by Rod Ellis
The interconnectedness of different L2 systems
In linguistics, grammar, phonology, lexis, and pragmatics are typically treated separately so it is no surprise that SLA researchers have generally adopted the same approach. However, it is clear that these different linguistic systems are interconnected in L2 development. Formulaic sequences are lexical, but they play an important role in the development of grammar (Eskildsen 2012). The acquisition of words necessarily includes acquisition of their grammatical properties (Meara 2009). The linguistic features of the pre-basic, basic, and post-basic varieties reflect the discourse strategies learners adopt (Klein and Perdue 1997). The development of syllable structure is dependent on the learners’ general proficiency, which dictates the extent to which they can focus on producing the final consonant in closed syllables (Abrahamsson 2003). Pragmalinguistic development feeds into grammatical development but is also dependent on it. What drives the acquisition of form-meaning mappings is the learners’ need to communicate more efficiently by utilizing a range of linguistic resources from all these systems. However, if communicative need can be satisfied through development of one particular system at the expense of others, as was the case with Wes, then development may be restricted.
Development is gradual, dynamic, variable, and non-linear
All the research we have examined in this chapter points to the gradual, dynamic, variable, and non-linear nature of L2 development. Learners do not move suddenly from one learner variety to another. Mastery of grammatical morphemes is a slow process and different morphemes are acquired at different rates. Nor does the acquisition of individual morphemes proceed in a straightforward way. There are periods when development is rapid, followed by a plateau, and then further development (Jia and Fuse 2007). The acquisition of negatives may manifest distinct stages of development, but these stages overlap, resulting in highly variable use of the different negative devices available at any one stage (Cancino et al. 1978). Perhaps it is usage-based analyses—such as Eskildsen’s (2012)—that most effectively capture the dynamic nature of development, as they show how forms and the uses they are put to evolve over time. They show the competition that exists between different forms and the meanings they realize, and how this is gradually sorted out.
Development is characterized by a set of universal processes
There are differences in how individual learners’ interlanguage develops. We have seen that the learner’s first language influences development—for example, in the word order that figures in the basic variety (Klein and Perdue 1997) or in the acquisition of voicing in final consonants (Eckman 1977). L2 sociopragmatic competence is heavily influenced by the pragmatic norms of the first language. However, it is also possible to identify a set of processes that are common to all learners and that suggest that, to some extent at least, development is systematic and predictable. These include:
Analysis of formulaic sequences. Throughout this chapter, we have pointed to the role that formulaic sequences play, not just in enhancing learners’ communicative ability, but also in contributing to acquisition. Where once formulaic chunks were seen as separate from the rule-systems that learners draw on in their creative speech, they are now seen as feeding into grammatical development as learners discover how to segment and recombine the parts that comprise them.
Semantic and structural simplification (i.e. the omission of content words and grammatical functors, as in ‘no colour’ (= ‘I don’t have a coloured pencil’). This is especially prevalent in the early stages of development (in the pre-basic variety), but is also likely to occur whenever learners are under pressure to communicate spontaneously and have had no opportunity to plan.
Overgeneralization (i.e. the extension of a specific linguistic form to a context that does not require it in the target language, for example, ‘eated’). Such forms do not occur in the input the learner is exposed to and thus must have been ‘created’ by the learner. In other cases, however, learners overgeneralize forms that do occur in the input, as when verb + -ing is overused to refer to habitual actions (for example, ‘he coming every day’).
Restructuring (i.e. the process where the acquisition of a new linguistic feature leads to the reorganization of existing L2 knowledge). The change that takes place does not simply involve the addition of the new feature to the learner’s interlanguage, but a qualitative re-organization of it. We saw examples of this in Eskildsen’s (2012) study of L2 negation.
U-shaped behaviour where a specific linguistic form is target-like initially, but is then replaced by an interlanguage form before the target form finally reappears (for example, ate → eated → ate). U-shaped behaviour has been observed in both the acquisition of grammatical and phonological L2 features.
Conclusion
It is clear that there are regularities in the ways in which all learners approach the task of learning a second language and that these regularities are reflected in the general trajectory observed in L2 development. As VanPatten and Williams (2007) commented, ‘learners output (speech) often follows predictable paths with predictable stages in the acquisition of a particular structure’ (p. 12). Ortega (2009) similarly concluded ‘all L1 groups will traverse the same series of approximations to the target L2 systems’ (p. 34). This chapter provides evidence to support these claims.
Some researchers, however, have questioned the existence of ‘predictable paths’. They argue that the learner’s L1 is much more influential than universalist accounts acknowledge. Larsen-Freeman (2010) wisely warned of the danger of treating all learners as behaving the same and noted that context can also affect development. More fundamentally, sociocultural theorists have queried whether acquisition orders and sequences have any validity at all. Lantolf (2005), for example, claimed that ‘development is revolutionary and therefore unpredictable’ and dismissed the view that developmental trajectories are ‘impervious to instructional intervention’ (p. 339). Johnson (2004) likewise argued that the adoption of Vygotskian sociocultural theory ‘would require that we … eradicate the assertion that L2 acquisition progresses a predetermined path’ (p. 172).
