applications of linguistic theory and GenSLA to the domain of language pedagogy, I will look at the following: (i) aspects of grammar that do not or might not need to be taught; (ii) properties of grammar that might benefit from instruction, in part because of insufficiencies in the L2 input; (iii) problems with the kind of information currently provided in textbooks, which sometimes provide too much information, sometimes too little and sometimes information that is, in fact, misleading.
3.1 What we don't need to teach
A central tenet of generative linguistic theory is that there are many grammatical phenomena that cannot be learned from input alone. Accordingly, UG is invoked to explain how language acquisition is possible, at least in the case of L1 acquisition. Principles of UG predetermine certain aspects of language structure. These properties are considered to be innate and do not have to be learned on the basis of input. Assuming that UG holds for non-native acquisition as well, as researchers working in the GenSLA tradition argue, there will be properties of the L2 that fall out from UG, which will not need to be taught. In the words of Slabakova (Reference Slabakova, del Pilar García Mayo, Gutiérrez-Mangado and Martínez Adrían2013), they ‘come for free’.
An example is provided by the Overt Pronoun Constraint (OPC), a principle of UG proposed by Montalbetti (Reference Montalbetti1984). The OPC determines the distribution and interpretation of overt pronouns in languages where the subject can be left unexpressed, such as Spanish or Japanese. In these languages, overt pronouns are restricted as to the kinds of expressions that can serve as their antecedents. In particular, they cannot have quantified expressions (such as everyone, someone) or wh-expressions (such as who) as antecedents. In contrast, overt pronouns can take referential antecedents, as can null pronouns. Consider the Japanese sentences in (1) (examples from Kanno, Reference Kanno1997). In (1a), the pronoun in the embedded clause, whether overt or null, can refer to Mr. Tanaka. In (1b), on the other hand, the overt pronoun kare cannot have a wh-antecedent, whereas the null pronoun can.
(1) a. Tanaka-sani wa [∅i/karei kaisya de itiban da to] itte-iru.
Tanaka-Mr. TOP company in best is that saying-is.
‘Mr. Tanaka is saying that (he) is the best in the company.’
b. Darei ga [∅i/kare*i kuruma o katta to] itta no?
Who NOM (he) car ACC bought that said Q.
‘Who said that (he) bought a car?’
In contrast to Japanese, in languages like English that do not permit null subjects, overt pronouns are not restricted as to their antecedents, as can be seen by the translations of (1a) and (1b), both of which are grammatical. The similarities and differences in the distribution of pronouns in Japanese and English are shown in Table 1.
Table 1. Permitted antecedents for pronouns
This means that English-speaking learners of Japanese have to ‘unlearn’ the possibility of overt pronouns taking quantified/wh antecedents, an issue that is not taught in Japanese L2 classrooms and is unlikely to be noticeable in the L2 input, since it means noticing the absence of an option in a relatively infrequent construction. Studies have demonstrated that learners of Japanese are sensitive, without instruction, to this rather subtle restriction (Kanno, Reference Kanno1997; Okuma, Reference Okuma2015). They treat null and overt pronouns differently in contexts like (1b), recognizing that an interpretation that is possible in the L1 does not hold in the L2. In other words, the OPC holds in the interlanguage grammar. In the case of this example, linguistic theory has identified a universal principle and GenSLA research has demonstrated that it operates in the interlanguage grammar, supporting the idea that there are some linguistic properties that are acquired and acquirable without any kind of instruction and in the absence of input. The only input that they need to notice is that Japanese is a language that allows null arguments; once this is acquired, the OPC falls out for free. There are some aspects of language, then, that do not need to be taught; this example is by no means the only one.
3.2 What we might not need to teach
We turn now to a slightly different situation. Here, linguists have identified certain linguistic properties as being related on theoretical grounds, such as markedness relations, implicational scales or parametric clusters. L2 researchers (not necessarily working in the generative framework) have explored the possibility that, by teaching only one of a related set of properties, the others are acquired ‘for free’.
