句法复杂性发展（2017）

A Cross-Linguistic Perspective on
Syntactic Complexity in L2
Development: Syntactic Elaboration
and Diversity
BASTIEN DE CLERCQ1 and ALEX HOUSEN2
1
Vrije Universiteit Brussel, TALK, Pleinlaan 2, 1050 Brussels, Belgium Email: [email protected]
2
Vrije Universiteit Brussel, TALK, Pleinlaan 2, 1050 Brussels, Belgium Email: [email protected]
Syntactic and linguistic complexity have been studied extensively in applied linguistics as indicators of
linguistic performance, development, and proficiency. Recent publications have equally highlighted the
reductionist approach taken to syntactic complexity measurement, which often focuses on one or two
measures representing complexity at the level of clause-linking or the sentence, but eschews complexity
measurement at other syntactic levels, such as the phrase or the clause. Previous approaches have also
rarely incorporated measures representing the diversity of syntactic structures in learner productions.
Finally, complexity development has rarely been considered from a cross-linguistic perspective, so that
many questions pertaining to the cross-linguistic validity of complexity measurement remain. This article
reports on an empirical study on syntactic complexity development and introduces a range of syntactic
diversity measures alongside frequently used measures of syntactic elaboration. The study analyzed 100
English and 100 French second language oral narratives from adolescent native speakers of Dutch, sit-
uated at 4 proficiency levels (beginner–advanced), as well as native speaker benchmark data from each
language. The results reveal a gradual process of syntactic elaboration and syntactic diversification in
both learner groups, while, especially in French, considerable differences between learners and native
speakers reside in the distribution of specific clause types.
Keywords: complexity; syntax; L2 French; L2 English; cross-linguistic
A CONSIDERABLE BODY OF RESEARCH ON between these components (Miestamo, 2009).

Second Language Acquisition (SLA) has been One of the assumptions in second language (L2)
concerned with finding objective ways to de- complexity research has been that, as L2 develop-
scribe language development and assess linguistic ment progresses, learners use more elaborate and
proficiency (Larsen–Freeman, 2006). This re- diverse language, as characterized by, for instance,
search has identified linguistic complexity, along- longer sentences, more embedding, and a greater
side accuracy and fluency, as a key component range of vocabulary and morphology (e.g., Ellis,
of linguistic proficiency, an indicator of language 2009; Housen & Kuiken, 2009; Wolfe–Quintero,
development and a descriptor of linguistic per- Inagaki, & Kim, 1998).
formance (Housen, Kuiken, & Vedder, 2012; More recently, this assumption has been chal-
Ortega, 2012). In general terms, linguistic com- lenged and several researchers have called for an
plexity can be defined as an absolute, essentially approach to linguistic complexity that, on the one
quantitative property of a (linguistic) unit or sys- hand, considers it as a multidimensional notion
tem, determined by the number of its compo- (e.g., Bulté & Housen, 2012; Norris & Ortega,
nents and the number and types of connections 2009) and, on the other hand, specifies the re-
lation between complexity and proficiency more
The Modern Language Journal, 101, 2, (2017) accurately (e.g., Pallotti, 2009). Previous research
DOI: 10.1111/modl.12396 has been criticized for adopting a ‘reductionist’
0026-7902/17/315–334 $1.50/0 approach to L2 complexity (Bulté & Housen,

C 2017 The Modern Language Journal 2012) through its singular focus on syntactic
15404781, 2017, 2, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/modl.12396 by Cochrane Japan, Wiley Online Library on [19/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
316 The Modern Language Journal 101 (2017)
complexity and by calculating only one or two role of linguistic complexity in the process of L2
measures that exclusively target sentence-level development with special attention to the poten-
elaboration or amount of subordination while ig- tial role of cross-linguistic factors. Next, we de-
noring other aspects, such as phrasal elaboration scribe the research questions and methodology.
and the diversity of structures at various syntactic We conclude with the results, which are further
levels (Bulté & Housen, 2012; Lambert & Kormos, discussed in the concluding section.
2014; Norris & Ortega, 2009). While recent stud-
ies have taken some of these criticisms into ac- DEFINING AND MEASURING SYNTACTIC
count by considering a wider range of syntactic COMPLEXITY
phenomena (e.g., Biber, Gray, & Poonpon, 2011;
Byrnes, Maxim, & Norris, 2010; Crossley et al., Previous definitions of syntactic complexity
2007; Lu, 2010, 2011; Verspoor, Schmid, & Xu, and of the broader notion of linguistic com-
2012; Vyatkina, 2012, 2013; Yoon & Polio, 2016), plexity allude to a variety of seemingly disparate
the influence of many relevant factors in deter- notions, ranging from linguistic phenomena such
mining syntactic complexity and its development as recursion, ellipsis, or the degree of paradig-
remains underexplored. For instance, few stud- matic and syntagmatic variation, to factors such
ies have incorporated syntactic diversity measures, as redundancy and markedness, as well as cog-
even though the notion of diversity has domi- nitively or developmentally related aspects such
nated previous research on lexical and, more re- as a feature’s difficulty or developmental timing
cently, morphological complexity (e.g., Malvern (Bartning & Kirchmeyer, 2003; Bulté & Housen,
et al., 2004; Pallotti, 2015). Furthermore, previ- 2012; Ellis, 2009; Housen & Simoens, 2016;
ous SLA research has left under-investigated the Salamoura & Saville, 2010; Szmrecsanyi, 2004;
potential influence of cross-linguistic variables on Szmrecsanyi & Kortmann, 2012). These aspects
the link between complexity in L2 production can broadly be related to two different ap-
and linguistic proficiency. Studies in language proaches to complexity: an absolute approach,
typology have concentrated on (grammatical) which views complexity as an inherent property of
complexity trade-offs, whereby the degree of the linguistic system and of linguistic structures,
complexity in one linguistic dimension is off- that is, as a function of the number of elements
set in another (cf. Ehret & Szmrecsanyi, 2016; in a system or structure and the number of rela-
Fenk–Oczlon & Fenk, 2014). The cross-linguistic tionships among these elements, and a relative
variability of linguistic complexity thus further approach which defines complexity in relation
challenges the extent to which measures tapping to the language user and the cognitive cost or
into syntactic complexity relate to linguistic de- difficulty invoked in processing and acquiring
velopment and proficiency in a similar way across certain linguistic structures (cf. Bulté & Housen,
languages. 2012; Dahl, 2009; Han & Lew, 2012; Miestamo,
This article is primarily concerned with syntac- 2008, 2009).