However, there is far too much evidence of these developmental trajectories to dismiss them so lightly. It is, perhaps, not so much a question of whether they do emerge in learners’ spontaneous communicative use, as examining how individual learner factors (such as the learner’s first language) and context impact on them. It is necessary to reconcile the claim that there are universal tendencies in the way a second language is acquired with the variability that is evident in learners’ use of a second language. We will consider this in the next chapter.
Notes
1 Pica (1983) argued that in considering whether or not a feature has been mastered we need to consider not just whether learners can use it accurately in obligatory occasions, but also whether they overuse it—i.e. use it in contexts when it is not obligatory. An example of overuse of third-person -s is ‘The children goes home at four o’clock’. She proposed a formula for investigating ‘target-language use’, which takes account of both suppliance in obligatory contexts and of overuse.
2 A similar problem arises with the use of emergence as the measure of acquisition. Pienemann (1984) defined emergence as the first occurrence of a specific grammatical feature in learners’ creative speech. This method also requires a clear distinction to be made between formulaic sequences (which are discounted) and creative speech.
3 This description is based on Dimroth (2012). However, the English examples illustrating the different patterns have been constructed by me.
4 Variational features were identified in Meisel, Clahsen, and Pienemann (1981), but in subsequent publications have received little attention. It is not clear to me whether there is a theoretical basis for deciding whether a feature is variational or developmental or whether variational features can only be established empirically.
5 However, there is also evidence that production can sometimes precede speech perception (Sheld
on and Strange 1982).
6 Abrahamsson’s explanation supports Skehan’s (1998) claim that learners will prioritize one aspect of L2 use (for example, fluency) over another (for example, accuracy). Abrahamsson’s study indicated that the aspect learners prioritize is determined not just by the nature of the task they are asked to perform but also by their developmental stage.
5
Variability in learner language
Introduction
In the last chapter, we saw that learners’ second language development is characterized by regularities that—to some extent at least—are universal. However, although it may be possible to identify stages in the development of specific linguistic features, there is variability within each stage. As Cancino, Rosansky, and Schumann (1978) noted in their study of the acquisition of negatives by Spanish learners of English:
Our attempt to write rules for the negative proved fruitless. The constant development and concomitant variation in our subjects’ speech at any one point made the task fruitless (p. 209)
Thus, while a specific negative device may be dominant at a given point of development (for example, ‘no’ + verb), other devices (for example, ‘don’t’ + verb; auxiliary + negative + verb) also occur. Interlanguage systems are dynamic and, as a consequence, variability is endemic. This chapter will explore the nature of this variability and also examine its significance for identifying the factors that shape L2 development.
In this chapter, I will consider three approaches for investigating variability in learner language. The first draws on variationist sociolinguistics (Labov 1970; Tarone 1983), which accounts for variability in terms of variable rules. The second makes use of the Dynamic Paradigm (Bickerton 1975; Huebner 1983) to examine variability in terms of learner varieties. The third—Dynamic Systems Theory (Verspoor, de Bot, and Lowie 2011)—examines variability within a general theory of change and development. These three approaches differ in how they view variability. In the variationist sociolinguistic tradition, ‘the concern is with discovering the underlying systematicity of variable learner production’ (Bayley 2005: 2). The Dynamic Paradigm is used to account for learner varieties in terms of the evolving form-function systems that characterize interlanguage development. In Dynamic Systems Theory, however, change is viewed as constant and the concomitant variation as largely chaotic and unpredictable.
In this chapter, I will also explore two types of variation in learner language: type 1 variation is evident when learners vary between the use of a target and an interlanguage form; type 2 variation occurs when learners alternate between two or more target language forms although not necessarily in the same way as native speakers.
Variationist sociolinguistics
L2 researchers in this tradition have drawn heavily on the work of Labov. I will briefly outline this approach before examining some of the SLA research that has been based on it. Two constructs are of particular importance in this paradigm; speech styles and variable rules. The paradigm also acknowledges that there are multiple factors that influence speakers’ variable use of linguistic forms.
Speech styles
Labov (1970) noted that ‘there are no single style speakers’ (p. 19) and that ‘all speakers vary their language to some degree in accordance with the social context or topic’. He proposed that styles can be arranged along a single dimension, measured by the amount of attention paid to speech. In other words, language users vary in the degree to which they monitor their speech in different situations, with the least attention evident in the vernacular style (i.e. the style associated with informal, everyday speech) and the most in the careful style (i.e. the style associated with formal language use, as in a public speech). That is, speakers style shift in accordance with situational factors, in particular with how they perceive their relationship with the person they are addressing, which affects their choice of linguistic forms.