An example is provided by L2 research relating to the Noun Phrase Accessibility Hierarchy (NPAH) (Keenan & Comrie, Reference Keenan and Comrie1977), which identifies universal constraints on relative clause formation. According to the NPAH, languages vary with respect to which NP positions can be relativized; an implicational hierarchy is involved, such that clauses with relativized subjects are most accessible/least marked; see (2a). This is the only type that is found in all languages. There are many other relative clause types; the L2 research has mostly focused on two of them, namely, relativized objects (2b), and relativized oblique objects (objects of prepositions in the case of English) (2c). The implicational scale states that if a language allows relativization lower on the hierarchy, it will allow relativization higher on the hierarchy but not vice versa: the possibility of relativized obliques implies relativized objects and subjects; the possibility of relativized objects implies relativization of subjects. The possibility of relativized subjects does not imply any other type of relative clause.
(2) a. The person who saw me.
b. The person whom I saw.
c. The person whom I talked to.
It is this implication that researchers have explored as far as L2 acquisition is concerned. Several studies have shown that teaching relative clauses types that are less accessible/more marked on the hierarchy results in L2ers coming to know how to deal with relative clauses that are less marked, without specific instruction on the less marked types (Doughty, Reference Doughty1991; Eckman et al., Reference Eckman, Bell and Nelson1988; Gass, Reference Gass, Hines and Rutherford1982). In other words, learners instructed on relative clause formation relating to relativized obliques were successfully able to generalize relative clause structure to subject and object relativization but not vice versa. Teaching a linguistic property lower on an implicational scale allows other (related) properties to fall out for free. Interestingly, this involves the possibly counter-intuitive idea that teaching harder aspects of structure first is beneficial.
3.3 What we might need to teach
We turn now to some syntactic and morphological properties that GenSLA researchers have identified as potentially benefitting from specific instruction, based on analyses available in the linguistic literature.
Before turning to what might need to be taught, we must consider the role of input. Every theory recognizes the importance of input though placing different emphases on it. It is only by having a sophisticated theory of grammar that one can determine the precise nature of the input necessary to lead to grammar acquisition and change. UG, as well as the L1 grammar, will determine, in part, what aspects of the input are sufficient or insufficient. A close consideration of the input may lead to suggestions as to what linguistic properties might need to be highlighted in language learning contexts.
Broadly speaking, input can be characterized as falling into two classes, known as positive evidence and negative evidence. Positive evidence is provided by the language that the learner hears, in naturally occurring input, and is usually implicit. Negative evidence, in contrast, provides information about ungrammaticality. It is typically explicit, provided by means such as correction and grammar teaching.
In the case of L2 acquisition, the L1 grammar can make it difficult or impossible to arrive at properties of the L2 on the basis of positive evidence alone. If the L1 grammar allows only a subset of structures found in the L2, availability of such structures should become evident through positive evidence from L2 input, even if the learner initially assumes that the L2 is like the L1. In contrast, if the L1 grammar permits structures that constitute a superset of those permitted in the L2 and if there is transfer from the L1, then it is, in principle, impossible to lose the L1 structure on the basis of naturalistic positive evidence. The L2 input will be consistent with one of the possibilities found in the L1 but will not rule out the other; it will simply be non-occurring. In such cases, there is a potential role for negative evidence in the form of specific grammar teaching and correction in the classroom. This raises the question of the extent to which explicit data, in the form of grammar teaching, can affect acquisition of the unconscious, implicit system. There is ongoing debate over this question in the GenSLA literature; I assume here that explicit data can feed into the implicit system. (See Marsden et al. (Reference Marsden, Whong and Gil2018) and White (e.g. Reference White1991) for discussion; see Schwartz (e.g. Reference Schwartz1993) for counter-arguments.)
3.3.1 Reflexives
Consider an example involving constraints on the distribution of reflexive pronouns, such as himself, herself (Binding Principle A (Chomsky, Reference Chomsky1981)). There are interesting differences between English and Japanese in how reflexives work. In English, reflexives are restricted, such that their antecedents must be local, roughly speaking within the same clause, as shown in (3). In (3a), himself can refer to John, whereas in (3b) and (3c) it cannot, because the reflexive and John are not within the same clause.
