Becoming multilingual: The macro and the micro time perspective

1 Introduction

In research on the acquisition of languages in individuals, two major points of view can be distinguished. One is the concern with the speaker’s multilingual competence and practices, how this develops and changes shape over time, the stages and sequences of acquisition, etc.

The other focuses on the specific events of language use, the speaker’s options and activity, and the conditions and factors that govern the acquisitional process in such situations. In both cases, we are dealing with states and processes which are located in time. Looking closer at some of the different aspects involved, I will discuss the two viewpoints here in terms of two different time levels. At one level we have to do with the development of a person’s competence in the languages over a shorter or longer period of time, maximally the lifespan, and the nature of the evolving linguistic repertoire. This is what I will call here the macro perspective. The other level comprises the processes that take place in specific situations of language use and their governing conditions, which may lead to the acquisition of pieces of linguistic knowledge and skills and thus contribute to the development over time. Focusing on this latter level is what I call the micro perspective.

It is obvious that different aspects of language development proceed at different speeds, and thus different timescales apply for the various kinds of processes and change in language (de Bot 2012, 2015; Lemke 2000). In the words of de Bot (2012: 143), “ … language development takes place at many interacting timescales, ranging from the milliseconds of lexical retrieval to pragmatic development at the scale of the life span.” The distinction of two basic time levels as defined here naturally recognizes this variation. The focus is somewhat different, however. While it is understood that different processes in language take place at a variety of timescales, the point here is to foreground the different perspectives in studying language use and development, where the concern with the speaker’s activity in situations involves different issues than the concern with the person’s linguistic repertoire and its development.

The topic of the following pages is to demonstrate and discuss how the two levels differ in the aspects that they involve, and how they connect to each other. At each of the levels, current research on multilingualism and third language acquisition presents certain problems which will be addressed in the following. One case in point is the problematic use of the constructs first, second and third language in current literature (Hammarberg 2014), a problem which can be more clearly understood if the two different perspectives are taken into account.

2 The general orientation

A person’s knowledge and use of language involves variation, continuous adaptation, creative action and development. I will look upon these processes essentially from a functional and cognitive point of view. When dealing with the individual’s language development at the macro and micro levels, certain notions appear fundamental. In particular, the following three points should be mentioned as a basic frame of reference for the subsequent sections.

One point is the adoption of a usage-based approach (Langacker 1987, 1988, 1999; see also Bybee 2010; Croft and Cruse 2004; Kemmer and Barlow 1999). According to this view, a speaker’s linguistic system develops from recurrent events of language use (interaction, comprehension, production) in specific situations. Language structure emerges when chunks of speech are identified by repeated occurrence and get established in the speaker’s memory as units of language. Speakers are sensitive to the frequency of use. The repeated activation of linguistic items in communication will lead to their gradual entrenchment, i. e. cognitive routinization and automatization. Languages are not construed as given sets of rules to be discovered by the learner, but adaptable systems of linguistic patterns emerging and developing from the speaker-listener’s experience of concrete usage events, i. e. actual instances of language use (Langacker 1999: 9). Grammar is seen as “a network built up from the categorized instances of language use” (Beckner et al. 2009: 5). It is an ongoing process of shaping, maintaining and reshaping language structure which is thought to characterize a speaker’s language competence both in its buildup stages and in stages of mature proficiency.

A second point is the idea that language use and competence forms a developing complex system, which is dynamic and variable and contains interacting elements and forces (see e. g. Ellis and Larsen-Freeman 2009; Larsen-Freeman and Cameron 2008; Verspoor et al. 2011). Systems consist of subsystems and are themselves part of a larger system. Thus, for example, a person’s repertoire of languages constitutes a complex system, and so do the various language competences forming the repertoire; similarly, a speech situation forms a complex system at the micro time level, as well as the many interacting elements that are involved in the situation. A usage-based approach fits in with this framework. Language is seen as an adaptable system which is socially situated: it develops through the user’s interaction with the social environment, especially with other language users, as well as through internal self-organization. There is no fixed end state of development. (Parts of) a system may settle for a shorter or longer period into attractor states, such as a state of fossilization. Development is driven by use of the system; a change may involve not only growth but also decline (such as attrition of language competence due to an extended period of non-use).

