ChapterPDF Available

Acquisition of additional languages as reorganization in the multilingual constructicon


Abstract and Figures

Recent years have seen an increasing interest in applying Construction Grammar to additional language (AL) acquisition as well as in constructionist approaches to language contact and multilingualism, in particular Diasystematic Construction Grammar (DCxG; Höder 2018). This paper combines both perspectives by proposing a usage-based constructionist model of AL acquisition as emerging multilingualism. In line with earlier work on DCxG, we assume that multilingual speakers store and process all of their languages in terms of constructions that are organized into one common constructicon. From that perspective, AL learning amounts to an extension and reorganization of the constructicon, resulting not only in the gradual entrenchment of new constructions that represent (a learner variety of) the AL, but also in modifications of previously acquired constructions and the links between them. The model is illustrated by examples from different kinds of AL acquisition scenarios and also discussed in relation to current key concepts within non-constructionist research in the field of AL acquisition.
Content may be subject to copyright.
Additional language acquisition as emerging multilingualism. A Construction Grammar
Steffen Höder (Kiel University), Julia Prentice (University of Gothenburg) & Sofia Tingsell
(Swedish Language Council)
1 Introduction: Why a constructional approach?
Non-native speakers typically use linguistic elements in ways that differ from native speakers’
usage. The production of non-canonical utterances is usually assumed to reflect the fact that non-
native speakers have a mental representation of the languages grammar that is different from
native speakers’ linguistic knowledge. Typically (though by no means necessarily), non-canon-
ical utterances appear to follow grammatical patterns of a speaker’s first language (L1) rather
than the canonical patterns of the additional language (AL).1 In (1), for example, a speaker
whose L1 is Macedonian produces a clause in her AL Swedish that is non-canonical in that it,
traditionally speaking, violates a syntactic rule:
(1) Swedish, L1 Macedonian (SW1203-uppsatser)2
Därför mitt förslag till kommunen är en ny belysning
therefore my proposal to municipality-DEF is a new illumination
‘So, my proposal to the municipality is a different illumination.’
In Swedish declarative main clauses, the finite verb always follows immediately after the first
constituent (verb-second word order). Therefore, the finite verb would be expected after the ini-
tial adverbial (Standard Swedish: Därför är mitt förslag till kommunen). In the actual utter-
ance, however, both the adverbial and the subject precede the finite verb (Därför mitt förslag till
kommunen är), following a conventional pattern in Macedonian, namely subject-verb word
While one could argue that the match between subject-verb word order in the speaker’s AL
Swedish and the L1 Macedonian is accidental and that the use of this word order pattern might
also be explained by, for example, a tendency towards using ‘unmarked’ rather than ‘marked’
1We prefer the terms additional language (AL) over second language (L2), third language (L3) and so on,
since it is unclear whether there is any categorical difference between L2s and L2+ns in terms of acquisitional
processes (if indeed there are categorical differences between L1s and L1+ns at all). This terminology is not
new, but has gained increasing popularity in recent years (see e.g. Ortega 2009).
2SW1203-uppsatser is a subcorpus of the Swedish Learner Language corpus (SweLL), consisting of student es-
says by advanced learners of Swedish as an AL (Volodina et al. 2016). It is available (access restricted) through
the Swedish corpus tool Korp (
structures (Eckman 2010; cf. Haspelmath 2006 for a critical discussion of markedness), other
cases are clearer, such as (2):
(2) English, L1 Norwegian (Johansson 2008: 144)
Here is probably an abortion a good solution.
‘Here, abortion is probably a good solution.’
The speaker, a learner of English with Norwegian as her L1, produces a clause that follows a
subject-finite inversion pattern that would be obligatory in Norwegian declarative main clauses
with a topicalized adverb but is restricted to auxiliary verbs in specific contexts in English (Gold-
berg 2006: 166–182).
Similarly, the speaker in (3), who has Finnish as an L1, produces a noun phrase, Finlands na-
tionaldagen ‘Finland’s national day’, that contains both a prenominal possessive (Finlands ‘Fin-
land-POSS’) and a definite noun (nationaldagen ‘national-day-DEF’), whereas prenominal possess-
ives are usually combined with indefinite nouns in Swedish (Standard Swedish: Finlands natio-
naldag). This reflects the fact that Finnish, unlike Swedish, does not inflect nouns for definite-
ness and, hence, there are no structural restrictions pertaining to definiteness as an inflectional
(3) Swedish, L1 Finnish (SW1203-uppsatser)
Därför ska jag gärna ta del svenska traditioner –
therefore shall I gladly take part Swedish traditions
och fira Finlands nationaldagen i December.
and celebrate Finland-POSS national-day-DEF in December
‘That’s why I’ll gladly participate in Swedish traditions – and celebrate Finland’s Na-
tional day [Independence Day] in December.’
Traditionally, non-canonical utterances such as in (1)–(3) would often be classified as containing
different types of transfer phenomena, loosely definable as the application of grammatical struc-
tures that originally belong to a speakers L1 to the production of an utterance in her AL, due to
her (as of yet) imperfect acquisition of the AL grammar. The learner’s incomplete knowledge, as
it were, of the AL corresponds to what is often called her interlanguage (a term originally intro-
duced by Selinker [1972]) – an idiosyncratic and changing variety of the AL, reflecting the
learner’s current hypotheses about the language system that she is acquiring (cf. Section 2.3). In
this view, the L1 and the AL form two separate grammatical systems, and the AL learner gradu-
ally accumulates linguistic knowledge in the AL, striving for (and possibly in the end achieving)
native-like language competence.
In contrast, current theoretical approaches to AL acquisition conceptualize the learner as an
emerging multilingual speaker, whose knowledge of the AL does not evolve independently of
previously acquired languages, but rather builds upon her prior linguistic knowledge, in particu-
lar in her L1. Usage-based research within the field of language acquisition has drawn attention
to the fact that exposure to input is crucial for language attainment, leading to the gradual en-
trenchment of linguistic units (Divjak & Caldwell-Harris 2019, Schmid 2017). While L1 acquisi-
tion typically involves a much higher degree of exposure to L1 material than AL acquisition,
both ultimately rely on the same cognitive mechanisms at the level of individual structural ele-
ments, regardless of the order of acquisition, and are processed in essentially the same way.3
From this perspective, then, the process of acquiring linguistic material in the AL is fundament-
ally the same as the acquisition of some new set of linguistic material in, say, a specific register
in the L1, and both entail some type of reorganization of the learner’s overall linguistic know-
ledge. On the functional side, specifically, the acquisition of additional linguistic material is usu-
ally related to the learner becoming familiar with new referential meanings, grammatical func-
tions, or socio-pragmatic aspects such as specific domains or communicative settings. This view
on language acquisition resonates well with usage-based and Construction Grammar (CxG) ap-
proaches to the organization and acquisition of linguistic knowledge in general (cf. Tomasello
2003, Diessel 2013, 2019, Ellis & Wulff 2019, Hilpert 2019: 241–243, Matthews & Krajewski
2019) and current constructionist approaches to language contact and multilingualism in particu-
lar (Boas & Höder 2018), most importantly Diasystematic Construction Grammar (DCxG; cf.
Section 2.1).
This paper does not present an empirical study, but is theoretical in nature, primarily motiv-
ated by the constructionist claim that “it’s constructions all the way down” (Goldberg 2006: 18)
– which must also be assumed to hold for AL learners’ linguistic knowledge –, and additionally
by the need to capture the details of emerging multilingualism in coherent terms. Our aim is to
propose a model for the constructional organization of learners’ emerging multilingual know-
ledge that extends DCxG so as to be applicable to AL acquisition. The paper is structured as fol-
lows: Section 2 discusses previous approaches to AL acquisition (constructionist and otherwise)
3 This view is confirmed by neuroscientific findings (e.g. Perani & Abutalebi 2005), which suggest that AL pro-
cessing involves the same type of brain activity as L1 processing.
that are relevant to the proposed model, focussing on DCxG (Section 2.1), CxG work on AL ac-
quisition (Section 2.2), and key insights from non-constructionist research (Section 2.3). Our
proposed model is spelled out in Section 3, focusing on the nature of entrenchment (Section 3.1)
in AL acquisition, the internal structure of the emerging multilingual constructicon (Section 3.2),
and the details of the reorganizational processes that are assumed to be at work when AL know-
ledge is integrated into the constructicon (Section 3.3). Finally, Section 4 gives a summary.
2 Building on previous research
2.1 A constructionist approach to multilingualism: Diasystematic Construction Grammar
Diasystematic Construction Grammar (DCxG; Höder 2012, 2014ab, 2018, 2019) is a usage-
based Construction Grammar approach to language contact situations, including a wide range of
types and degrees of individual as well as collective multilingualism. While the development of
DCxG focused mainly on phenomena that are related to contact-induced language change (e.g.