(3) a. John respects himself.
b. *John thinks that Mary respects himself.
c. *John wanted Mary to respect himself.
In contrast, in Japanese, there is no such restriction; the reflexive zibun (‘self’) can take an antecedent that is ‘far away’, or non-local. It can also take local antecedents, like English. In other words, in sentences like (4) (from Hirakawa, Reference Hirakawa1990), zibun can refer to either Taro (non-local) or Akira (local).
(4) Taro-wa Akira-ga zibun-o butta to itta.
Taro-TOP Akira-Subj self-DO hit COMP said.
‘Taro said that Akira hit himself.’
The situation facing a learner of English whose L1 is Japanese is as follows. Japanese allows a superset of the sentence types permitted by English: any clause includes local clauses, and both local and non-local antecedents are possible. This means that a Japanese-speaker must restrict the options in terms of what an L2 reflexive can do. This situation is illustrated on the right-hand side of Figure 3. Assuming that a Japanese speaker transfers from the L1 to English the possibility of non-local antecedents, this would be a case where positive evidence from the L2 would be insufficient to disconfirm the L1-based analysis. It would simply confirm one of the possible analyses allowed in Japanese (the local one).
Figure 3. Locality domains for reflexives and the subset/superset problem
This problem rests on the assumption that Japanese-speaking learners of English transfer locality domains from the L1 to the L2. A number of studies have looked at this issue in L2 English, for Japanese or Korean-speakers. One example is Hirakawa (Reference Hirakawa1990). Japanese-speaking learners of English were tested on sentences like those in (3) and asked to pick from a list of possible antecedents. Non-local responses were given 23% of the time for finite clauses like (3b) and 37% of the time for nonfinite clauses like (3c), suggesting L1 influence from Japanese. Hirakawa did not explore the issue of instruction in this case but this looks like a situation where it might be worth providing explicit instruction (negative evidence) on the need for English reflexives to take only local antecedents.
Turning to the opposite situation, where English-speakers are acquiring Japanese, assuming that L2ers start off with an analysis based on their L1, the properties of the L2 are acquirable on the basis of positive evidence, as illustrated on the left-hand side of Figure 3. In other words, the learner will hear sentences like (4) uttered in a context where it is clear that a non-local antecedent is intended (Taro in this example) and can thus infer that Japanese allows a wider range of interpretations for reflexives than English.
For this situation, there has been a study addressing pedagogical implications. White et al. (Reference White, Hirakawa and Kawasaki1996) involved a teaching intervention. Learners of Japanese in an intensive summer course were instructed on various properties of zibun over a 4-week period, including the fact that zibun allows non-local antecedents. L1s included English and French, which do not allow non-local antecedents, as well as Korean and Chinese, which do. Learners were pre-tested prior to instruction, and post-tested twice on a truth-value judgment task. At the pre-test, L2ers rejected non-local binding, regardless of L1. At both post-tests (i.e. after instruction) more than half of the participants accepted non-local antecedents, suggesting an effect of instruction. Missing from this study was a group that received no instruction. Since there is positive evidence of non-local binding in the L2 input, such a group might also have come to know the relevant properties of Japanese without instruction.
In summary, linguistic theory has identified properties of reflexives that differ cross-linguistically, GenSLA research has shown that there is L1 transfer in such cases. Consequently, pedagogical intervention may be advisable, particularly in the case where the L2 grammar is in a subset relation with the L1.
3.3.2 Unaccusatives
We will now consider a different kind of example where learners persistently misanalyse the L2, and where L2ers’ difficulties do not appear to be attributable to the L1. The case involves unaccusative verbs, a sub-class of intransitive verbs (Burzio, Reference Burzio1986). Compare the examples in (5).
(5) a. Mary ran.
b. The door opened.
In (5a), the intransitive verb involves an agentive subject, who is performing the action of her own volition. In (5b), the subject of the intransitive verb is the theme of the action, not the agent. It is this class that is known as unaccusative. Unaccusative verbs are said to take an underlying object, which moves to subject position. In this respect, they are similar to passives. They differ from passives in that the verb form is active, so does not show any passive morphology. A well-known error by L2ers, regardless of L1, is to passivize unaccusatives, as in (6a) (from Zobl, Reference Zobl, Gass and Schachter1989). In contrast, L2ers do not passivize other intransitives (6b) (Oshita, Reference Oshita1997).