tic complexity and complements our two previous Despite the focus on the absolute perspective
studies on cross-linguistic variability of lexical in this article, the two approaches are clearly
(De Clercq, 2015) and morphological complexity linked. On the one hand, complexity research
(De Clercq & Housen, 2016). It aims to bridge is concerned with finding objective, quantitative
the gap in current research by investigating ways to describe the multifaceted and often non-
how syntactic complexity develops in relation to linear nature of the language learning process
general L2 development. It adopts a systematic (e.g., Bulté & Housen, 2012; Pallotti, 2009, 2015;
approach to syntactic complexity that not only Larsen–Freeman, 2006). On the other hand, a
considers syntactic complexity through syntactic related research agenda extends these aims by
elaboration but also through the diversity of explicitly searching for measures that capture the
syntactic structures in oral production data. To construct of language proficiency and focuses
this end, we use a corpus of 250 oral narratives, on language features that are expected to be
representing 100 L2 French and 100 L2 English linearly or closely related to proficiency (e.g.,
productions at four proficiency levels as well as 25 Han & Lew, 2012; Lambert & Kormos, 2014;
native speaker benchmark productions for each Ortega, 2012). Yet some researchers from both
language. typological and SLA backgrounds have preferred
In the next section, we examine the notion of to clearly distinguish the two notions. Pallotti
linguistic and, more specifically, syntactic com- (2015), for instance, argues that the confla-
plexity in greater detail, focusing on measure- tion of the notions ‘complex,’ ‘advanced,’ and
ment practices in particular. We then review the ‘difficult’ may lead to such circular or tautological
Bastien De Clercq and Alex Housen 317
statements as “complex [structurally complex] complexity measures have relied on wide-ranging
structures are often more complex [difficult] and operationalizations of the underlying construct
complex [acquired late]” (p. 119). Furthermore, and have included measures of clause length
as will be argued in the next section, competent or (Norris & Ortega, 2009), the length of phrases
proficient language use is not necessarily linguis- surrounding the verb (Pallotti, 2015), and mea-
tically more complex in the absolute sense, since sures counting the length of noun phrases(NPs)
effective communication may rely on simple(r) (Bulté & Housen, 2014) or the number of
language as well. In accordance with this view, this modifiers per NP (Crossley et al., 2007).
study therefore investigates linguistic complexity Measures based on clausal subordination have
in a primarily descriptive way in order to account equally been a staple in SLA research on syntactic
for how the complexity of various components complexity, with many studies relying on a ratio
of the linguistic system can increase, decrease, or calculating the number of (subordinate) clauses
level off throughout language development. per multiclausal unit (sentence, AS-unit, T-unit),
Syntactic complexity is arguably the most fre- even though such measures have not been with-
quently studied dimension of linguistic complex- out criticism, either. Bulté and Housen (2012),
ity in SLA research, more so than morphological for instance, find subordination measures prob-
or lexical complexity, so that syntactic complexity lematic since they only represent complexity at
is frequently treated as isomorphic with gram- the supraclausal level, and not at the clausal or
matical complexity or linguistic complexity in phrasal level, nor do they gauge clause-linking
general (Bulté & Housen, 2012; Ellis, 2009). through coordination. Moreover, Bulté and
Syntactic complexity has long been associated Housen (2012) point out that subordination
with syntactic elaboration, reflected by syntactic measures are sometimes used as measures of rela-
length and subordination.1 Beaman (1984), for tive complexity (e.g., difficulty), since they single
instance, states that syntactic complexity implies out a specific structure assumed to be of partic-
“longer sentences and more subordinate clauses ular importance in the SLA process (e.g., they
that reveal more complex structural relation- allegedly pose more learning difficulties or are ac-
ships” (p. 45). Syntactic length (in words or mor- quired later than other syntactic linking devices).
phemes), then, is one of the most commonly used The specificity of subordination as the sole means
indicators of syntactic complexity. Such measures of complexification has equally been criticized for
assume that changes in length reflect underlying being less representative of complexity in written
structural changes as well. While such an ap- production than in oral production (Beaman,
proach may seem superficial or simplistic, the use 1984; Biber & Gray, 2010; Biber et al., 2011).