A third point is the awareness that multilingualism is the normal form of (mature) linguistic competence in humans. I define the terms mono-, bi- and multilingualism here as the knowledge of one, two and three and more languages, respectively, at some significant level of proficiency. It should be noted that this terminology draws a distinction between bi- and multilingualism, and that the terms do not presuppose a level or type of proficiency in each language. In the course of life, people tend to pick up knowledge of the languages with which they get into active contact. In this process, monolingualism and true bilingualism represent intermediate or preliminary stages of a potential multilingual development. Adult monolingualism occurs as a consequence of limited language contact, typically with speakers of a large and dominant majority language, or else with people who keep living in a local, more secluded language environment. As many have pointed out, individual multilingualism is extremely widespread in the world’s population, and is found to be increasing in the modern society with its globalization, mobility and fast communication (see Aronin and Singleton 2008, 2012). Judging by what we can observe, the multilingual potential appears to be a natural and characteristic part of the human mental equipment (Hammarberg 2010, 2014). Multilingualism constitutes a clear example of a developing complex system, with its embedded and interacting subsystems.

3 Macro time

Multilingual development in macro time involves the buildup of a repertoire of languages. Developing a multilingual repertoire in an individual includes adding knowledge of new languages over time, but also other aspects such as

1. Changes in level of proficiency in the respective languages: gradual growth, also attrition

2. Changes in depth and type of knowledge, familiarity with variation (dialects, sociolects, genres, styles, trends etc.), familiarity with the cultural context of each language and the social norms governing its use

3. Comparative knowledge of different languages

4. Metalinguistic awareness

5. Experience of language use in situations

6. Experience and skill in learning languages; strategic skills in using and acquiring language

7. Attitudes, affective relations to specific languages

   etc.

The notion of a linguistic repertoire has been defined from a sociolinguistic point of view by Gumperz (1972: 20) as “the totality of linguistic resources (i. e. including both invariant forms and variables) available to members of particular communities”. Whereas Gumperz refers primarily to language communities, this concept of a repertoire has also been applied to the language knowledge of individuals (Blommaert and Backus 2013). This latter approach is the one taken here.

It is obvious that not all languages in a repertoire develop to the same extent, or in the same mode of learning, or to the same resulting form. Blommaert and Backus (2013: 17) distinguish broadly between ‘comprehensive’ language learning, leading to a ‘maximal’ set of resources in the respective languages, ‘specialized’ language learning which produces various kinds of more restricted or partial forms of linguistic knowledge and skills, and ‘encounters’ with language which refers to learning “very small bits of language” such as single words, slang expressions or fragments picked up by temporary contacts. All of these cases of language learning have roles to play in a person’s total repertoire. According to Blommaert and Backus, even the mere ability to recognize and identify languages counts as parts of the repertoire, irrespective of any competence to use the languages in question. Informal learning of a language yields different knowledge than formal, instructed learning; often both types occur in combination. In a description of one highly multilingual speaker’s repertoire, Blommaert and Backus (2013) sort the languages into four large categories: maximum, partial, minimal and recognizing competence. It is important to be aware, however, that there are no sharp borderlines between such categories.

The L1/L2 distinction is one basic dimension which relates to the way a language becomes internalized and stored by the user. In standard usage in SLA research, the terminological distinction between a first and a second language (L1, L2) is based on whether the language has been acquired from early infancy on (a native language), or added at a later stage (a non-native language). Due to the fact that a person may have more than one L1 and more than one L2, these terms have lost their literal sense of the chronologically first and second language acquired by the speaker. The difference is related to the stage of cognitive maturity that changes with age during early childhood. In several works, Paradis (2004, 2008, 2009) discusses the neural basis for the distinction between an L1 and an L2. He distinguishes between implicit linguistic competence in L1 and explicit metalinguistic knowledge of L2. The former is acquired incidentally, stored implicitly, used automatically, sustained by procedural memory, and involves different parts of the brain than the latter, which is learned consciously, stored explicitly, consciously controlled when used, and sustained by declarative memory (Paradis 2008: 343). ^[1] Paradis thus posits two distinct mechanisms for L1s and L2s, respectively, but also emphasizes that the explicit knowledge of an L2 may become automatized, i. e. replaced by implicit competence, through practice.