Höder 2012, 2014a), it has more recently also been applied to other types of contact phenomena,
such as synchronic multilingual practices (cf. the papers by Bourgeois, Namboodiripad, Lepic,
and Urban, this volume). Drawing on insights from current contact linguistics, DCxG differs
from more traditional linguistic approaches (including constructionist ones) in assuming that
‘languages’ do not have any a priori status in the organization of grammar, neither in individual
speakers’ linguistic knowledge nor in the social conventions that represent the grammar shared
by a specific speaker group. Rather, speakers organize their linguistic knowledge in its entirety
into one constructicon, which, consequently, may contain constructions that are used in different
languages. From a contact linguistic perspective, this reflects the well-established concept of the
linguistic repertoire (Matras 2009: 208–209), from which multilingual speakers choose the ele-
ments that they deem to be appropriate in a given communicative situation. From a CxG per-
spective, the existence of multilingual constructicons is motivated by the general claim that the
acquisition and organization of linguistic knowledge is exclusively governed by domain-general
cognitive factors and mechanisms, such as input frequency, saliency, and abstraction and gener-
alization based on perceived similarity. If these do not lead to a differentiation between distinct
‘languages’ in the constructicon, then there is no reason to assume that such a differentiation is
cognitively real.
On the other hand, multilinguals do not normally use constructions from different ‘languages’
randomly, but rather according to specific patterns that reflect conventionalized associations
between (sets of) linguistic elements and (sets of) communicative contexts, defined by, for in-
stance, discourse topics, interlocutor constellations, or – in a broader sense – communicative do-
mains. These patterns, which have been investigated in countless studies on what is often la-
belled ‘code choice’ and ‘code-switching’ (cf. Gardner-Chloros 2009: 42–59), reflect a more
general principle in multilingual communication known as the Complementarity Principle (Gros-
jean 2008: 22–34): multilinguals tend to use their different languages for different purposes. In
DCxG, this is reflected in terms of pragmatic meaning: a linguistic element – i.e. a construction –
that ‘belongs to language A carries a specific type of pragmatic meaning that restricts its use to a
specific set of communicative contexts that are conventionally associated with ‘language A’, and
at the same time it marks the current communication as belonging to this set of contexts.4 If, for
example, a German-Swedish bilingual family living in Sweden has established a convention to
use German at home and Swedish at work or at school, the language-specificity of the German
word Küche ‘kitchen’ is represented as a pragmatic restriction on the functional side of the cor-
responding construction. This can be formalized as [Küche ‘kitchen’ Chome ], with a shorthand
notation specifying the communicative context (Chome) in angle brackets, as opposed to the
Swedish equivalent [kök ‘kitchen’ Cwork; school ]. Constructions carrying pragmatic meaning of this
type ( Cx) are called idioconstructions in DCxG.
Language-specificity is, however, an optional property of constructions, since there are con-
structions that are not restricted to specific communicative contexts in this way. More schematic
constructions, in particular, can often be used across all communicative contexts. For example,
polar questions as illustrated in (4) share the same syntactic structure in both German and
Swedish in that they contain an initial finite verb, followed by a subject (typically a noun phrase
4This is in line with the view expressed by e.g. Cappelle (2017), who argues that both (referential) semantics
and pragmatics should be included on the functional side of constructions and that pragmatic information
should not be treated as extra-constructional. Similarly, Goldberg (2019: 7) views constructions as “emer-
gent clusters of lossy memory traces that are aligned within our […] conceptional space on the basis of
shared form, function, and contextual dimensions”.
in the nominative case, if applicable), and optionally other elements (whose presence and order
are governed by additional constructions). They also encode the same type of functional informa-
tion (illocutionary force):
(4) a. German
Kommst du bald nach Hause?
come you soon home
b. Swedish
Kommer du hem snart?
come you home soon
‘Are you coming home soon?’
This corresponds to a schematic construction that can be formalized as [Vfin¹ SBJ² polar ques-
tion ]. While the construction does not specify any additional pragmatic meaning for the schema
itself, pragmatically unspecified constructions of this kind (diaconstructions) of course are often
instantiated along with idioconstructions. In (4), for example, language-specific lexical construc-
tions fill the finite verb and subject slots in the polar question diaconstruction.
Particularly schematic diaconstructions can, in principle, be posited at fairly high levels of ab-
straction and schematicity (cf. Höder 2018: 62–64, 2019). As DCxG aims to model multilin-
guals’ linguistic knowledge in a socio-cognitively realistic fashion, the question whether highly
schematic constructions actually exist in speakers’ cognition is highly relevant. While there is a
broad consensus in the constructionist (and cognitivist) literature that the organization of lin-
guistic knowledge depends more on lower-level schemas than constructions with a higher degree
of schematicity (e.g. Hilpert 2019: 67–72), it also follows from common assumptions about rich
memory representations in usage-based linguistics (Bybee 2010: 14–32) that representations at
different levels of schematicity need not be mutually exclusive.5
As a consequence, DCxG assumes that schematic diaconstructions and less schematic
idioconstructions coexist in the multilingual constructicon, and are connected via inheritance
links (Figure 1; cf. Höder 2019: 341–342). Formal and functional properties can, but need not be
specified both at the diaconstructional and at the idioconstructional level; the only categorical
5 Furthermore, the assumption of diaconstructions is supported by different types of empirical evidence, such as
findings pointing to interlingual productivity (Höder 2018: 59–60, 2019: 347–348), i.e. the productive use of a
construction in one of the languages involved when it results in non-canonical, but perfectly comprehensible ut -
terances in one of the languages involved (cf. Hilpert 2019: 151–153).
difference between the two levels is the type of pragmatic information that idioconstructions
formX(+ formA)
meaningX(+ meaningA) CA
formX(+ formB)
meaningX(+ meaningB) CB
Figure 1: Diaconstructions and idioconstructions within the multilingual constructicon
Furthermore, the constructions that are used in a particular ‘language’ are not necessarily specific
to it. Rather, the linguistic knowledge that is needed in order to represent and process ‘language
A is composed of both diaconstructions and idioconstructions.
Previous work on DCxG has focused on multilingualism within what has been characterized
as ‘stable-ish’ multilingual communities, i.e. “groups that are stable enough in terms of their spa-
tial, temporal, and social structure as to allow for linguistic conventions to emerge and stabilise”
(Höder 2018: 42, fn. 9), such as communities that are traditionally diglossic, with relatively
firmly established conventions on the association between domains and ‘language’ choice. Con-
sequently, “the object of inquiry is defined as the set of cxns shared by a specific (multilingual)
speaker community” (Höder 2019: 340), i.e. the multilingual constructicon is viewed as having
both a cognitive side at the individual level and a conventional side at the collective level (com-
munity-specific grammar).
When it comes to AL acquisition by individual speakers that are not part of a stable(-ish) mul-
tilingual community, it is evident that community-specific linguistic conventions and individual
linguistic knowledge do not mirror each other in the same way as in multilingual communities
with shared social conventions about when to use what language. It is, indeed, often the case that
learners are isolated in the sense that no one else in their social network uses the same set of lan-
guages in the same type of distribution across communicative settings as they do. Therefore, ap-
plying DCxG to AL acquisition as emerging individual multilingualism requires a slightly differ-
ent view of the ontological status of the constructicon and, indeed, specific constructions, namely
one that focuses primarily on the cognitive organization of linguistic knowledge at the individual
level. So, instead of defining constructions as “conventional, learned form-function pairings”
(Goldberg 2013: 17), it is sufficient to just define them as cognitive units, i.e. learned form-func-
tion pairings in the learner’s emerging multilingual constructicon, whether or not they reflect
conventional units within the AL speaker group or some community that the learner belongs to.6
2.2 Additional language acquisition and Construction Grammar
As Ellis (2013: 365) points out, “[i]f the units of language are constructions, then language ac-
quisition is the learning of constructions. So L2A [L2 acquisition] depends upon learners’ experi-
ence of language usage and upon what they can make of it.” Consequently, there has been an in-
creasing interest in usage-based and in particular constructionist approaches within the field of
AL acquisition over the past decades; constructionists have also, conversely, been increasingly
interested in AL acquisition (for an overview cf. Ellis 2013 and De Knop & Gilquin 2016).
Among the abundant studies in the field, many findings suggest major differences in language
usage and, more crucially, the cognitive organization of linguistic knowledge between L1 and AL
speakers. Such findings include the observation that AL speakers appear to be relying more on
schematic patterns in language processing than L1 speakers (e.g. Gries & Wulff 2005). Structural
interaction between L1 and AL constructions has also been the focus of numerous studies, point-
ing to the possibility of analysing them in terms of constructional transfer mechanisms, as op-
posed to ‘mere’ lexical or grammatical transfer (e.g. Römer, O’Donnell & Ellis 2014).