of length may nonetheless function adequately as Finally, some researchers have called for a more
a proxy for the construct of syntactic complexity fine-grained analysis of subordination, claiming
and as an approximation of more fine-grained that various types of subordination may surface at
measures. Szmrecsanyi (2004), for instance, different points in development (Buysse, Housen,
found very high correlations (r = .976 – .989) be- & Pierrard, 2017; Lambert & Kormos, 2014). Sim-
tween length measures and syntactic node counts ilarly, Schleppegrell (1992) emphasizes that
in both written and oral English data. Previous subordinate clauses can fulfill more general
research practices, however, have been criticized discourse functions or serve as “simple parallel
for focusing almost exclusively on syntactic length statements” (p. 118), rather than establish links
at the supraclausal level (e.g., Terminal (T)-units of clausal dependency, so that a more nuanced
in writing research [Hunt, 1965], Analysis of analysis of subordination may yield a different
Speech (AS)-units in spoken language research perspective on the development of clause-linking.
[Foster, Tonkyn, & Wigglesworth, 2000]), while If considerable attention has been devoted to
complexity may manifest itself differently at the measurement of complexity through syntactic
other syntactic levels, notably the phrasal and elaboration, few studies have considered syntactic
clausal level, reflecting proficiency, modality, or complexity in the form of diversification in equal
genre differences (Biber & Gray, 2010; Biber measure. Although Ellis and Barkhuizen (2005)
et al., 2011; Byrnes et al., 2010; Kuiken & Vedder, mention both the use of “elaborated language”
2007, 2012; Norris & Ortega, 2009; Pallotti, 2015; and the use of “a wide range of structures” as
Yoon & Polio, 2016). With regard to the phrasal key dimensions of complexity (p. 139), measures
level, Norris and Ortega (2009) have argued that of diversity are mostly restricted to the study of
phrasal elaboration constitutes an advanced area the lexicon (e.g., De Clercq, 2015; Jarvis, 2013;
of syntactic development, particularly in formal Malvern et al., 2004), for which a considerable
and academic writing. Recent uses of phrasal number of indices have been developed, and,
more recently, morphology (Brezina & Pallotti, view on complexity advocated by, for instance,
2017; De Clercq & Housen, 2016; Pallotti, 2015). Pallotti (2015). Finally, previous complexity re-
If syntactic complexity can be related to “the ex- search has also shown a preference toward simple
tent to which the learner has been able to create indices representing the complexity of a specific
a full syntactic tree in his or her interlanguage” language dimension by a single value (e.g., the
(Towell, 2012, p. 55), complexity in production mean length of a syntactic unit as a measure of
can thus in part be seen as a manifestation of the elaboration), while the analysis of the distribution
degree of elaboration allowed by the learner’s of structures does not offer such a simple metric.
interlanguage, but also of the range of mastered While a simple measure such as the mean length
structures. Since researchers have emphasized of clause is easily interpreted on a scale from less
that high structural elaboration alone does not to more complex, the percentage of nonfinite
necessarily imply high proficiency or effective adverbial clauses in a text is less transparent as
communication (Han & Lew, 2012; Lambert a measure of complexity unless it is accompa-
& Kormos, 2014; Ortega, 2003; Pallotti, 2009), nied by measures representing other types of
diversity measures may thus further refine the subordinate clauses showing how each structure
descriptive toolset used in complexity research by contributes to the overall diversity of the text.
considering the variation of syntactic structures in
relation to proficiency. For instance, speakers
A CROSS-LINGUISTIC PERSPECTIVE
who overuse some seemingly complex syntactic
ON THE DEVELOPMENT OF SYNTACTIC
patterns (e.g., I think that + complement clause)
COMPLEXITY
may achieve high syntactic elaboration using con-
ventional measures but low diversity, so that the The variation of linguistic complexity across
combined use of syntactic elaboration and diver- languages has long been the focus of typological
sity measures may lead to a better description and research, which has been particularly concerned
understanding of (syntactic) complexity as a with investigating the equi-complexity hypothe-
whole. sis (Dahl, 2009). This hypothesis states that all
Previous studies interested in mapping the languages share a comparable degree of total (ab-
range of structures in learners’ productions solute) complexity and that linguistic subdimen-
have typically examined the distribution of sions (e.g., syntax and morphology) entertain
specific structures (e.g., Buysse et al., 2017, and trade-off relationships so that high complexity in
Welcomme, 2013, for types of subordinate clauses one domain is offset by low complexity in another.