The distinction betweeen an L1 and an L2 that Paradis draws is one between any native and any non-native language, which means that different non-native languages are assumed to have a neurolinguistic affinity with one another that is not shared with native languages. Bardel and Falk (2012), in examining the implications of Paradis’ L1/L2 distinction for L3 learning, emphasize this affinity between non-native languages and propose that Paradis’ theory can offer a neurolinguistic explanation for the influencing effect of prior L2 knowledge on the learning of an L3 (the L2 status effect). The dichotomy of L1 versus L2 is standard usage in the SLA research tradition, but when multilingual users and learners come in focus and a more differentiated set of terms is needed, some confusion tends to arise in the literature. One practice which is common in dealing with multilinguals is to order the speaker’s languages chronologically with respect to the start of acquisition of each language, using ordinal numbers: first, second, third, fourth language etc. (L1, L2, L3, L4, …). Although this terminology refers to the buildup of a speaker’s multilingual repertoire, it has turned out to be problematic in several respects (Hammarberg 2010, 2014). Firstly, it is not compatible with the cognitively based standard use of the terms L1 and L2 in SLA research which we just discussed; these terms then become ambiguous.

Secondly, a time ranking by ordinal numbers may be difficult or even impossible to apply. A basic problem is that we have to know which languages to count as parts of a person’s repertoire before languages can be given ordinal numbers. Is there a certain level of proficiency that has to be attained in order for a language to be counted? How should various kinds of partial or specialized knowledge of a language be judged (such as, for example, mere reading comprehension, or speaking but not writing ability, or a “basic variety” type of competence in the sense of Klein and Perdue 1997, or purely metalinguistic knowledge)? Should even fragmentary pieces of knowledge be considered (as Blommaert and Backus 2013 do)? What about “bonus languages” (Hammarberg 2010: 94), i. e. languages that are so close and similar to a language the person knows well that they are by and large comprehensible, like for example Danish and Norwegian to a speaker of Swedish? Should these rather be regarded as extended cases of the speaker’s understanding of dialects in the better known language? Related to these problems of inclusion/exclusion is the problem of sequencing. There is no ideal way of handling simultaneous acquisition. The practice of chronological numbering suggests a strictly sequential series of one language at a time. Languages introduced simultaneously will have to receive the same ordinal number, leaving subsequent gaps in the series (e. g. L1+L1, L3, …, or L1, L2+L2, L4 … ; cf. the list of various such combinations given in Cenoz 2000: 40). The ordering by the first encounter does not take into account the common case that different languages are used in parallel over time and continue to develop simultaneously in the user’s repertoire even if the acquisition has started at different points in time. Sequencing can be especially difficult in cases of intermittent acquisition. These are cases where a speaker first comes into limited contact with a new language, and then later in life, after acquiring one or more other languages in the meantime, renews the contact with the former language and acquires it more comprehensively. Which point in time should then be regarded as the starting point for this language?

Thirdly, as these various questions show, the practice of chronological sequencing by ordinal numbers strongly oversimplifies the picture of how multilingual competence develops. It is not general enough to fit all multilingual speakers well and will not capture the complexity of developing multilingual repertoires.

Summarizing my observations on the chronological ordering model, it (1) is incompatible with the standard definition of L1 and L2 in the SLA research tradition, leading to ambiguous use of these terms, (2) involves problems of inclusion/exclusion and sequencing which can make the ordering scale difficult to apply, and (3) oversimplifies the conception of multilingual development. Its frequent use in the literature may be due to its apparent usefulness as a simple historical “skeleton”, a schematic overview of the speaker’s contact with different languages. However, it focuses on the start of acquisition of each language, rather than the time periods of their use which seems a more representative way of locating a person’s languages chronologically. In sum, these various observations cast doubt on the value of the chronological numbering practice. Rather, it seems better to stay with the cognitively grounded definitions of the categories L1 and L2 as used in SLA studies even when focusing on multilingual repertoires. The presence of more than one L1 and/or L2 is then a natural part of the picture. The use of the term L3 is an issue to which we shall return below.