Drawing on Labov’s work on style shifting, Tarone (1983) suggested that interlanguage is also characterized by a continuum of styles, each of which can be elicited by tasks varying in the extent to which they induce learners to attend to form (see Figure 5.1). An important implication of Tarone’s model is that if we want to investigate learners’ L2 competence (or ‘capability’ as she preferred to call it), we will need to use a variety of different tasks to collect data, as competence is heterogeneous, not homogeneous. Another implication, which—as we will later see—is only partly supported by research, is that learners are more likely to use target-language variants in the careful style and interlanguage variants in their vernacular style.
Figure 5.1 Style-shifting in learner language (from Tarone 1983)—see Ellis (2008: 142)
Variable rules
A categorical rule states that a specific linguistic form occurs invariably in a specific context. For example, in standard varieties of English, verb + -s occurs after the pronouns ‘he’, ‘she’, and ‘it’ in the present simple tense. Even within a single variety, however, there are some linguistic features that are variable (i.e. they can be realized by two or more linguistic forms). For example, French negation can be realized by ne + verb + pas (for example, ‘Je ne vais pas à Londres’) or by verb + pas (for example, ‘Je vais pas à Londres’). To capture such phenomena, a variable rule is needed. This states that a given variable (negation in French) is manifest as either one variant (ne + verb + pas) or another variant (verb + pas) with different levels of probability depending on the situational context. Ashby (1981), for example, reported that omission of ne occurred 61 per cent of the time in the informal speech of native speakers of French but only 40 per cent in their formal speech. However, more recent research (Armstrong 2002) showed that style-shifting in this feature is disappearing as many native French speakers now omit ne in both their vernacular and careful styles.
Linguistic context
The choice of variant, however, is not influenced solely by the situational context. The linguistic context also plays a role. That is, the variant a speaker chooses is dependent in part on the words that precede and follow the variable being investigated. For example, Armstrong (2002) reported native French speakers were likely to delete ne following a pronoun, but retain it after a full noun phrase. Research indicates that—by and large—the linguistic context exerts a stronger effect on the choice of variant than the situational context.
Multiple factors influencing variation
Romaine (2003) pointed out that ‘variation is usually conditioned by multiple causes’ (p. 431). Increasingly, researchers in the Labovian tradition have switched from investigating single factors to examining the multiple external (i.e. social) and internal (i.e. linguistic context) factors that influence the choice of variants. They have employed a statistical program called VARBRUL to determine the probabilistic weight of a whole range of factors on the operation of a variable rule. The aim is to account for the complexity and multidimensionality of speech variation. VARBRUL enables the researcher to investigate the different contributions of different factors to the choice of a specific variant (for example, omission of ne in French negatives). For those readers interested in the application of VARBRUL Young and Bayley (1996) provide a helpful account.
SLA research in the Labovian paradigm
Much of the early work on variability in SLA utilized the methodology of variationist sociolinguistics to investigate the effects of the situational and linguistic context on L2 learners’ choice of linguistic variants. More recent work has focused on identifying multiple influences using VARBRUL. For a comprehensive survey of this research see Howard, Mougeon, and Dewaele (2013). Here I will focus on a number of typical studies.
Style shifting in learner language
One of the earliest studies was Dickerson’s (1975) study of Japanese learners’ use of /z/, as in ‘boys’. Dickerson showed that the target language variant was used least frequently in free speech and most consistently when the learners read word lists aloud. Performance on a task that required learners to read di
alogues aloud was intermediate. Using a similar methodology, Bayley (1996) found that native Mandarin learners of English who had lived in the United States for varying lengths of time were less likely to delete final -t/-d in consonant clusters in their careful style (i.e. when reading aloud) than in their casual style (i.e. in informal conversation). These studies provide evidence of the same kind of style shifting found in native-speaker speech. They suggest that the attention that L2 learners pay to their speech influences the extent to which they deploy the L2 variants when speaking.
Other studies, however, have reported that style shifting in learner language does not always occur, or is not always as regular as Dickerson (1975) and Bayley (1996) found. Sato (1985) collected data over a ten-month period by means of free conversation, oral reading of a continuous text, and elicited imitation of words and short phrases. She looked at target final consonants and consonant clusters. The accuracy ranking for target final consonants on the three tasks changed from one time to the next. Greater consistency, however, was evident for consonant clusters. Sato concluded that not all variables yield the same pattern of variation.
Form-function relations also need to be taken into account to provide a full explanation of style shifting. Tarone (1985) investigated three grammatical morphemes (third-person -s, the article, and plural -s) and one grammatical structure (direct object pronouns) using data collected from three tasks designed to elicit different degrees of attention to form. In the case of third-person -s, the results demonstrated the expected pattern of style shifting: that is, the learners were most accurate in a test and least accurate in an oral narrative. However, the pattern was reversed for the article and direct object pronouns—the learners were most accurate in the task that had been designed to require the least attention to form. Tarone explained these results by suggesting that the article and object pronouns served as important markers of discourse cohesiveness in the narrative task as this led the learners to attend to discourse cohesiveness to a greater extent than in the other two tasksNOTE 1.