in L2 French; Ågren, Granfeldt, & Schlyter, 2012, Such trade-offs between morphological and syn-
for, among others, variety in verbal tense use). tactic complexity have been statistically observed
Such methods have, however, rarely been in- in a wide range of languages, both diachroni-
tegrated in more traditional L2 complexity cally and synchronically (cf. Ehret & Szmrecsanyi,
research (for exceptions, see Biber et al., 2011, 2016; Fenk–Oczlon & Fenk, 2008), even though
and Verspoor et al., 2012). Moreover, in some correlations between complexity dimensions do
cases the link with previously established ways of not necessarily imply that these subdimensions
studying syntactic complexity is not always clear. are in perfect balance and overall complexity
For instance, in a study on planning effects on L2 differences may thus exist regardless of trade-
writing in English, Ellis and Yuan (2004) comple- offs (Fenk–Oczlon & Fenk, 2014). Cross-linguistic
mented their analysis of syntactic elaboration (a complexity differences may also be driven by stylis-
subordination ratio) with an ambiguous measure tic or rhetorical preferences rather than strictly
of ‘syntactic variety,’ which only represented the grammatical factors, so that the decision to en-
variety of verb forms. While verb form variety may code information using a specific syntactic re-
be an interesting variable in its own right, it is source (e.g., coordination or subordination) is
unclear how this measure represents structural not necessarily grammaticalized but rather indi-
syntactic variety rather than morphological diver- cates a preference of the language community
sity. As for measures counting the occurrence of (Doherty, 2005). For instance, while both French
specific structures, the overall pool of syntactic and English allow for comparable paratactic and
features is so extensive that an arguably unavoid- hypotactic devices, coordination may be a more
able preference toward developmentally relevant preferred way of clause-linking in English while
structures, that is, structures that are expected subordination is more favored in French (Cosme,
to appear at specific stages of development, may 2007). Similarly, Cosme (2007) notes that infor-
make this method irreconcilable with the absolute mation expressed at the clausal level in English is
typically integrated at the phrasal level in French. start producing simple clauses organized around
Finally, some languages may have dedicated mor- finite verbs (Klein & Perdue, 1997). Subsequent
phological means to encode subordinate clauses developmental stages involve the shift from sim-
(e.g., the subjunctive in many European lan- ple clauses to the use of clause linking and
guages), while others may not, such as Finnish, a progressive diversification of functional rela-
where most subordinate clauses do not differ mor- tions between clauses and their formal encoding,
phologically or syntactically from main clauses. with clause types such as infinitival or participial
The impact of such cross-linguistic differences clauses typically being found at higher proficiency
has rarely been explicitly considered in SLA levels. Simultaneously, learners gradually attempt
studies on complexity. Previous research has im- to master an increasingly wide array of grammati-
plicitly assumed the relationship between com- cal devices expressing, for instance, negation and
plexity and development or proficiency to be question formation, which present specific hur-
similar across languages. However, if the degree dles to overcome in the two languages (e.g., split
of syntactic complexity and its morphological cor- negatives in French and do-support in English).
relates vary cross-linguistically, its use as an in- Several authors have suggested that development
dicator of linguistic development or proficiency at advanced levels, especially in written academic
may vary accordingly. For instance, Martin et al. language, involves noun phrase elaboration, as in-
(2010) claims that subordination-based measures formation previously encoded as a clause is em-
may be less revealing in terms of developmen- bedded at the phrasal level (Byrnes, 2009; Ortega,
tal schedules for a language like Finnish, for 2012).
which subordination ratios would thus need to be While these studies seem to emphasize com-
complemented with other complexity measures. mon tendencies across L2 English and L2 French,
Previous explicitly cross-linguistic studies on their results should nonetheless be considered
syntactic complexity in SLA are scarce (e.g., with care, since they are equally the product
Kuiken & Vedder, 2012, on L2 Italian and L2 of different research methodologies. First, these
Dutch), although individual studies on the two studies involve learners from different L1 back-
languages investigated in this article, English grounds and learners may transfer rhetorical
and French, indicate converging developmental strategies from their first languages, which may in
tendencies. In the case of L2 French, the devel- turn affect the degree of complexity in their use
opment of syntactic complexity (and of clause of L2 French or English (cf. Neff et al., 2004).
linking in particular) has been well documented Second, the studies cover language develop-
for L1 Swedish learners (e.g. Ågren et al., 2012; ment in both writing and speaking, yet modality
Bartning & Kirchmeyer, 2003; Bartning & can equally account for considerable complexity
Schlyter, 2004; Gunnarsson, 2012), for L1 Dutch differences (Biber et al., 2011; Kuiken & Vedder,
(Welcomme, 2013), L1 Arabic (Perdue, 1993; 2007, 2012). The use of subordination as a means
Véronique, 2009), and L1 Spanish (Benazzo & of complexification, for instance, may be more
Starren, 2007; Perdue, 1993; Véronique, 2009). characteristic of speaking than writing, for which
These French L2 studies have often considered phrasal embedding plays a more important role
clause linking and syntactic development in their (Biber & Gray, 2010). The study of complexity in
own right, emphasizing qualitative analyses. In oral data poses a number of specific methodolog-
contrast, relevant studies on L2 English have ical challenges as well. Utterances in oral data
typically adopted a quantitative approach and are typically more fragmented and elliptic than
have focused on a broader notion of linguistic in writing, so that other syntactic units, especially
complexity that encompasses both syntactic and at the supraclausal level, have been proposed for
lexical complexity, which is often considered in spoken language analysis (notably the AS-unit by
relation to the notions of fluency and accuracy Foster et al., 2000). In addition, the analysis of
(e.g., Bulté & Housen, 2014; Iwashita et al., 2008; oral data has revealed different functions of some
Kormos & Trebits, 2012; Verspoor et al., 2012; see coordinating and subordinating conjunctions.
also Bulté & Housen, 2012, and Wolfe–Quintero For instance, whereas the use of coordinating con-
et al., 1998, for reviews). junctions is often considered an important clause-
Generally, these studies have indicated that, linking device in early language development,
in both languages, utterances in initial stages of in spoken language it often has a filler function,
acquisition are often noun-based verbless struc- serving as a discourse-organizing device rather
tures that appear alongside rudimentary clauses, than as a true syntactic coordinator (Beaman,
typically with nonfinite verbs. Gradually, learners 1984).