Such findings fit in with Ellis’s (e.g. 2006: 184) emergentist view of the initial disposition of
the mind in AL acquisition as a tabula repleta (‘replenished table’, in contrast to the tabula rasa
in L1 acquisition): Since the learner already possesses linguistic knowledge, consisting of L1
constructions, AL acquisition differs from L1 acquisition in that the learner does not build up her
mental constructicon from scratch, eventually leading to the establishment of constructions emer-
ging from usage, but rather integrates newly acquired knowledge into what is already there.
On the other hand, constructionist studies also find similarities between L1 and AL acquisi-
tion, particularly with regard to the role of linguistic input (cf. Ellis & Wulff 2014). For example,
6 Findings from related fields also suggest that the cognitive representation of grammar can be fairly independent
of linguistic conventions, including in (nearly) monolingual communities, even though it is of course shaped by
social practices. For example, Dąbrowska (2019: 231), in her survey article, refers to the idea that (first) lan-
guage acquisition leads to identical mental grammars for all speakers as a ‘myth’, and Sabino (2018: 75–99)
even speaks of the ‘illusion of shared grammar’.
several recent studies investigate high school students with various degrees of access to AL in-
put. Hendrikx (2019) and Van Goethem & Hendrikx (this volume) study the effects of a Content
and Language Integrated Learning model (CLIL) on the acquisition of specific constructions in
English and Dutch as ALs. Loenheim (2019) investigates Swedish compounds as interpreted by
L1 and AL speakers with varying ages of onset, reflecting different amounts of exposure to the
AL. Both studies show that learners’ ability to use their linguistic knowledge (productively and
receptively) similarly to L1 speakers correlates with the amount of exposure to the AL. This in-
cludes, in Loenheim’s (2019) study, the accessibility of more abstract schemas as opposed to lex-
ically filled constructions. Such findings point to the importance of individual constructions in
AL acquisition, as opposed to the role of languages as a priori categories.
2.3 From interlanguage to the multilingual turn: insights from non-constructionist
It has, of course, long been acknowledged in AL research that L1 structures have an impact on
AL acquisition. One of the first and most well-known theoretical models that take L1 influence
into account was the notion of interlanguage (IL), introduced in Selinker’s (1972) oft-quoted art-
icle. Interlanguage is understood as a learners linguistic system at any given point in time during
the acquisition of a new language (the target language [TL] in Selinker’s terminology) in addi-
tion to the L1 (in Selinker’s terms, the native language [NL]) or, more broadly, any previously
acquired language(s). Originally, the concept was developed within the framework of the Inter-
language Hypothesis, which involves a set of theoretical postulates, one of them being that a
learner’s IL can be characterized as a “separate linguistic system based on observable output”
(Selinker 1972: 214). IL is thus viewed as a system in its own right, building on a basic system
of hard-wired language universals (what Selinker [1972: 212] calls a “latent language
structure”), rather than an erratic product of the learner’s deficient knowledge in the TL.7 AL pro-
duction involves five fundamental processes in language learning, viz. (a) transfer from the NL,
(b) transfer of training, (c) strategies of language learning, (d) strategies of second language
communication, and (e) overgeneralization of TL material. The theory also includes the concept
7 The “latent language structure” is related to the Chomskyan notion of universal grammar and understood as a
pre-existing arrangement in the brain.
of fossilization, which denotes non-TL structures in advanced or possibly final states of language
acquisition. The five processes are different in kind: While transfer of training, strategies of lan-
guage learning, and strategies of second language communication (processes (b)−(d)) are based
on general learning or coping strategies employed in language learning, transfer from the NL
(process (a)) and overgeneralization of TL material (process (e)) explicitly involve the cognitive
processing of pre-established linguistic knowledge and TL input, respectively.
In mainstream, present-day AL research, the term interlanguage is still used to refer to the
learner’s variety of her AL, but in many cases the concept is used in a way that is theoretically
underspecified. Regardless of whether the term is understood in its original context or as more
widely applicable tool, more recent dynamic views on language learning and language practices
(cf. Cook & Li Wei 2016, García 2009, Ortega 2009, Larsen-Freeman 2011) challenge some of
the underlying theoretical assumptions of the concept, since it presupposes that AL acquisition is
merely an accumulative process, in which new linguistic material is added to existing know-
ledge. As a consequence, the term also implies a one-dimensional process leading from a starting
point (at which the learner has no knowledge of the AL) towards a hypothetical endpoint, at
which the learner has acquired the TL system in its entirety, resulting in her ability to use the TL
in a native-like way.
The notion of interlanguage, even in its original sense, deserves credit for emphasizing the
systematicity of learner language, including systematic influence from the L1. However, it is ob-
vious that, firstly, Selinker’s fundamental claim that the interlanguage builds on universal struc -
tures does not resonate well with a usage-based view. Secondly, while the view that the IL con-
stitutes a separate system in the sense that it differs from native varieties of the TL is obviously
true, viewing the IL as a separate system in cognitive terms is irreconcilable with the DCxG as-
sumption that the mental constructicon of multilingual speakers, including emerging multilin-
guals, is not partitioned into different languages. However, Selinker’s (1969, 1972) discussion at
least in part already suggests that the accumulative and, at any given point in time, static nature
of interlanguage may not have been entirely intended by the original design of the concept: “…
we can speculate that as part of a definition of ‘learning a second language’, ‘successful learning’
of a second language for most learners, involves, to a large extent, the reorganization of lin-
guistic material from an IL to identity with a particular TL”, Selinker (1972: 224) states, but ex-
plicitly refers to future research for the implications of such a belief. This more nuanced view
could be seen as (compatible with) an embryonically emergentist position, which allows us to
avoid thinking of the IL as reflecting a transitional process between an L1 and the (imagined) fi-
nal state of the TL in the learner and, instead, to conceive of the IL as representing a subset of
constructions within an evolving multilingual constructicon: In a (potentially) ever-growing net-
work, there is no need to identify a starting point and an endpoint – there is only an expansion, in
many directions, involving both L1 and AL material. Thus, it is possible to reinterpret the notion
of IL in terms of a dynamic, changing, and growing constructional network representing the
learner’s AL within her mental constructicon.
The same holds true for other key concepts in AL research, such as the notion of transfer,
which has also changed from being seen as a rather static and unidirectional mechanism to more
dynamic definitions. In current research on cross-linguistic influence (CLI), transfer is viewed as
a “highly complex cognitive phenomenon that is often affected by language users’ perceptions,
conceptualizations, mental associations, and individual choices” (Jarvis & Pavlenko 2008: 13).
While transfer in earlier days was often seen as a mechanism of interference of L1 in AL produc-
tion and, hence, a major obstacle for AL acquisition (e.g. Selinker 1972, Lado 1957), today’s
view includes all of the languages a language user has knowledge of and their influence on each
other (Jarvis & Pavlenko 2008: 13), including influence of the AL on the L1. Findings support-
ing this view were already published decades ago (e.g. Ringbom 1978, Weinreich 1953) but
have, like other findings indicating the complexity of transfer effects, only reached full recogni-
tion in more recent years (Jarvis & Pavlenko 2008: 12).8
On a larger scale, the field of AL acquisition has seen two major paradigm shifts in recent
decades, often referred to as the multilingual turn (or bilingual turn) and the social turn (cf. May
2013, Ortega 2009, 2019). The shared motivation between the two turns is the shift of focus from
cognitive and linguistic structure alone to a sociolinguistically more realistic perspective, con-
templating the AL learner as an active individual within a social context. While the social turn is
also highly relevant for AL acquisition in general, it is obvious that the multilingual turn is more
8 This also entails the insight that CLI not only affects formal but also semantic or functional aspects (see Jarvis &
Pavlenko 2008: 11–13).
crucial with regard to a usage-based approach to AL acquisition as emerging multilingualism.9
The multilingual turn opposed the monolingual bias, the notion that it is “normal” for an indi-
vidual or a society to speak one language only. This notion affected not only the general view on
language, language learning, and language planning, but also had an impact on linguistic theory,
including AL theory. Prior to the multilingual turn, the field of AL acquisition, perhaps surpris-
ingly, did not challenge the monolingual norm as such, but instead, in most cases, built upon
monolingual theories of language and language acquisition and introduced add-ons in order to
describe the acquisition of additional languages and the mind of the multilingual speaker. Thus,
the multilingual turn has, as May (2013: 1) puts it, “usefully foregrounded multilingualism,
rather than monolingualism, as the new norm of applied linguistics”. The foregrounding of mul-
tilingual competence(s) is not entirely revolutionary, though, as the term paradigm shift may im-
For example, the concept of multicompetence (e.g. Cook 1991, 1993, Cook & Li Wei 2016),
originally developed within Universal Grammar theory, refers to a “compound state of a mind
with two grammars” (Cook 1991: 112), which is described as different from a mere coexistence
of two monolingual systems, echoing Grosjean’s (1989: 4) oft-quoted claim that “the bilingual is
not two monolinguals in one person”. In Cook’s (1993: 3) metaphor, “an L2 user does not just
have a second L2 competence tacked on the original L1 competence, an extension built on at the
back of the house; rather, the L2 user’s mind is different as a whole – the whole house has been
This rewiring of the entire linguistic setup is also at the core of Li Wei’s and García’s extens-
ive work on translanguaging (cf. Li Wei 2011, 2018, García & Li Wei 2014). While the process
of translanguaging is given different definitions for various purposes, Li Wei (2011: 1223)
defines translanguaging as including “the full range of linguistic performances of multilingual
language users for purposes that transcend the combination of structures, the alternation of sys-
9 The social turn has raised awareness of how the multilingual individual is often socially disadvantaged. Elimin -
ating the traditional dichotomy between native and non-native speakers, Pavlenko (2003) suggests, is beneficial
for language learners who are unable or unwilling to identify themselves as either learners or native speakers. In
order to experience ownership of and to invest in (cf. Peirce 1995) a (new) language, learners need an ‘imagined
community’ in which mastering the new language is meaningful. When identification with neither a community
of native speakers nor a community of learners is possible or desirable for the learner, there is a void of imagined
communities. Offering an imagined bilingual community fills that void and makes it possible for the learner to
invest in the new language in order to become a full member of that community.