It is essential to keep in mind that the chronology of an individual’s contact with different languages relates to the development of a repertoire in macro time. The roles of the different languages in the processes of use and acquisition in micro time are a different matter, to which we now turn.

4 Micro time

Assuming a usage-based model of language, the development of a linguistic repertoire rests on the speaker’s experience of events of language use in specific situations. Conversely, the processes of acquisition in such situations are dependent on the current state of the speaker’s linguistic repertoire. Put differently, the processes in micro time both feed and draw on the development in macro time.

Research on language acquisition in multilinguals has made a differentiation of the traditional concept of second language (L2) necessary, in particular since it has become clear that the acquisition of a new language differs significantly depending on whether or not the learner has prior experience of acquiring some other L2. For example, there is today ample evidence that both L1s and prior L2s can be sources of cross-linguistic influence, and one line of research has been to investigate the roles and relative significance that the learner’s different languages have in this respect (cf. Falk and Bardel 2010; Williams and Hammarberg 2009 [1998]). Focus on multilingual settings has also brought up important issues of linguistic awareness (Jessner 2006), use of strategies (Mißler 1999, 2000), multilingualism as an asset in language learning (Cenoz 2003), and other aspects. In the rapidly expanding literature on third language acquisition (TLA), the term L3 has become an established concept. As mentioned above, the terms L1, L2, L3 are frequently used here in a purely chronological sense, referring to a simple sequential order of acquisition. However, as we saw in Section 3, this practice has the drawback of being simplistic and of deviating from common usage in SLA research. A definition of L3 which is grounded in the cognitively based conception of L1 and L2 that is common in SLA research is the following:

In dealing with the linguistic situation of a multilingual, the term third language (L3) refers to a non-native language which is currently being used or acquired in a situation where the person already has knowledge of one or more L2s in addition to one or more L1s.

(Hammarberg 2010: 97, italics in original.)

Defined in this way, L3 is a concept which is specifically related to the level of micro time processes. An L3 is identified in relation to a specific situation of language use. The concepts of L1 and L2, on the other hand, refer to languages characterized by the way in which they are internalized by the speaker and stored as parts of the developing repertoire, as described above. In a communicative situation, the speaker’s current knowledge of (one or more) L1s and L2s may become activated and influence the processes of language use and acquisition (cross-linguistic influence, transfer). Thus, L1s and L2s are parts of a complex system both at the macro and the micro time level. An L3 is then one of the multilingual speaker’s non- native languages (L2s) which occupies the role of the currently used language in the situation.

This yields a three-way characterization of the languages involved in situations of use: L1(s) and prior L2(s) established in the repertoire and functioning as background languages in the situation, and the L3, being the language currently in use.

One observation that motivates TLA research is that learners of an L3 are differently equipped for the task than those who learn their first L2. In the framework known as the Factor Model, Hufeisen (1998, 2005, 2010 and elsewhere) posits bundles of interacting factors which exert an influence on the acquisition/learning of L1, (the first) L2 and L3. ^[2]

According to this model, new factors are added step by step, thus changing the total set of factors that influence the learning of a new language. Some factors apply already to L1 acquisition; these are neurophysiological factors, (general language acquisition capability, age etc.) and learner external factors (learning environment, types of input etc.). In L2 learning some further factors are added: emotional factors (motivation, anxiety, assessment of own language proficiency, attitudes, individual life experiences, etc.), and cognitive factors (language awareness, learning awareness, learning strategies, individual learning experiences, etc.), as well as the prior knowledge of L1. The distinctive characteristic of L3 learning is the adding of foreign language specific factors (individual foreign language learning experiences and strategies, interlanguages of previous languages and target language), plus the updated linguistic factors, viz. the prior knowledge of L1 and L2. Some further discussion of these factors is contained in Hufeisen (2003).

The set of factors is thus found to expand from L1 to L2 to L3, successively adding different types of factors, which shows that there are qualitative differences in the conditions that apply to these three steps. On the other hand, if more languages are learned after the L3, this does not involve further types of factors; thus there is a three-level hierarchy in the acquisition/learning of languages: L1, (first) L2 and subsequent languages. The original model presupposes a sequential order of adding languages, one at a time. Since the aim is to point out the different conditions that apply when a learner encounters new languages, the focus is specifically on the initial stage for each language. Later versions (especially 2010) also to some extent consider simultaneous acquisition and the accumulated interlanguage of the target language. Hufeisen (2010: 200) points out that the model is in a continuous state of revision depending on our changing state of knowledge about the factors influencing language learning.