Other task effects seem to influence the degree complexity as syntactic elaboration (length of
of complexity in production in considerable ways AS-unit and subordination ratio) and also consid-
as well. For instance, the degree of interaction- ers this complexity dimension at the phrasal and
ality has been found to influence complexity in clausal level. At the same time, our study exam-
the sense that monologic tasks facilitate the use ines the hitherto under-investigated dimension
of higher syntactic elaboration more than dialogic of syntactic diversity by introducing a num-
tasks (Michel, Kuiken, & Vedder, 2007). Another ber of methods for measuring this complexity
study on L2 writing in different genres (Yoon & dimension.
Polio, 2016) also found significant differences in We seek to answer the following research ques-
terms of syntactic elaboration at the level of the tions:
phrase, clause, sentence, and T-unit between ar-
gumentative and narrative writing, but found no RQ1. How does syntactic elaboration develop in
differences for syntactic subordination. Moreover, L2 French and L2 English?
Ferrari (2012) remarks that effective communica- RQ1a. Which aspects of syntactic
elaboration surface throughout
tion in some tasks can be associated with simpler
development?
language, while Lambert and Kormos (2014) sim- RQ1b. To what extent is the role of
ilarly emphasize that successful communication syntactic elaboration different
may be a matter of “practiced mastery in efficient or similar in L2 French and L2
and effective message formation” (p. 612), rather English?
than the use of increasingly complex language.
Finally, although the study of syntactic com- RQ2. How does syntactic diversity develop in L2
plexity development has seen considerable French and L2 English?
advances with regard to the measurement of RQ2a. Which aspects of syntactic
diversity surface throughout
complexity, with detailed, multi-dimensional in-
development?
vestigations of complexity (e.g., Biber et al., 2011; RQ2b. To what extent is the role of
Bulté & Housen, 2014; Yoon & Polio, 2016) and syntactic diversity different or
the increased use of automatic tools in writing similar in L2 French and L2
research especially (e.g., Lu’s [2010] Syntactic English?
Complexity Analyzer and Graesser et al.’s [2004]
Coh-Metrix), this is not yet the case for all the
aforementioned studies. Thus, the observed METHODOLOGY
increase, decrease, or leveling off of syntactic Participants and Data
complexity in previous studies sometimes only
applies to the degree of subordination, but does The participants in this study are 581 native
not consider the diversity of structures or possible speakers of Dutch, aged 12 to 18, learning French
changes at the phrasal level. We therefore argue and English as part of the foreign language learn-
that, if the effects of cross-linguistic differences on ing curriculum in secondary school in Flanders,
complexity in production are to be studied more Belgium. The data were gathered in the context
extensively, explicitly comparative studies are of a larger research project at the Vrije Univer-
necessary that, to the extent possible, neutralize siteit Brussels that seeks to describe the outcomes
or account for these various factors. It is therefore of foreign language education in Flanders. The
imperative that the relationship between syntactic participants were mostly L2 learners at the onset
complexity and linguistic development and pro- (12–13 years old, n = 146 for English, n = 189
ficiency is studied in comparable contexts across for French) and at the end (17–18 years old, n =
languages, using a sufficiently wide range of 131 for English, n = 87 for French) of foreign lan-
measures that captures linguistic complexity in its guage education in secondary school. They told
multidimensionality. an oral narrative in both French and English.
An additional group of 14–15-year-old learners of
AIMS AND RESEARCH QUESTIONS French (n = 28) was added to compensate for the
higher proficiency level of a considerable part of
This study aims to compare the development the younger English group, which had initially re-
of syntactic complexity in oral productions of L2 sulted in a greater range of linguistic proficiency
English and L2 French over a considerable part of levels in English but not French. Native speaker
the L2 developmental spectrum, from beginning benchmark groups (9–11-year-old speakers of En-
to advanced learners. It uses a toolset that elab- glish from the UK [n = 26] and 12–13-year-
orates on commonly used measures representing old speakers of French from Brussels, Belgium
[n = 32]) were also included. The benchmark sensitive in one language than in the other. In
groups were chosen to correspond to the younger the case of the percentage of correct clauses, for
learners. We acknowledge that the age difference example, the overall lower scores in French can
between the older learners and the native speak- partly be explained by frequent errors on gram-
ers, as well as the age gap between the native matical gender (e.g., *le grenouille [frog-*masc]
speakers in both languages, may hamper the com- instead of la grenouille [frog-fem]).
parison between learners and native speakers; we While we recognize that accuracy and profi-
will return to these issues in the discussion. ciency are not isomorphic, there are nonetheless
The study participants were asked to tell the strong links between both constructs (cf. Bartning
‘frog story’ (Mayer, 1969)—a wordless picture & Schlyter, 2004; Iwashita et al., 2008; Verspoor
story frequently used in language acquisition et al., 2012; Wolfe–Quintero et al., 1998) and we
research—in individual interviews with (near-) believe that the use of accuracy measures resulted
native speakers of the target languages. Partici- in groupings more representative of overall lin-
pants looked at the pictures beforehand and kept guistic proficiency.