tems, the transmission of information and the representation of values, identities and relation-
Similarly, complexity theory (cf. Larsen-Freeman 2011) offers a framework that views lan-
guage as emerging from usage; all patterns, or constructions, are products of language use, and,
hence, “[t]he system changes every time a form is used” (Larsen-Freeman 2011: 53). The system
itself is thus ever-changing: “in short, language learning is not just about adding knowledge to an
unchanging system. It is about changing the system” (Larsen-Freeman 2011: 57). The categories
within the system are not discreet, but rather gradual, and Larsen-Freeman (2011: 53) refers to
the granularity of constructions to illustrate this point.
To sum up, a constructionist approach to AL acquisition that views AL learners as emerging
multilinguals is not only consistent with current thinking about multilingualism in Construction
Grammar, but also compatible with recent developments in the field of (non-constructionist) re-
search on AL acquisition.
3 Modelling AL acquisition in Diasystematic Construction Grammar: a proposal
3.1 Gradual entrenchment of constructions
From a cognitive perspective, the acquisition of linguistic material amounts to the establishment
of units of linguistic knowledge on the basis of linguistic input. In usage-based linguistics, en-
trenchment is seen as a key process in language acquisition, whether in L1 acquisition (cf.
Theakston 2017) or AL acquisition (MacWhinney 2017; for a general discussion from a con-
structionist perspective, cf. Hilpert & Diessel 2017). Schmid (2017: 24) defines entrenchment as
“the ongoing reorganization and adaptation of individual communicative knowledge, which is
subject to exposure to language and language use and to the exigencies of domain-general cog-
nitive processes and the social environment”. In his Entrenchment-and-Conventionalization
Model, Schmid views entrenchment as a dynamic, gradual process (2015, 2016, 2017), eventu-
ally leading to a degree of entrenchment that mirrors a construction’s degree of conventionaliza-
tion within the relevant speaker community.10 In this view, constructions are not entrenched as
10 Schmid (2017: 25) describes linguistic knowledge as “being available in one format only, namely, associations”.
While Schmid does not explicitly place the Entrenchment-and-Conventionalization Model within the framework
ready-made form-function pairings that are represented as static units in the individual’s mental
constructicon, but rather form “more or less strongly entrenched symbolic associations between
forms and meanings” (Schmid 2017: 25). Constructionhood, in this view, is not binary; rather, it
reflects a high degree of entrenchment and conventionalization (cf. Hilpert & Diessel 2017: 67–
69; Divjak & Caldwell-Harris 2019), based on generalizations over recurrent tokens in the in-
put.11 As dynamic entities, constructions can both be strengthened, whenever they are accessed in
language processing, and weakened as a consequence of disuse (disentrenchment; cf. Steinkrauss
& Schmid 2017: 370). In an AL context, this means that every token of usage strengthens the
mental representation of a certain form, its referential or grammatical meaning, and its pragmatic
association with the specific context in which the learner uses or encounters the AL, thus en-
abling the formation of a new construction. However, entrenchment of AL material does not ne-
cessarily correspond to any degree of conventionalization, since even completely unconven-
tional, idiosyncratic constructions can be fully entrenched in the learner’s repertoire.
Entrenchment, as well as disentrenchment, according to Schmid (2017), does not only involve
the symbolic, intraconstructional links between forms and functions, but also interconstructional
links, i.e. links between different constructions, and pragmatic links between construction and
extra-linguistic contexts and experiences, i.e. associations between constructions and communic-
ative contexts. The latter are playing an important role in the context of AL acquisition, espe-
cially in a model that draws on the theoretical approach of DCxG, which, as mentioned previ-
ously, views constructions not as a priori language-specific, but rather as context-specific (Höder
2018: 43–44). The difference between DCxG and Schmid’s description of the role of context is
that Schmid does not include pragmatic associations within the scope of the symbolic associ-
ations between forms and functions, whereas DCxG sees pragmatics as part of constructions’ se-
mantic properties, i.e. as an integrated part of their function. This difference is not a theoretical
contradiction per se but rather due to the fact that DCxG, in line with general tenets of CxG, con-
of Construction Grammar, the two are generally compatible; symbolic links between constructional form and
function as well as interconstructional links within the constructicon can, in principle, be conceptualized as (dif -
ferent types of) associations.
11 This is related to the notion of exemplars, i.e. “categories formed from tokens of experience that are judged to be
the same” (Bybee 2013: 53). With every token of usage (receptive or productive), the mental representation of an
exemplar is strengthened: details of an exemplar’s form and usage are registered and stored in memory and can,
thus, become part of the mental representation of a construction.
ceptualizes the language system as consisting of constructions, the links between them, and noth-
ing else. In other words, a more strictly constructionist approach requires a somewhat stronger
formalization of the units of language than Schmid’s more general socio-cognitive approach to
linguistic knowledge in terms of a range of different kinds of associations. The latter is well
suited to explain the complexity of the interplay between linguistic and extralinguistic informa-
tion language users and learners have to be able to process and to acquire. It is, however, harder
to capture in a construction-based model of AL acquisition and development, one problem being
that the line between a semantic and a pragmatic function of a construction is as difficult to draw
as the one between its lexical and grammatical form. The socio-cognitive reality can, however,
be assumed to be the same, whether the pragmatic context is seen as part of the construction or
not: the AL learner, on the basis of experience from input, entrenches information about the con-
text in which a certain form is used and about its “social values and meanings” (Schmid 2016:
For the purposes of our model we propose that knowledge about the pragmatic context in
which a certain form is used is entrenched as part of constructional meaning, in the same way as
other aspects of the same form-meaning pair. There is, in this view, no fundamental cognitive
difference between a German learner of Swedish entrenching the link between the target form
tjena with the semantic meaning ‘greeting’ and entrenching the link between the form tjena and
the pragmatic context of, say, ‘informal conversation with Swedish-speaking acquaintances’ (as
opposed to more unmarked or even formal ways of greeting someone, such as hej or god dag).
Hence, the association between the form tjena, its referential meaning and its pragmatic function
could be described as a construction [tjena ‘greeting’ Cinformal, Swedish acquaintances ].
A learner, hence, needs to entrench the association between form and function described
above, and will do so gradually on the basis of the input, i.e. after a range of experiences of how
the form tjena is used conventionally. In an acquisitional context, one could imagine a situation
where the learner, on the basis of too little and skewed input, draws the conclusion that tjena is
the unmarked way of saying hello in Swedish and, hence, applicable in all contexts when greet-
ing another person. The learner would then, at least to some degree, entrench an unconventional
construction that could be described as [tjena ‘greeting’ CSwedish interlocutors ]. This construction is
likely to undergo subsequent pragmatic narrowing, provided the learner also experiences negat-
ive evidence in an increasing and more evenly distributed input, i.e. observes a lack of instances
of tjena in formal situations. For a German learner of Swedish, acquiring the construction [tjena
‘greeting’ Cinformal, Swedish acquaintances ] could, then, involve the following steps:
a. the learner is exposed to the form tjena in the AL input;
b. the learner encounters recurrent instances of the form tjena as a greeting formula in the
context CSwedish interlocutors;
c. too little and too skewed input prevents the acquisition of the conventionalized, i.e. more
specific, context Cinformal, Swedish acquaintances;
d. a construction tjena is weakly entrenched;
e. a construction [tjena ‘greeting’ CSwedish interlocutors ] is entrenched gradually; this entails reor-
ganizational processes within the speakers’ constructicon (see Section 3.2);
f. increased and more evenly distributed input AL-input, containing
experiences that allow the learner to differentiate between contexts like ‘formal con-
versations with Swedish-speaking strangers’, ‘informal conversations with Swedish-
speaking friends’ and ‘unmarked conversations with Swedish-speaking interlocutors’,
no experiences of the form tjena used in formal or even unmarked conversations with
Swedish-speaking interlocutors, but several encounters with the form tjena in in-
formal contexts with Swedish-speaking acquaintances (along with several encounters
with other forms used in those contexts, like god dag or hej);
g. [tjena ‘greeting’ CSwedish interlocutors ] is gradually narrowed to [tjena ‘greeting’ Cinformal, Swedish
acquaintances ].