The focus on the initial stage for each language and the fact that the languages are assumed to be encountered in linear succession sets a limit to the scope of the model. This obviously has to do with its main purpose of demonstrating the introduction of new types of influence on the learning process which appear with the first L2 and then with L3. The model was devised primarily with a view to foreign language learning, and practical implications for the teaching of an L3 as the second foreign language have been explored elsewhere (e. g. Hufeisen and Marx 2007; Marx and Hufeisen 2010; see also Jessner 2008).

The question arises whether the Factor Model as proposed by Hufeisen can be extended into a situation-related model that covers all possible cases and forms of language acquisition or learning and kinds of linguistic knowledge: foreign language learning as well as informal second language acquisition, different stages of proficiency, different types of knowledge of a language, successive, simultaneous as well as intermittent acquisition, and – importantly ‒ representing the parallel knowledge of the languages that exist currently in the repertoire.

Figure 1 proposes a general template for a variable model of language use by a potentially multilingual user/learner. The processes of language use and acquisition in the situation take place in the individual who interacts with his/her current linguistic environment as he/she receives and/or produces utterances in context. The interaction may be oral or written (or take place by other means, such as signing). The boxes around the processing box contain bundles of factors which together influence the process. Those to the left contain contextual knowledge (encyclopaedic “knowledge of the world”, understanding of the situation, awareness of the preceding discourse), and language knowledge (one or more L1s, L2 knowledge if any – parentheses in the figure indicate ‘if occurring’). The boxes to the right are at the present stage assumed to be essentially those of Hufeisen’s Factor Model. They will have to be further elaborated and be open to revisions. The extended scope of the present model will entail some modifications in how various factors apply. At this point, the details of this will have to be a topic for future discussion. But my assumption is that Hufeisen’s factor approach is a useful point of departure.

Figure 1:

Language use by a multilingual speaker. Template for a variable model.

This model will appear in different realizations depending on the language user’s stage of mono-, bi- or multilingual development. Figure 2 applies the template to the case of an L3 user. The language knowledge factors will here have to be specified according to the user’s current state of background knowledge of one or more L1s, one or more L2s, and the L3 interlanguage. The factors contained in all the five boxes to the right also apply in their current state of development.

Figure 2:

L3 use (non-native language use by a multilingual speaker).

In representing events of language use at stages where the repertoire of languages is less developed, more limited versions of the model will apply. Figure 3 illustrates the use of a first L2. There may still be more than one L1 as background languages, but the language currently being used is the L2, and to the extent that this has developed beyond an initial stage, there is an L2 interlanguage. In Hufeisen’s model, the bundle of specific L2 learning factors is not represented in the case of L2 learning, which is logical considering that her model focuses on the initial stage of L2. But the existence of a developing L2 interlanguage in the present model motivates including this type of factors here.

Figure 3:

First L2 use.

The most limited case is the representation of language use by a monolingual, as shown in Figure 4. This variant corresponds to the stage of L1 acquisition in Hufeisen’s Factor Model, although not confined to the initial stage.

Figure 4:

Language use by a monolingual.

As I mentioned above, the extended scope of this model will have some consequences for how the various factors apply, as compared to the initial stage focus in Hufeisen’s model. For example, the person’s age, which is there listed as a neurophysiological factor present already in L1 acquisition, will be operative variably even for other languages throughout the lifetime and influence other factors as well. We do not learn the same parts of languages at different times in life, or in the different contexts in which we communicate and think, nor do we master languages in the same way (cf. Blommaert and Backus 2013: 27 f.).

In a wider perspective, this Extended Factor Model can be combined with others dealing with related aspects, such as models for the production and comprehension of utterances and the selection of language to use (cf. de Bot 2004; Levelt 1989, 1993), Green’s activation-inhibition model (Green 1986, 1998), Grosjean’s concept of language mode (Grosjean 2001), and also with models for acquisition through interaction. The primary intention here is to argue that the factor approach is fruitful in devising a variable model which is at the same time situation-related and adaptable to the dynamics of developing linguistic repertoires.