them throughout the interview though they were For the final corpus, we then randomly selected
concealed from the interviewers. With the excep- 25 participants in each L2 and L1 group, in or-
tion of a small number of students, most partici- der to arrive at homogeneous group sizes for the
pants completed the task in French and English statistical analyses (see Table 1 for an overview).
on separate days. The order of performance was In the resulting cross-sectional corpus, 33 texts
counterbalanced for the 12–13-year-old learners. per language, 12 of which were categorized at dif-
While the counterbalancing of the order of per- ferent proficiency levels in both languages, were
formance was not an explicit consideration when produced by the same learners in English and
gathering the data for the 17–18-year-old learners, French.
de facto, the order of performance turned out to The narratives were prepared for the syntactic
be variable. complexity analyses in CHAT format (MacWhin-
The learner narratives were grouped into four ney, 2000). The raw transcriptions were adapted
proficiency levels per target language. We define to exclude repetitions, hesitations, and false starts
linguistic proficiency somewhat loosely here, as (Example 1a). Clauses with missing obligatory
a person’s knowledge of the linguistic structures constituents or that contained a main verb not
(patterns, rules, etc.) of the target language and in the target language were excluded as well
the ability to employ these in language use. For (Example 1b). Finally, since the monologic char-
the purposes of this study, this broad construct acter of the narratives was sometimes variable,
is operationalized in terms of (a) age as a proxy, with some participants more frequently address-
which reflects hours of instruction received, and ing the interviewers, all dialogic exchanges were
(b) the participants’ level of linguistic accuracy, removed from the analyses (Example 1c). Many
reflecting the distance to target language norms. of these dialogic fragments contained structurally
Linguistic accuracy was represented by three mea- different elliptic clauses, which resulted in asym-
sures, representing the overall percentage of cor- metries between the strictly monologic narratives
rect clauses per text, the percentage of correct and more interactive exchanges.
verb forms, and the percentage of accurately used
prepositions (see Buysse & De Clercq, 2014, for EXAMPLE 1
more details on the choice of these measures). (1a) [they run th] they fly after the dog.
The additional use of accuracy measures was (1b) and the dog and the bees komen [Dutch:
deemed necessary since groups based solely on ‘come’] the dog.
age would have been particularly heterogeneous. (1c) - why is he allowed to take one frog home?
Thus, the scores for the accuracy measures were - because the grand frog is used to live with
used to divide each age group (12–13 and 17– him.
18 for English, 12–15 and 17–18 for French) into
two further groups, resulting in four general pro- Syntactic Complexity Measures
ficiency levels. Group membership was decided
on the basis of the average accuracy scores within An array of judiciously selected measures rep-
each language: Participants achieving higher than resenting syntactic complexity through syntactic
average scores on at least two out of three mea- elaboration and syntactic diversity were applied
sures were placed into a higher proficiency level. to the transcriptions. The measures of syntactic
Language-specific averages were chosen as cut- elaboration follow the suggestions of Norris &
off points because some measures were more Ortega (2009) and Pallotti (2015) and include
TABLE 1
Overview of Groups and Accuracy Scores
Language English French
Proficiency group 1 2 3 4 1 2 3 4
N 25 25 25 25 25 25 25 25
Age 12–13 12–13 17–18 17–18 12–15 12–15 17–18 17–18
% error-free clauses Mean 22% 61% 75% 88% 8% 34% 56% 74%
SD 12% 14% 8% 5% 7% 20% 10% 13%
% correct prepositions Mean 55% 80% 85% 90% 33% 68% 84% 94%
SD 16% 12% 6% 7% 23% 18% 9% 5%
% correct verb forms Mean 28% 61% 85% 98% 35% 75% 92% 97%
SD 28% 31% 17% 3% 17% 17% 6% 3%
length measures calculated at three syntactic lev- in-text standard deviation may distinguish texts
els, that is, the noun phrase (LenNP), the clause that simple length measures would consider iden-
(LenC), and the AS-unit (LenAS), as well as a tical (Examples 3a and 3b).
clause-linking measure representing the number
of clauses per AS-unit (C/AS). The AS-unit is a EXAMPLE 3
multiclausal unit encompassing an independent (3a) Average clause length: 6 words; Standard
clause and its dependent clauses and is thus simi- deviation of clause length: 0 words
lar to the frequently studied T-unit (Hunt, 1965) and he climbed on a rock.
in writing research. In contrast to the T-unit, the and held himself on a tree.