In addition to the entrenchment and disentrenchment of constructions, AL learning also involves
the entrenchment of additional links between already entrenched constructions in the multilin-
gual constructicon and the disentrenchment of links that have been entrenched on the basis of too
little or too skewed input, as a cognitive reaction to increased AL experience; these are discussed
in detail in Section 3.2.
Summing up so far:
a. A distinction is made between the cognitive dimension (entrenchment) and the social di-
mension (conventionalization), while still recognizing that they interact.
b. Entrenchment is seen as a gradual and reversible process, and constructionhood is not
binary, i.e. a construction is more or less of a construction.
c. Constructions can be (more or less) entrenched in the learners interlanguage (i.e. her
emerging multilingual constructicon) without being conventionalized patterns in the AL
in question.
Entrenchment (along with its opposite disentrenchment) is thus considered the central cognitive
process for AL acquisition and development and the driving force behind the development of the
learner’s emerging multilingual constructicon.
3.2 The emerging multilingual constructicon
In the following we will illustrate and explain the interplay of the different components of the
model, i.e. how AL input affects both the (more or less) pre-entrenched linguistic constructions
and the (more or less) newly acquired ones within the learner’s constructicon.
multilingual constructicon
xare applied to
lead to reorganization
Figure 2: The emerging multilingual constructicon
As illustrated in Figure 2, the AL input can be described as resulting from the learners’ expos-
ure to linguistic material in the AL, which in part can be processed by applying pre-existing con-
structions, whereas other AL tokens lead to the establishment and reorganization of the learner’s
linguistic knowledge. This can be exemplified by an imagined speaker S who, up to this point,
has been monolingual in language A (LA in Figure 2) and is now acquiring language B (LB) as an
AL. Exposure to AL material implies a need to process tokens in language B. A subset of these
tokens can be processed as instances of pre-existing constructions (cxn), i.e. constructions in lan-
guage A that S has acquired at an earlier stage. Note that while S has acquired these construc-
tions in communicative contexts where language A was used, none of them can be used to mark
these contexts as different from anything else – the fact that, from a linguist’s perspective, these
constructions ‘belong to language A has no counterpart in S’s mental constructicon. In technical
DCxG terms, this means that they do not carry pragmatic meaning of the type CA. Therefore,
no reorganization of linguistic knowledge is required: The pre-existing constructions can simply
be applied to AL material as well, i.e. they function as diaconstructions in the emerging multilin-
gual constructicon. For example, a German-speaking monolingual who acquires Swedish does
not need an additional Polar Question Construction to process Swedish polar questions; the con-
struction that she already uses in German is sufficient for both languages (see example (4) in
Section 2.1).
Other cases, however, require a reorganization of S’s pre-existing linguistic knowledge. Some
structural patterns that she encounters in language B will not match any of the pre-existing con-
structions in her constructicon. As they occur exclusively in communicative contexts that are
(coming to be) associated with B, they are entrenched as constructions carrying, among other
things, pragmatic meaning of the type CB. In technical terms, they are idioconstructions
(cxn CB). Simple examples are AL words that do not have cognates in the learners L1.
Finally, S will recognize that while some pre-existing constructions can be applied to AL ma-
terial, others cannot. In order to prevent using instances of such constructions in inappropriate
contexts, S will have to acquire a pragmatic restriction that specifies the set of communicative
contexts in which they can be used, i.e. contexts that are associated with language A. This corres-
ponds to the acquisition of a pragmatic meaning of the type CA for a subset of the pre-existing
constructions (cxn CA).
Eventually, the emerging multilingual constructicon comprises
a. pre-existing, strongly entrenched constructions that function as diaconstructions;
b. newly acquired, weakly entrenched constructions restricted to CB;
c. pre-existing, but modified constructions that now are restricted to CA.
Crucially, though, none of these subsets correspond to S’s knowledge of language B in its en-
tirety. Rather, ‘language B’ is represented partly by diaconstructions, partly by CB idioconstruc-
tions. Note that not all constructions that represent language B have actually been acquired dur-
ing AL acquisition. On the contrary – and paradoxically, from a linguocentric point of view –, the
acquisition of AL material involves as much modification of pre-existing constructions as the en-
trenchment of new ones.
3.3 Reorganizational processes
From a DCxG perspective, as discussed above, AL acquisition may also require a reorganization
of pre-existing constructions. Such reorganizational processes can be modelled in terms of the
addition and removal of information at three different levels, reflecting the gradual entrenchment
or disentrenchment of linguistic knowledge:
a. Constructions: Whole constructions are added to or removed from the multilingual con-
structicon. AL idioconstructions are added whenever the learner’s input contains AL
structures that cannot be processed via pre-existing constructions. Moreover, abstraction
from and generalization over formally or functionally similar idioconstructions result in
the addition of diaconstructions that encapsulate the forms and functions that are shared
across different communicative contexts. Conversely, progress in the acquisition process
may render an already established construction superfluous or useless, which is then re-
moved from the constructicon (in a process of gradual disentrenchment).
b. Interconstructional links: Links between different constructions are added to or removed
from the constructicon. For example, a construction can become linked to a more schem-
atic construction via an inheritance link if the learner’s input indicates that the two con-
structions are formally and/or functionally similar. The removal of an inheritance link
may reflect the learner’s insight that two constructions are not as similar as previously
c. Constructional properties: Formal and/or functional properties are added to or removed
from pre-existing constructions. If, for instance, a pre-existing construction undergoes a
type of pragmatic specialization in that it becomes pragmatically restricted to a particular
set of communicative contexts that are conventionally associated with ‘language A’, then
this amounts to the addition of pragmatic information (Ø → CA). The opposite process,
i.e. the kind of pragmatic bleaching that leads to a construction being used in a wider
range of contexts, implies the loss of information on pragmatic context ( CA Ø). In
addition to or independently of the addition or removal of functional information, even
formal properties can be added or removed. These processes reflect the entrenchment or
disentrenchment of intra-constructional links (i.e. symbolic links between form and func-
tion, including pragmatic links sensu Schmid [2017]).
AL acquisition typically involves reorganizational processes at all three levels, often combined
into more complex reorganizational sequences. A relatively straightforward case is the addition
of an AL idioconstruction. Consider example (5):
(5) L1 Swedish
De tycker att det är roligt med fotboll.
they think that it is fun with football
‘They think that football is fun.’
The formal side of the Swedish construction exemplified in (5) can be represented as [SBJ[det],
VARA, ADJPpred[, (…)] PP[med NP]], i.e. an expletive third-person neuter subject pronoun det
with an inflectional form of the copula vara and a predicative adjective phrase containing an ad-
jective in the indefinite singular neuter form, followed by a prepositional phrase containing the
preposition med ‘with’ and a noun phrase. The meaning of this construction can be described as
an evaluation of the NP’s referent as having the property referred to by the adjective (cf. the
entry for det_är_AP_med_NP in the Swedish Constructicon, SweCcn).12
An isomorphous construction does not exist in German (*es ist lustig mit Fußball). Hence, an
L1 speaker of German acquiring Swedish as an AL will find that there is no equivalent construc-
tion in her constructicon that could be used to interpret constructs such as in (5) and that there is
not even a formally or semantically similar construction that could function as a starting point for
generalizing across the two languages and form a diaconstruction based on structural similarities.
The only option, then, for the learner to store and process this new piece of linguistic knowledge
will be to establish a new idioconstruction [SBJ[det], VARA, ADJPpred[, (…)] PP[med NP]
CSwedish ].
12 SweCcn ( treats this construction on a slightly more schematic level, as even
other copula verbs (such as bli ‘become’, låta ‘sound’, verka ‘seem’) and similar verbal expressions (such as
passa bra ‘fit well’) are included in the constructional description in addition to the copula vara. While it is clear
that the variant with vara forms part of a family of constructions, we will restrict the discussion to its most prom-
inent member.
Similarly, (6) and (7) are examples of constructions that require reorganization in the con-
structicon of an L1 speaker of English acquiring Swedish as an AL:
(6) a. Swedish
Han vet mycket om mig.
he knows much about me
b. Swedish
Jag känner hans bror.
I know his brother
c. English
He knows much about me.