(6) a. My mother was died.
b. My mother was laughed.
Numerous studies have reported errors like (6a) with unaccusatives, using both corpora and experimental data. In the light of such findings, Hirakawa (Reference Hirakawa, Whong, Gil and Marsden2013) investigated whether instruction might help. Japanese-speaking learners of English were divided into two groups. One group received instruction on unaccusatives, including negative evidence relating to the ungrammaticality of passivized unaccusatives (sentences like (6a)), while the other group did not. Learners who received instruction improved in terms of recognition of ungrammaticality, though not for all types of unaccusative verbs that were tested. The uninstructed group also showed some improvement, though not as much as the instructed group.
Here, the contribution of linguistic theory has been to identify two different verb classes, and the linguistic behaviour of each class. Using this analysis, GenSLA research has provided an explanation of error types produced by L2ers, as well as why certain errors are not found, and has identified situations where instruction might be useful.
3.3.3. Morphology
Another grammatical domain where instruction might be beneficial involves morphology. Cross-linguistic differences are often found between how the L1 and the L2 map meanings onto morphological forms. Slabakova (Reference Slabakova2008) proposed the Bottleneck Hypothesis, arguing that meanings themselves are not problematic; working out how they are expressed via L2 morphology is what causes problems, creating an acquisition bottleneck in her terms.
An example of difficulties in form-meaning mapping involves English articles and the range of meanings they can convey, including definite versus indefinite, specific versus nonspecific, as well as generic, the issue that we consider here. While these meanings are available universally, languages differ as to whether or not they are explicitly realized, and how they are realized. If the L1 has no articles, such as Japanese, and the L2 requires articles, such as English, it may be hard for L2ers to work out exactly how these meanings map onto the English system, given that they are not expressed overtly in the L1.
Problems with article acquisition in L2 are well known. Here, we focus on one aspect of English articles, namely generic reference, in particular, so-called kind reference, as in the examples in (7). When referring to a class or kind, the singular definite article is possible, as in (7a), as well as a bare plural, as in (7b). Singular indefinites, on the other hand, cannot convey this kind of generic meaning; see (7c).
(7) a. The dinosaur is extinct.
b. Dinosaurs are extinct.
c. *A dinosaur is extinct.
Umeda et al. (Reference Umeda, Snape, Yusa and Wiltshier2019) conducted a study investigating the teaching of various properties of English articles to Japanese-speakers, using an acceptability judgment task. Learners were pre-tested, then instructed on English articles, then post-tested several times, the last time more than a year after the instruction. Results showed that bare plurals (as in 7b) were accepted before and after instruction, possibly due to the fact that this would be the way to express kind reference in the L1 Japanese. Following instruction, there was an increase in acceptance of definite singulars (as in 7a) although by the last post-test this had dropped back to pre-instruction levels. Instruction initially also resulted in an increase in the ungrammatical indefinite singular (7c) but acceptances eventually dropped back to pre-instructional levels. What these results suggest is that even when the properties of generics as identified by linguists are taught, the teaching is not fully effective. The bottleneck remains. Nevertheless, the linguistic description of different types of generics and the proposal for the Bottleneck Hypothesis have helped to identify particular form-meaning mapping problems relating to article use, where other kinds of instructional intervention may prove beneficial.
3.4 Pitfalls
The final set of situations that we will consider involves pitfalls that can arise because of problems in how textbooks present grammatical properties; sometimes, a linguistic analysis of what the textbook is trying to achieve can offer suggestions for improvement. We will consider cases where textbooks offer too much information, cases where they offer too little information and cases where they offer misleading information.
3.4.1 Too much information
Bruhn de Garavito (Reference Bruhn de Garavito, Whong, Gil and Marsden2013) reviewed how 15 textbooks commonly used for introductory Spanish in North America deal with object clitic pronouns, which are pronouns that attach closely to the verb, preverbally in the case of finite verbs (8a) and postverbally in the case of nonfinite verbs (8b), and that change their form depending on number and gender. (The subject is null (∅) in these examples.)