AS-unit also includes subclausal units, which are the tree was n(o)t a tree.
not necessarily organized around a verb (Foster (3b) Average clause length: 6 words; Standard
et al., 2000). Our working definition for clauses deviation of clause length: 2 words
identified these as syntactic units headed by a the frog is n(o)t in the jar anymore.
verb, whether finite or nonfinite. The NP is here and he looks sad.
he looks everywhere in the room.
defined as a syntactic unit headed by a nominal el-
ement, and was operationalized by including only
Obviously, both traditional length measures
NPs with a common head noun which were not
and standard deviation are holistic, global metrics
embedded in another NP. Only words of the tar-
of structural surface complexity that cannot dis-
get language or idiosyncratic words that had a
tinguish the underlying structural differences of
phonological or morphological structure corre-
utterances in (4a) and (4b).
sponding to the target language were included in
the counts. EXAMPLE 4
(4a) and he does n(o)t find him. (6 words)
EXAMPLE 2 (4b) the guy find the frog not. (6 words)
(2a) The boy roeping [Dutch: CALL+ING]
again. (4 words) However, the use of these measures is based
on their status as proxies of syntactic complex-
Since few L2 studies have included syntactic ity and, in the case of length measures specifi-
diversity measures, the ones used in this study cally, on their high correlation with other types
are exploratory. We used three types of measures. of measures such as node-based measures (Szm-
First, as a proxy of syntactic diversity analogous to recsanyi, 2004). The advantage of these diversity
the length measures, the within-text standard de- measures is that they are easy to calculate and in-
viation of the syntactic length is used.2 The un- terpret, are analogous to length-based measures,
derlying assumption for this measure is that, if and can be applied to several syntactic levels. As
length can be used as a proxy of underlying struc- such, they will be applied to the same units as
tural differences, as is the case for most frequently the length measures, that is, noun phrase diversity
used syntactic elaboration measures, more syntac- (DivNP), clause diversity (DivC), and AS-unit di-
tically diverse texts will include more units of vary- versity (DivAS).
ing length and have higher standard deviations Second, we used a measure representing
for length measures. Thus, measures based on the percentage of various clause types in each
text. We distinguished three general types of EXAMPLE 7
clauses in the corpus: matrix clauses (%mat),
[sub.adv.fin = S.ADV-F] when the boy wakes up.
subordinate clauses (%sub), and coordinated
[mat = M] he sees.
clauses (%coor). Following Cosme (2007) and
[sub.com.fin. = S.C-F] that the frog is n(o)t in his
Biber et al. (2011), we further distinguished
jar anymore.
two formal subtypes of subordinate clauses (fi-
Clause-linking pattern = S.ADV-F+M+S.C-F
nite/nonfinite) and three functional subtypes
(complement/adverbial/relative). Note that in An automatic tool (Brezina & Pallotti, 20153 )
our analysis the category of coordinated clauses then segmented each text into random samples
only covers coordinated clauses with ellipsis of five AS-units, for which the number of different
(of the verb and/or the subject). This decision syntactic codes within and across each sample was
was made to account for the frequent use of counted. After 100 repetitions of this process, the
coordinating conjunctions such as ‘and’ as filler tool averaged the number of unique codes within
items in oral data rather than as true connectives each sample and across each sample and calcu-
(Beaman, 1984; Kerr–Barnes, 1998) (see Exam- lated the SDI using the following formula:
ple 5).
uniq ue c ode s within s ample s

EXAMPLE 5 uniq ue c ode s acr os s s ample s
+ − 1
[mat] he look after a tree. 2
[mat] and there is his frog with a girl frog. Since the SDI requires the presence of at least
[mat] and they have a lot of children. two 5-code samples in order to calculate across-
[mat] and they are playing with them. segment diversity, six Level 1 French texts that
[mat] and then see he leaves. contained fewer than 10 AS-units were excluded
[mat] and they are watching at him. from the SDI analysis.
The high frequency of such connectives in our

Statistical Analyses
data thus led us to consider these utterances as
separate matrix clauses. Utterances where part of The statistical analyses were carried out using
the second clause was elliptic were nonetheless SPSS (IBM, 2015). The normality of data dis-
analyzed as coordinated clauses, since these could tribution was assessed by visually inspecting Q–
not be considered syntactically independent but Q plots and performing a Kolmogorov–Smirnov
nonetheless did not correspond to subordinate test. These tests revealed that the distribution
clauses. These include coordinated clauses with of the scores in the different participant groups
ellipsis of the subject (6a) or of one or more other largely followed a normal distribution. For mea-
constituents (6b). sures that adhered to this pattern (that is, those
for which the scores of at least four out of five
EXAMPLE 6 groups followed a normal distribution), a one-
way analysis of variance (ANOVA) was carried
(6a) [mat] and the boy does his hands for his
head. out. Since variances in most groups were unequal,
[coor] and hide him. as assessed by Levene’s test of homogeneity of
(6b) [mat] the boy fells into a big river. variances, Games–Howell post-hoc tests were per-
[coor] and his dog too. formed. Post-hoc test results will be presented
as homogeneous subsets, separating nonsignifi-
Third, we developed a new measure of syntactic cantly different groups with commas and statisti-
diversity conceptually based on Pallotti’s (2015) cally significantly different groups (p < .05) with
measure of morphological diversity. This mea- angled brackets. For example, ‘1<2<3,4<4,NS’
sure, called the Syntactic Diversity Index (SDI), would mean that the differences between Levels
represents the diversity of the combinatorial pat- 1, 2 and 3 are statistically significant. The differ-
terns of clause types found within AS-units by ana- ence between Levels 1 and 2, on the one hand,
lyzing each AS-unit as a linear string of the clause and Level 4, on the other hand, is also statistically
types described previously. The procedure for cal- significant, while the difference between Levels 3
culating the measure first labeled each AS-unit and 4 is not. Similarly, the native speakers attain
with a code representing its underlying clause- significantly higher scores than Levels 1, 2, and
linking structure. 3, but the difference between the L1 group and
Level 4 is not statistically significant. Effect sizes only between Levels 1 and 2 and between Levels
for parametric ANOVAs will be reported in the 3 and 4 (p < .05). Unlike in the French group,
form of partial eta-squared values (η2p ). no significant differences were observed between
When the data did not follow a normal distri- the native speakers of English and Levels 3 and
bution, a Kruskal–Wallis test for nonparametric 4 for the two AS-unit measures, nor between the
data was performed (indicated as ‘np’ in the re- native speakers and any of the learner groups for
sult tables). The use of nonparametric tests was LenNP and LenC. Effect sizes are generally lower
warranted by nonnormal distributions for some in English than in French and are again higher for
of the clause type measures, more specifically, for the AS-unit measures (η2p = .393 for LenAS and
the percentage of coordinated clauses with ellip- .350 for C/AS) than for NP length (η2p = .054) or
sis, adverbial clauses, relative clauses, and nonfi- clause length (η2p = .025).