A learner with limited experience in the AL will possibly identify tokens of the Swedish verb
veta ‘know’ as in (6a) with the English verb know, i.e. the construction [SBJ KNOW OBJ ‘know’], as
instantiated in (6c), leading to the assumption that the slots in the pre-existing construction can
also be filled with Swedish material. However, Swedish veta is more restricted than know, since
it is not conventionally used with animate objects in the same sense. Instead, this is expressed by
the verb känna (as in (6b)), whereas veta with animate objects means ‘know of sb., know that sb.
exists, know who sb. is’. Consequently, separate idioconstructions are needed for processing the
different form-function pairings in both languages (e.g. [SBJ KNOW OBJ ‘know’ CEnglish ], [SBJ VETA
OBJinanimate ‘know sth.’ CSwedish] and [SBJ känna OBJanimate ‘know sb.’ CSwedish ], although the similar-
ities may also lead to the entrenchment of a more schematic diasystematic construction such as
[SBJ KNOW-VERB OBJ ‘know (in a wider sense)’].
The reorganizational processes involved in other scenarios are far more complex, though. One
case in point is the acquisition of Swedish subordinate clause patterns by learners with German
as an L1. While word-order patterns in both languages can be considered to be fairly complex,
not least with respect to subordinate clauses, a common feature is that both Swedish and German
differentiate main and subordinate clauses systematically by means of word-order patterns. The
different patterns are usually described based on the relative position of the finite verb and other
constituents: While the default word order in main clauses (including declarative main clauses
and wh-questions) is verb-second (V2) in both languages, the finite verb tends to occupy a later
position in subordinate clauses. In German, subordinate clauses are canonically verb-final, al-
though extraposition can result in constituents being placed to the right of the finite verb (in the
so-called Nachfeld ‘post-field’). Example (7) shows main clause verb-second order in contrast
with verb-final word order (VF) in a subordinate clause as well as a variant with an extraposed
(7) L1 German
a. Ich möchte jetzt ein Bier trinken.
I want now a beer drink
‘I’d like to drink a beer now.’ (declarative main clause; V2)
b. …, dass ich jetzt ein Bier trinken möchte.
that I now a beer drink want
‘… that I’d like to drink a beer now’ (subordinate clause; VF)
c. …, dass ich ein Bier trinken möchte jetzt.
that I a beer drink want now
‘… that I’d like to drink a beer now’ (subordinate clause; VF with extraposed ad-
Swedish subordinate clauses, in contrast, possess only one pre-verbal slot that is optionally filled
by certain types of adverbials (mittfältsadverbial ‘central field adverbials’), in particular sentence
adverbials, negations and other fronted elements (this is called adverbial−finite word order, AF,
in Swedish grammaticography), often resulting in the finite verb occupying the third (instead of
the second) position (V3). Example (8) shows verb-second word order in main and subordinate
clauses as opposed to adverbial−finite word order (AF) in a subordinate clause containing a neg-
(8) L1 Swedish
a. Vi kommer dricka öl i kväll.
we will drink beer in evening
‘We’re gonna drink beer tonight’ (declarative main clause; V2)
b. … att vi kommer dricka öl i kväll.
that we will drink beer in evening
‘… that we’re gonna drink beer tonight’ (subordinate clause; V2)
c. … att vi inte kommer dricka öl i kväll.
that we not will drink beer in evening
‘… that we aren’t gonna drink beer tonight’ (negated subordinate clause; AF)
For the sake of simplicity, we will now focus on subjunctional subordinate clauses with canon-
ical verb-later word-order patterns. In terms of constructional form, the German pattern can be
described as [SUBJNinitial, SBJ, (…), Vfinitefinal], i.e. as containing a clause-initial subjunction, a subject,
(optional) additional elements, and a clause-final finite verb. The Swedish pattern, on the other
hand, has the constructional form [SUBJNinitial, SBJ¹, (CFA²), Vfinite³, (…)], i.e. it contains a clause-ini-
tial subjunction, followed by a subject in the first position, (optional) central field adverbials
(CFA) in the second position, the finite verb in the third position, and, post-verbally, (optional) ad-
ditional elements.
While the patterns obviously differ substantially, an L1 speaker of German acquiring Swedish
as an AL will be able to process many constructs in the Swedish input using the pre-existing con-
structions in her constructicon. A learner will be exposed to Swedish utterances such as the one
in (9), which are instantiations of the Swedish construction [SUBJNinitial, SBJ¹, (CFA²), Vfinite³, (…)
subjunctional subordinate clause ]; however, they can be decoded using the pre-existing Ger-
man construction [SUBJNinitial, SBJ, (…), Vfinitefinal subjunctional subordinate clause ] because of the〈 〉
structural overlap between the two patterns and the lack of additional elements (cf. the German
equivalent in (10)).
(9) L1 Swedish
De sa att nätet inte funkade.
they said that internet-DEF not worked
‘They said that the internet didn’t work.’
(10) L1 German
Sie sagten, dass das Internet nicht ging.
they said that the internet not worked
‘They said that the internet didn’t work.’
Input such as in (9) does not provide any cues indicating that the Swedish pattern is any different
from the German one. As a consequence, the learner will continue to use the genuinely German
verb-final construction as a diaconstruction, i.e. across all communicative contexts including the
ones where she speaks Swedish, as long as she is not exposed to a sufficient amount of input that
does not match the entrenched verb-final pattern. In AL production, this may lead to overgenera-
tion, i.e. the learner will produce non-canonical utterances with verb-final clauses such as in (11):
(11) AL Swedish
De sa att nätet hela veckan inte funkade.
they said that internet-DEF whole-DEF week-DEF not worked
‘They said that the internet didn’t work all week.’
However, additional input will gradually lead to a differentiation of word-order patterns. For ex-
ample, the learner will be exposed to canonical variants of the pattern in (11) that do not match
the pre-existing construction. This includes Swedish utterances as in (12) (cf. the German equi-
valent in (13)):
(12) L1 Swedish
De sa att nätet inte funkade hela veckan.
they said that internet-DEF not worked whole-DEF week-DEF
‘They said that the internet didn’t work all week.’
(13) L1 German
Sie sagten, dass das Internet die ganze Woche nicht ging.
they said that the internet the whole week not worked
‘They said that the internet didn’t work all week.’
This then prompts a reorganizational sequence within the constructicon that can be represented
in an idealized fashion as follows (cf. Figure 3):
a. The pre-existing construction [SUBJNinitial, SBJ, (…), Vfinitefinal subjunctional subordinate
clause ] undergoes a constructional split, resulting in two different constructions, namely
a CGerman idioconstruction on the one hand and a diaconstruction on the other hand. The
emerging idioconstruction is formally identical to the previous one, but with an added
functional property specifying a pragmatic restriction to CGerman ([SUBJNinitial, SBJ, (…), Vfinitefi-
nal subjunctional subordinate clause; CGerman ]). The emerging diaconstruction is function-
ally identical to the previous one, but does not specify word order except for the clause-
initial position of the subjunction; this reflects the removal of a formal property: [SUBJNini-
tial, SBJ, (…), Vfinite subjunctional subordinate clause ].〈 〉
b. An additional idioconstruction is established representing both the Swedish pattern and
the pragmatic restriction to CSwedish: [SUBJNinitial, SBJ¹, (CFA²), Vfinite³, (…) subjunctional subor-
dinate clause; CSwedish ].
c. Interconstructional links are added to the constructicon, reflecting the fact that both
idioconstructions are instantiations of the diaconstruction.
[SUBJNinitial, SBJ, (…), vnitenal]
[SUBJNinitial, SBJ, (…), vnite]
[SUBJNinitial, SBJ, (…), vnitenal ]
[SUBJNinitial, SBJ¹, (CFA²), vnite³, (…)]
constructional split
removed formal property
added pragmatic property additional idioconstruction
additional interconstructional links
Figure 3: Reorganizational sequence for German and Swedish subordinate clause word patterns
While the amount and the composition of reorganizational sequences involved in language ac-
quisition differ, depending on the structural ingredients (not least the degree of diasystematicity
and the diasystematic potential of the AL and previously acquired languages; cf. Höder 2018:
63–64), we propose that the integration of AL material in the emerging multilingual con-
structicon boils down to the same basic types of reorganizational processes across all kinds of
To conclude, from a DCxG perspective, the acquisition of AL material not only results in the
entrenchment of constructions that specifically represent AL structures, but it also involves the
reorganization of previously acquired linguistic knowledge, including the addition of pragmatic
restrictions to pre-existing constructions.
4 Summary and outlook
In this contribution, we introduced a Construction Grammar approach to AL acquisition as emer-
ging multilingualism. The model builds on an earlier constructionist approach to language con-
tact and multilingualism, DCxG (e.g. Höder 2018), as well as more general usage-based concepts
of entrenchment and disentrenchment (as in Schmid’s [2017] Entrenchment-and-Conventionaliz-
ation Model). At the same time, the model aims at compatibility with current non-constructionist
approaches that embrace what has come to be known as the multilingual turn in AL research,
while also drawing on the more traditional notion of interlanguage (e.g. Selinker 1972).