(8) ∅ Vi una blusa.
‘(I) saw a blouse.’
a. ∅ La compré.
‘(I) bought it.’
b. ∅ Quiero comprarla.
‘(I) want to buy it.’
Bruhn de Garavito notes that, in early stages of L2 Spanish, learners are overwhelmed with information about direct and indirect object clitics, including their forms and positions in the sentence. Some of this information turns out to be superfluous because a number of properties of object clitics are relatively easy to acquire; some of the information is unsuitable for the beginner level because it relates to properties of clitics in complex structures. Insights derived from GenSLA could be used to help textbook authors and language teachers determine which properties of clitics can be taught early on and which would more suitably be dealt with at later stages of proficiency, in other words to ensure that learners are not swamped with information all at once.
3.4.2 Too little information
The opposite problem is identified by Marsden et al. (Reference Marsden, Whong and Gil2018) and Gil et al. (Reference Gil, Marsden and Whong2019), who investigate the English negative polarity item (NPI), any (anything, anyone, etc.). Doing a survey of 26 textbooks, they note that properties of any are under-reported. All textbooks deal with the fact that any is required in questions and negatives, as in (9a) and (9b). None of them deals with other uses of any, such as after negative adverbs or in the case of verbs that carry negation in their meaning, as in (9c) and (9d).
(9) a. Do you want anything to eat?
b. I don't want any soup.
c. Mary ate hardly any soup.
d. She regrets having eaten anything.
In an experiment involving Chinese-speaking and Arabic-speaking learners of English, using an acceptability judgment task, these researchers found that learners accepted any in questions and negatives, in other words, those uses that they had been taught. On the other hand, they were much less likely to accept any with negative adverbs and verbs. The researchers suggest that any should not just be treated as on a par with other quantifiers such as many, much, some, and so forth, as is typically the case in textbooks. Rather, the semantics of negation should receive more attention and the various different contexts for the use of any should be presented.
3.4.3 Misleading information
Finally, we turn to potentially misleading information provided in textbooks or in the language classroom. Sometimes, explicit input (in the form of negative evidence) is provided that is not, in fact, in accordance with the intuitions of native speakers. In part, this may be an issue of providing prescriptive information that native speakers themselves ignore. Or it may be that, in the process of simplifying things for the learner, incorrect information is provided.
Bruhn de Garavito (Reference Bruhn de Garavito, Eckman, Highland, Lee, Mileham and Rutkowski Weber1995) provides a relevant example from Spanish, relating to a subjunctive rule found in L2 textbooks and explicitly taught in the classroom. According to this rule, the subject of any embedded subjunctive clause may not corefer with the subject of the matrix clause, as shown in (10a) (the subject is again null (∅) in these examples). However, there are, in fact, exceptions to this rule that are not mentioned in textbooks and not taught. In (10b), the subjunctive clause contains a modal; in (10c), it is an adjunct. In both cases, coreference is possible. If learners follow the rule they have been taught, they should consider such sentences to be ungrammatical, like (10a).
(10) a. ∅i quiero que ∅*i vaya a la fiesta.
want-1sg that go-*1 sg-subj to the party.
‘I want *me to go to the party.’
b. ∅i espero que ∅i pueda hablar con él hoy.
hope-1sg that can-1 sg speak with him today.
‘I hope that I will be able to speak with him today.’
c. ∅i voy a llamarte cuando ∅i llegue ]].
am-going-1sg to call-you when arrive-1 sg-subj.
‘I will call you when I arrive.’
Bruhn de Garavito investigated whether Spanish L2ers follow the (misleading) general rule that they are explicitly taught. Results from a truth value judgment task showed that many learners acquired the contrast in acceptability between the different kinds of subjunctive clauses. In spite of being provided with inaccurate information, these learners managed to work out the appropriate properties of the L2. However, some of the learners did not arrive at the relevant generalization, instead applying the rule that they were taught across the board, raising the question of whether they would have been successful if an accurate rule had been taught rather than the inaccurate one.