5 Relating micro and macro time

In order to exemplify how the usage events in situations connect to the development over time, I will cite some data drawn from a longitudinal test of word production (Hammarberg 1988). It is a test of phonological acquisition with adult L1 German learners of Swedish in Sweden at the initial and elementary stages. The example focuses on one test item and a particular phonological learning problem.

The subjects were tested and audio-recorded individually in a phonetic studio together with the test leader. They were shown a set of pictures of objects representing words with various phonological problems. The same test was run on three subsequent occasions: (1) at the initial stage of the learners’ encounter with Swedish, shortly after their arrival in Sweden, (2) one month later, and (3) two months after session 2. Initially, the test leader showed the pictures one at a time, while pronouncing the test word once. Then the set was run through again, and the learner was required to say the word represented by each picture. If the learner could render the word correctly or in an accented shape, the test leader would just proceed to the next item. Should the learner fail to recall the word or produce it in an erroneous form irrelevant to the phonological problem, the test leader would say it again once, and the learner would as a rule make one or more new attempts. The set of pictures was re-run four more times until by and large all words had been memorized. The same procedure was repeated in sessions 2 and 3, but without the introducing word production by the test leader, and with only one round in session 3. No written text was provided, nor any mention of pronunciation. The design was intended to appear as a test of lexical memorization capacity.

We will look at the attempts by three learners to pronounce one test word, viz. the word säng [sɛŋ] ‘bed’, illustrated by a picture of a bed. The phonological problem here consists in the fact that major varieties of German, including those of our three learners, require a word-initial alveolar fricative followed by a vowel to be voiced, which produces the structure #zV (word boundary, z, vowel) and rules out the structure #sV which is the target Swedish structure in this case. Although this German phonological pattern is phonetically motivated as voicing assimilation, it is relatively easy for German speakers to violate, which is evidenced by the fact that #sV occurs marginally in some loanwords. Yet there is an effective tendency for German speakers to produce a voiced [z] in this position. We can for these reasons foresee a conflict between two possible alternative learner solutions for this test item. We may expect a competitive situation between the German-based [z] and the input-based [s] as pronounced by the native Swedish test leader.

Figure 5 shows the results of the repeated pronunciation attempts by the three learners WG, GS and KG. Here 1:1 designates session 1, round 1; 1:2 stands for session 1, round 2, and so forth. The symbol § marks the places where the test leader intervenes and pronounces the test word again once.

Figure 5:

Pronunciation sequences for the Swedish word säng [sɛŋ] by three L1 German speakers.

Looking first at the speaker WG, we can see that there is an element of variation and uncertainty during session 1. In the first attempt (at 1:1), WG renders the form as given by the test leader. The [s] is kept also at 1:2, but the shift of vowel makes the test leader intervene, and WG repeats the word twice, still with [s]. Then at 1:3, there is a spontaneous shift to [z], which is maintained also at 1:4, even after the test leader has given the word again because of the wrong vowel. At 1:5 another spontaneous shift occurs, back to [s]. However, in session 2 one month later, WG seems to have settled on a stabilized use of the pattern #zV, which is then maintained in session 3. The development has reached what in complexity theory is termed an attractor state, a stable state over a shorter or longer period. In this case it is the L1-based variant #zV that prevails, and not the target form #sV. An early period of competition between variants is seen in session 1 and appears to be over (at least for the time being) by session 2.

The speaker GS appears to follow an opposite path. The form with [z] is firmly established throughout session 1, and there is no sign of hesitation or variation. But at 2:1 a partly devoiced variant appears, indicated here by a diacritical devoicing mark. After two more rounds with [z], GS spontaneously shifts to [s] at 2:4 and then maintains this form. The process of variation and change is thus seen in session 2. This takes place without any intervention by the test leader; apparently the learner’s experience with Swedish speech in the preceding month has effected this change.