Our contribution, so far, consists in the attempt to provide a detailed theoretical framework of
several of the reorganizational processes that are involved in the complex cognitive task of addi-
tional language acquisition. The main statements that our model seeks to emphasize can be sum-
marized as follows:
a. The learner’s linguistic knowledge, including her interlanguage variety of the AL, can be
redefined in terms of an emerging multilingual constructicon.
b. The process of AL acquisition does not only pertain to the processing of AL material or
the establishment of AL constructions, but affects all language(s) in the constructicon.
c. Reorganization in the constructicon is determined by general (i.e. not AL-specific) cog-
nitive processes, including the modification of pre-existing constructions, the addition of
necessary idioconstructions and the entrenchment and disentrenchment of diaconstruc-
tions and interconstructional links.
d. Before AL acquisition, a monolingual speaker’s linguistic knowledge consists of
(strongly entrenched) diaconstructions only, i.e. of constructions that are not restricted to
any specific set of communicative contexts. Language-specificity first becomes relevant
as a consequence of AL acquisition, with an emerging differentiation between dia- and
idioconstructions. To put it more simply: You do not have one language in your con-
structicon until you have two.
While this model primarily targets linguists working within the framework of Construction
Grammar, non-constructionist scholars working on AL acquisition should also find it useful.
With its built-in multilingual baseline and the focus on individual constructions in the learner’s
repertoire and reorganizational processes at the constructional level instead of different language
systems, the DCxG approach offers an architecture that is open for whatever new linguistic ma-
terial the learner acquires, without any need for an a priori distinction between different lan-
guages within individual learners’ repertoires. While this approach certainly resonates well with
work that emphasizes multilingual speakers’ freedom and creativity in using and combining lin-
guistic material from different languages, it by no means implies that it would be somehow use-
ful to completely abandon the notion of ‘languages’ (including acquiring or using different ‘lan-
guages’) in a more general sense, as long as the term refers to conventionalized communicative
practices that are socially relevant for speaker communities as well as AL learners. The monolin-
gual bias in research on both multilingualism and AL acquisition has certainly contributed to the
misunderstanding from a cognitive point of view that keeping languages apart is somehow
more natural than, for example, combining linguistic elements from different languages into mul-
tilingual utterances (cf. Sabino 2018 for a discussion of the ideological aspects). The DCxG ap-
proach does not provide any cognitive argument for such a view. However, linguistic knowledge
also entails knowledge about social conventions on language use: using one language at a time
means sticking to linguistic material that, by convention, belongs to the same set of communicat-
ive contexts within a speaker community. How and why such conventions are or are not ac-
quired, how L1 speakers of the learner’s AL react to violations of such conventions, and whether
or not groups of emerging multilinguals may establish divergent conventions (e.g. by conven-
tionalizing translanguaging patterns) – these are questions of a normative kind for societies and
educators, but rather empirical questions for linguists. To address these empirical questions, it is,
in turn, crucial to discuss and, if necessary, challenge linguistic norms and to acknowledge that
there are many contexts where the use and cognitive processing of several languages at the same
time are relevant (cf. Ortega 2019).
Boas, Hans C. & Steffen Höder. 2018. Construction grammar and language contact. An introduc-
tion. In Hans C. Boas & Steffen Höder (eds.), Constructions in contact. Constructional per-
spectives on contact phenomena in Germanic languages (Constructional Approaches to Lan-
guage 24), 5–36. Amsterdam/Philadelphia: Benjamins.
Bybee, Joan. 2010. Language, usage and cognition. Cambridge: Cambridge University Press.
Bybee, Joan L. 2013. Usage-based theory and exemplar representations. In Thomas Hoffmann &
Graeme Trousdale (eds.), The Oxford handbook of Construction Grammar, 49–69. Oxford:
Oxford University Press.
Cappelle, Bert. 2017. What’s pragmatics doing outside constructions? In Ilse Depraetere &
Raphael Salkie (eds.), Semantics and pragmatics. Drawing a line (Logic, Argumentation &
Reasoning 11), 115–151. Cham: Springer.
Cook, Vivian 1991. The poverty-of-the-stimulus argument and multicompetence. Second Lan-
guage Research 7, 103–117.
Cook, Vivian J. 1993. Wholistic multi-competence – jeu d’esprit or paradigm shift? In Bernhard
Kettemann & Wilfried Wieden (eds.), Current issues in European second language acquisi-
tion research (Tübinger Beiträge zur Linguistik 378), 3–8. Tübingen: Narr.
Cook, Vivian & Li Wei (eds.). 2016. The Cambridge handbook of linguistic multi-competence.
Cambridge: Cambridge University Press.
Dąbrowska, Ewa. 2019. Individual differences in grammatical knowledge. In Ewa Dąbrowska &
Dagmar Divjak (eds.), Cognitive linguistics, vol. 3: Key topics, 231–250. Berlin/Boston: de
De Knop, Sabine & Gaëtanelle Gilquin. 2016. Exploring L2 constructionist approaches. In
Sabine De Knop & Gaëtanelle Gilquin (eds.), Applied Construction Grammar (Applications
of Cognitive Linguistics 32), 3−18. Berlin/Boston: de Gruyter.
Diessel, Holger. 2013. Construction Grammar and first language acquisition. In Thomas Hoff-
mann & Graeme Trousdale (eds.), The Oxford handbook of Construction Grammar, 347–364.
Oxford: Oxford University Press.
Diessel, Holger. 2019. Usage-based Construction Grammar. In Ewa Dąbrowska & Dagmar
Divjak (eds.), Cognitive linguistics, vol. 2: A survey of linguistic subfields, 50–80. Berlin: de
Divjak, Dagmar & Catherine L. Caldwell-Harris. 2019. Frequency and entrenchment. In Ewa
Dąbrowska & Dagmar Divjak (eds.), Cognitive linguistics, vol. 1: Foundations of language,
61–86. Berlin/Boston: de Gruyter.
Eckman, Fred R. 2010. Linguistic typology and second language acquisition. In Jae J. Song
(ed.), The Oxford handbook of linguistic typology, 618–633. Oxford: Oxford University Press.
Ellis, Nick C. 2006. Selective attention and transfer phenomena in L2 acquisition. Contingency,
cue competition, salience, interference, overshadowing, blocking, and perceptual learning.
Applied Linguistics 27, 164–194.
Ellis, Nick. 2013. Construction grammar and second language acquisition. In Thomas Hoffmann
& Graeme Trousdale (eds.), The Oxford handbook of Construction Grammar, 365–378. Ox-
ford/New York: Oxford University Press.
Ellis, Nick C. & Stefanie Wulff. 2014. Usage-based approaches to SLA. In Bill VanPatten & Jes-
sica Williams (eds.), Theories in second language acquisition. An introduction. New York:
Routledge, 75–93.
Ellis, Nick & Stefanie Wulff. 2019. Second language acquisition. In Ewa Dąbrowska & Dagmar
Divjak (eds.), Cognitive linguistics, vol. 2: A survey of linguistic subfields, 182–207. Berlin:
de Gruyter.
García, Ofelia. 2009. Bilingual education in the 21st century. A global perspective. Chichester:
García, Ofelia & Li Wei 2014. Translanguaging: language, bilingualism and education. Basing-
stoke: Palgrave Macmillan.
Gardner-Chloros, Penelope. 2009. Code-switching. Cambridge: Cambridge University Press.
Goldberg, Adele E. 2006. Constructions at work. The nature of generalization in language. Ox-
ford: Oxford University Press.
Goldberg, Adele E. 2013. Constructionist approaches. In Thomas Hoffmann & Graeme Trous-
dale (eds.), The Oxford handbook of Construction Grammar, 15–31. Oxford: Oxford Univer-
sity Press.
Goldberg, Adele E. 2019. Explain me this. Creativity, competition, and the partial productivity of
constructions. Princeton: Princeton University Press.
Gries, Stefan Th. & Stefanie Wulff. 2005. Do foreign language learners also have constructions?
Annual Review of Cognitive Linguistics 3, 182–200.
Grosjean, François. 1989. Neurolinguists, beware! The bilingual is not two monolinguals in one
person. Brain and Language 36, 3–15.
Grosjean, François. 2008. Studying bilinguals. Oxford: Oxford University Press.
Haspelmath, Martin. 2006. Against markedness (and what to replace it with). Journal of Lin-
guistics 42, 25–70.
Hendrikx, Isa. 2019. The acquisition of intensifying constructions in Dutch and English by
French-speaking CLIL and non-CLIL students: Cross-linguistic influence and exposure ef-
fects. Louvain-la-Neuve: Université catholique de Louvain. Ph.D. dissertation.
Hilpert, Martin. 2019. Construction grammar and its application to English. 2nd edn. Edin-
burgh: Edinburgh University Press.
Hilpert, Martin & Holger Diessel. 2017. Entrenchment in Construction GrammarConstruction
Grammar. In Hans-Jörg Schmid (ed.), Entrenchment and the psychology of language learn-
ing. How we reorganize and adapt linguistic knowledge, 57–74.Berlin: De Gruyter Mouton/
Washington: American Psychological Association.