To summarize, explicit input that inadvertently introduces inaccuracies about language structure (as opposed to providing too much or too little information) is not uncommon. This suggests that authors of textbooks need to be sensitive to the subtleties of the language they are describing and wary of making overly general claims. Descriptions derived from linguistic theory can be helpful in this regard. GenSLA research has shown that, fortunately, many learners (but by no means all) are able to overcome such misleading information.
4. How to teach
So far, I have suggested that linguistic theory and GenSLA theory can shed light on which properties of language do not need to be taught and which might benefit from instruction, in other words, the what of language teaching. We turn now to the how. Here, it is much less clear whether linguistic theory and GenSLA have anything to say. Indeed, the words of Widdowson (Reference Widdowson2000, p. 28) are apposite. He points out that we are dealing with two distinct domains: ‘Looking into the relationship between … linguistic description and pedagogic design is a matter of mediating between two domains of inquiry, each with its own principles and conditions of adequacy.’
Linguistic theory and GenSLA, then, are distinct from language pedagogy and should not be expected to offer insights directly relating to language teaching. Indeed, linguists and researchers in SLA are usually not qualified to do so. However, this is often misunderstood in the applied linguistics field, as the following quotation from de Bot (Reference de Bot2015, p. 262) shows. Arguing against the UG approach to SLA, he says: ‘Ultimately the UG movement did not deliver … in terms of – crucially – how a language should be taught.’
The comments of de Bot are misconceived. It is not and never has been the role of UG theory, or any other linguistic theory, to offer insights into how languages should be taught.
5. Conclusion
To summarize, I have suggested that it is not the case that pedagogical applications of linguistic theory or GenSLA theory are necessarily to be expected. At the same time, there are cases where applications may be found, particularly relating to what aspects of language it is not necessary to teach, what aspects might be fruitful to teach, what kind of evidence (positive or negative) is helpful, and so forth. Table 2 summarizes the situations that have been discussed here.
Table 2. Possible applications
The examples in Table 2 relate to the what of language teaching. At the same time, I have suggested, with Widdowson (Reference Widdowson2000, Reference Widdowson2020) and Slabakova et al. (Reference Slabakova, Leal and Liskin-Gasparro2015), amongst others, and contra de Bot (Reference de Bot2015), that it is not the place of linguistic theory or GenSLA theory to offer suggestions as to how to teach. Linguists do not usually have expertise in language teaching, nor should they be expected to do so; language teachers cannot be expected to keep up with linguistic theories or theories of SLA. So, the question remains as to how to mediate between the two, so that useful insights can be made available in a comprehensible manner.
In conclusion, linguistic theory provides a characterization of universal and language specific properties of languages; GenSLA offers insights into underlying linguistic competence and its acquisition, including the role of UG, effects of the L1 grammar and effects of input. Insights from these domains provide potential applications for language pedagogy, including: (i) an understanding of when to expect effects of the L1 on the L2, what those effects might be and whether specific instruction would be beneficial in overcoming them; (ii) determining what properties of language to teach or not to teach; and (iii) helping to inform choices made by language teachers, textbook writers and curriculum designers. Nevertheless, uncovering applications and implications for language teaching is a bonus and not a requirement of either linguistic theory or GenSLA; as I have suggested, this bonus mostly relates to what might be taught rather than how. Theories of second language acquisition should not be judged according to their potential applications but according to what light they can shed on L2 acquisition.
Lydia White is James McGill Professor Emeritus and a Fellow of the Royal Society of Canada. She has a B.A. in Philosophy and Psychology from Cambridge University (1969) and a Ph.D. in Linguistics from McGill University (1980). She is Co-Editor of a book series (Language Acquisition and Language Disorders, published by John Benjamins) and she is on the Editorial Boards of several international journals, as well as Associate Editor of the Canadian Journal of Linguistics. Lydia White is internationally recognized as a pioneer in applying the generative linguistic framework to L2 acquisition, focusing on Universal Grammar and language transfer. She has published numerous scholarly articles and is the author of two acclaimed textbooks that have influenced the training of researchers world-wide.