The speaker KG shows greater difficulty in the beginning. He produces the word [bεd]. Various sources for this form seem possible. It may either be traceable to the English word bed, in that case a prior-L2 influence, or KG may possibly have been acquainted with the Swedish word bädd [bεd], which is a synonym. It is less frequent, but occurs in certain contexts, e. g. when speaking about hotel beds. Joint influence from the similar English and Swedish words is also conceivable, as well as supporting influence from the German counterpart Bett. In any case, KG gets corrected, but keeps struggling to recall and produce the target form, even through round 2:2. It is interesting to note that, from the moment when it is no longer a problem for him to remember the word säng and suppress bed/bädd (at 2:3), he gives up the attempt to render the [s] and settles on the German phonological pattern #zV. Apparently, the influence from the L1 pattern prevails when he releases control and no longer makes efforts to pay attention to the input form. The system has then reached an attractor state.

This test was arranged so as to enable us to observe both usage events in individual situations and the patterns of development that emerge over a period of time. The test yielded sequences of learner attempts on three different timescales: immediate succession (of repeated attempts at each round when the card was shown), between subsequent rounds (1–2 minutes intervals), and from session to session (1–2 months, with other exposure to the target language in the meantime). Taking the distinction between speech situation and longitudinal development into consideration, we may observe the learner attempts at each round (1:1, 1:2 etc.) from the point of view of single usage events (micro time perspective), and take the sequences across rounds and sessions as showing developments over (shorter and longer) periods of time (macro time perspective). The phonological phenomenon chosen is one where a competition between alternative learner solutions arises, a case in which crosslinguistic influence from L1 or a prior L2 and the readiness and ability to render the input form are of comparable strength. The outcome illustrates how this dynamic situation forms a basis for a developing pattern, and how the current state of development is reflected in each subsequent usage event. Variable production in each situation and fluctuations from time to time reflect the dynamics of ongoing change, whereas invariable production reflects periods of stabilization. The variable representation in the learner’s mind originates in situations of use when the reception of the target form from the interlocutor is confronted with the pre-established L1 pattern. But we can note that this unstable state may stay on and may lead to an apparently spontaneous change in a subsequent round (e. g. WG 1:5; KG 2:3) or in a later session (GS session 2), where there has been no immediate fresh input. This indicates that this part of the learner’s linguistic repertoire is still in a dynamic state at this point, and that this influences the process in new situations of use.

6 Summary and conclusion

A main purpose of this paper has been to demonstrate that the development of a person’s repertoire of languages over time and the processes of use and acquisition in situations represent different although connected types of complex systems which need to be studied each in its own right. We have defined two perspectives on language acquisition as the macro and the micro time level. Each of them poses certain problems for the current research on SLA and TLA to deal with. In part, these problems involve how the terms for the multilingual speaker’s languages (L1, L2, L3) are conceived.

At the macro time level, the acquisition of a language constitutes a part of the development of a linguistic repertoire and should be seen in this wider context, in connection with other languages also developing within the repertoire, and various aspects of how an individual knows languages. One concern in the present discussion is that the cognitively based distinction of L1 and L2 as commonly adopted in SLA research would also need to be applied consistently in TLA studies. Moreover, the way of describing multilingual development by means of a chronological sequence of languages, L1, L2, L3, L4, …, is not well suited to the nature of a linguistic repertoire and has effects of distorting or at best limiting the perspective.

The concept of L3 is specifically related to the level of micro time processes. One of the basic topics in TLA research has been to demonstrate that the acquisition of an L3 takes place under different conditions than a previously monolingual learner’s acquisition of an L2. For this reason, it has been emphasized that the acquisition of an L3 and a first L2 need to be distinguished in research, in contrast to traditional SLA practice. One model which has the aim of clarifying this distinction is Hufeisen’s Factor Model which deals with the sequential acquisition of an L1, a (first) L2 and an L3. In the present chapter I have suggested an extension of the factor approach to a more general model, an Extended Factor Model. This retains Hufeisen’s variable set of conditioning factors, but allows for different stages and types of language knowledge and different chronologies of development in the repertoire while focusing explicitly on the situation at the micro time level.

The connection between the micro time processes in specific situations and the development over a period in macro time is here illustrated by a longitudinal test of a particular phonological phenomenon which exposes both individual usage events and an evolving pattern. This has given glimpses of the dynamic nature of both the situations of use and the linguistic repertoire. Further research on the relation between micro and macro processes in a unified design could be fruitful in order to shed more light on the ways these two different levels function together.