Höder, Steffen. 2012. Multilingual constructions: a diasystematic approach to common struc-
tures. In Kurt Braunmüller & Christoph Gabriel (eds.), Multilingual individuals and multilin-
gual societies (Hamburg Studies on Multilingualism 13), 241–257. Amsterdam: Benjamins.
Höder, Steffen. 2014a. Constructing diasystems. Grammatical organisation in bilingual groups.
In Tor A. Åfarli & Brit Mæhlum (eds.), The sociolinguistics of grammar (Studies in Language
Companion Series 154), 137–152. Amsterdam: Benjamins.
Höder, Steffen. 2014b. Phonological elements and Diasystematic Construction Grammar. Con-
structions and Frames 6, 202–231.
Höder, Steffen. 2018. Grammar is community-specific: Background and basic concepts of Dia-
systematic Construction Grammar. In Hans Boas & Steffen Höder (eds.), Constructions in
contact. Constructional perspectives on contact phenomena in Germanic languages (Con-
structional Approaches to Language 24), 37–70. Amsterdam: Benjamins.
Höder, Steffen. 2019. Phonological schematicity in multilingual constructions: a diasystematic
perspective on lexical form. Word Structure 12, 334–352.
Jarvis, Scott & Aneta Pavlenko. 2008. Crosslinguistic influence in language and cognition. New
York, Abingdon: Routledge.
Johansson, Stig 2008. Contrastive analysis and learner language: A corpus-based approach.
Oslo: University of Oslo.
Lado, Robert. 1957. Linguistics across cultures. Applied linguistics for language teachers. Ann
Arbor: University of Michigan Press.
Larsen-Freeman, Diane. 2011. A complexity theory approach to second language development/
acquisition. In Dwight Atkinson (ed.), Alternative approaches to second language acquisition,
48–72. London: Routledge.
Li Wei. 2011. Moment Analysis and translanguaging space: Discursive construction of identities
by multilingual Chinese youth in Britain. Journal of Pragmatics 5, 1222–1235.
Li Wei. 2018. Translanguaging as a practical theory of language. Applied Linguistics 39, 9–30.
Loenheim, Lisa. 2019. Att tolka det sammansatta. Befästning och mönster i första- och andra-
språkstalares tolkning av sammansättningar (Meijerbergs arkiv för svensk ordforskning 43).
Göteborg: Meijerbergs institut för svensk etymologisk forskning, Göteborgs universitet.
MacWhinney, Brian. 2017. Entrenchment in second-language learning. In Hans-Jörg Schmid
(ed.), Entrenchment and the psychology of language learning. How we reorganize and adapt
linguistic knowledge, 343–366. Berlin: De Gruyter Mouton/Washington: American Psycholo-
gical Association.
Matras, Yaron. 2009. Language contact. Cambridge: Cambridge University Press.
Matthews, Danielle & Grzegorz Krajewski. 2019. First language acquisition. In Ewa Dąbrowska
& Dagmar Divjak (eds.), Cognitive linguistics, vol. 2: A survey of linguistic subfields, 159–
181. Berlin: de Gruyter.
May, Stephen (ed.). 2013. The multilingual turn. The implications for SLA, TESOL and bilingual
education. New York/London: Routledge.
Ortega, Lourdes. 2009. Understanding second language acquisition. London. Routledge.
Ortega, Lourdes. 2019. SLA and the study of equitable multilingualism. The Modern Language
Journal 103, 23–38.
Pavlenko, Aneta. 2003: “I never knew I was a bilingual”: Reimagining teacher identities in
TESOL. Journal of Language, Identity & education 2, 251–268.
Peirce, Bonny Norton 1995. Social identity, investment, and language learning. TESOL
Quarterly 29, 9–31.
Perani, Daniela & Jubin Abutalebi. 2005. The neural basis of first and second language pro-
cessing. Current Opinion in Neurobiology 15, 202–206.
Ringbom, Håkan. 1978. The influence of the mother tongue on the translation of lexical items.
Interlanguage Studies Bulletin 3, 80–101.
Römer, Ute, Matthew B. O’Donnell & Nick C. Ellis. 2014. Second language learner knowledge
of verb-argument constructions. Effects of language transfer and typology. The Modern Lan-
guage Journal 98, 952–975.
Sabino, Robin. 2018. Languaging without languages. Beyond metro-, multi-, poly-, pluri- and
translanguaging (Brill’s Studies in Language, Cognition and Culture 18). Leiden/Boston:
Schmid, Hans-Jörg. 2015. A blueprint of the Entrenchment-and-Conventionalization Model.
Yearbook of the German Cognitive Linguistics Association 3, 3–25.
Schmid, Hans-Jörg. 2016. Why cognitive linguistics must embrace the social and pragmatic di-
mensions of language and how it could do so more seriously. Cognitive Linguistics 27, 543–
Schmid, Hans-Jörg. 2017. A framework for understanding linguistic entrenchment and its psy-
chological foundations. In Hans Jörg-Schmid (ed.), Entrenchment and the psychology of lan-
guage learning. How we reorganize and adapt linguistic knowledge, 9–36. Berlin: De Gruyter
Mouton/Washington: American Psychological Association.
Selinker, Larry 1969. Language transfer. General Linguistics 9, 67–92.
Selinker, Larry. 1972. Interlanguage. International Review of Applied Linguistics 10, 209–232.
Steinkrauss, Rasmus & Monika S. Schmid. 2017. Entrenchment and language attrition. In Hans-
Jörg Schmid (ed.), Entrenchment and the psychology of language learning. How we reorgan-
ize and adapt linguistic knowledge, 367–384. Berlin: De Gruyter Mouton/Washington: Amer-
ican Psychological Association.
Theakston, Anna L. 2017. Entrenchment in first language learning. In Hans-Jörg Schmid (ed.),
Entrenchment and the psychology of language learning. How we reorganize and adapt lin-
guistic knowledge, 315–342. Berlin: De Gruyter Mouton/Washington: American Psycholo-
gical Association.
Tomasello, Michael. 2003. Constructing a language. A usage-based theory of language acquisi-
tion. Cambridge: Harvard University Press.
Van Goethem, Kristel and Isa Hendrikx. This volume. Intensifying constructions in second lan-
guage acquisition: A diasystematic-constructionist approach.
Volodina, Elena et al. 2016. SweLL on the rise: Swedish Learner Language corpus for European
Reference Level studies. In Nicoletta Calzolari et al. (eds.), Proceedings of the Tenth Interna-
tional Conference on Language Resources and Evaluation (LREC 2016), Portorož, Slovenia,
May 23–28, 2016.
Weinreich, Uriel. 1953. Languages in contact. Findings and problems (Publications of the Lin-
guistic Circle of New York 1). New York.
The aim of the current paper is to reinterpret some results of two previous studies on the mastery of figurative expressions from the perspective of usage-based linguistics. The reanalysis aims to shed more light on the learning and use of figurative language by multilingual students by exploring the complex interplay of linguistic creativity, expressivity, and conventionality in figurative expressions. The reinterpretation shows that many of the examples that were previously categorized as novel figurative expressions used in students’ writing, can be analyzed as instances of regular patterns, i.e. constructions, with certain lexical idiosyncrasies. Modifications of conventionalized figurative expressions are discussed and reinterpreted in terms of strength of entrenchment of links between form and meaning within certain constructions or links between constructions and conventionalized pragmatic information in the multilinguals’ mental construction. Implications for the treatment of Swedish figurative expressions in the second language class room are, in line with previous research, that focusing on regularity might reduce unpredictability, often seen as the core difficulty in the learning of such expressions in an L2. The paper also offers some directions for further investigation of the socio-cognitive processes involved in the learning of figurative language in an additional language.
Full-text available
When communicating across closely related languages or varieties (e.g. in interdialectal communication or in regions such as Mainland Scandinavia), speakers have to learn how to decode words that show partial phonological differences from the equivalents in their L1. Although contact situations like these are rather common, interlingual decoding has scarcely been addressed in the CxG literature. As a contribution to this field of research, the paper discusses how (a particular stage in) emerging receptive multilingualism can be modelled from a CxG perspective. Specifically, it deals with the idea that repeated interlingual decoding generates partially schematic cross-linguistic constructions mirroring the speaker’s knowledge about sound correspondences, as suggested by Diasystematic Construction Grammar ( Höder 2019 ).
Full-text available
This article discusses the role of intra-word phonological schematicity in multilingual con-structicons from a Diasystematic Construction Grammar perspective. It argues that, in particular with communities that use two or more typologically similar and/or closely related languages, many lexical elements (e.g. cognates) exhibit regular sound correspondences that can be analysed as consisting of different types of phonological schemas. In this view, there is a division of labour between schematic constructions that specify the words' referential meaning and others that specify their belonging to one of the 'languages', with language-specificity defined as a pragmatic property of constructions. The focus is on the question whether generalizations at this level of schematicity and abstraction are cognitively real and what can count as evidence for their existence from a usage-based perspective.