nite clauses. While a direct comparison between both lan-
guages is difficult since equivalence of proficiency
levels is not guaranteed, a few tendencies surface
RESULTS nonetheless. First, the English scores tend to be
Syntactic Elaboration higher than the French scores at Level 1 and
Level 4. This trend is the most pronounced for
Table 2 presents the results for the syntactic the AS-unit measures LenAS and C/AS, but does
elaboration measures in both languages. A one- not extend to the native speaker groups, in which
way ANOVA revealed statistically significant differ- the French participants consistently score higher
ences for mean length of NP (LenNP, F(4,120) = than their English peers, with the exception of
9.09, p < .05), mean length of AS-unit (LenAS, LenC.
F(4,120) = 34.57, p < .05), and the number of In sum, we found a stronger increase of syntac-
clauses per AS-unit (C/AS, F(4,120) = 39.05, p < tic elaboration in the French learner data than
.05), but not for mean length of clause (LenC, in the English learner data. In both languages,
F(4,120) = 1.292, p = .277). this increase was mainly restricted to elabora-
Pair-wise comparisons between groups revealed tion at the AS-unit level. Additionally, more En-
only small and statistically nonsignificant differ- glish learners attained scores similar to the native
ences between the French learner groups for benchmark group than the French learners did.
LenNP and LenC. LenAS in the L2 French groups
increased progressively, especially from Levels 1 Syntactic Diversity
to 3, where the difference between Levels 1 and
2 on the one hand and 3 and 4 on the other The results for syntactic diversity, represented
hand was statistically significant, as indicated by a by the standard deviation (DivNP, DivC, and
Games–Howell test (p < .05). Similarly, a progres- DivAS) and the number of different clause-
sive increase was found between all proficiency linking patterns (SDI), indicate stronger in-
levels for C/AS, with statistically significant differ- creases at the AS-unit level than at the (noun)
ences observed between Levels 1, 2, and 3 (p < phrasal or clausal level in both languages
.05). Compared to the native speakers, the learn- (Table 3).
ers of French consistently produced shorter noun In French, a one-way ANOVA revealed sta-
phrases and AS-units, and also used fewer clauses tistically significant differences for NP diversity
per AS-unit; a Games–Howell post-hoc test re- (DivNP, F(4,120) = 15.63, p < 0.05), clause diver-
vealed the difference between native speakers and sity (DivC, F(4,120) = 6.55, p < 0.05), AS-unit di-
learners at all levels to be statistically significant (p versity (DivAS, F(4,120) = 31.61, p < 0.05), and
< .05). The stronger differences found for LenAS the Syntactic Diversity Index (SDI, F(4,114) =
and C/AS are reflected in the higher effect sizes, 42.09, p < 0.05). DivNP tended to increase with
of respectively .535 and .566, compared to .233 for proficiency, although the differences are not sta-
LenNP and .009 for LenC. tistically significant in the learner groups. At the
As for the English learners, a one-way ANOVA clausal level, a Games–Howell post-hoc test re-
revealed statistically significant differences for vealed a similar increase to be statistically signif-
LenAS (F(4,120) = 19.44, p < .05) and C/AS icant between Level 1 and Levels 3 and 4 (p <
(F(4,120) = 16.162, p < .05), but not for LenNP .05). Finally, AS-unit diversity measures (DivAS
(F(4,120) = 1.72, p = .15) or LenC (F(4,120) = and SDI) showed strong and statistically signif-
.764, p = .55). At the AS-unit level, LenAS and icant increases between Levels 1 and 2 and 3
C/AS again increased progressively, although the and 4 in French (p < .05). Note that, while not
difference was in both cases statistically significant statistically significant, SDI scores in French also
TABLE 2
Summary of Syntactic Elaboration Measures
Measure 1 2 3 4 NS Post-hoc η2p
LenNP French N 25 25 25 25 25 1,2,3,4<NS .233

Mean 2.49 2.47 2.57 2.58 2.92
SD .43 .27 .21 .20 .34
English N 25 25 25 25 25 1,2,3,4,NS .054
Mean 2.56 2.39 2.59 2.51 2.49
SD .35 .26 .33 .25 .26
LenC French N 25 25 25 25 25 1,2,3,4,NS .009
Mean 5.38 5.41 5.72 5.54 5.62
SD .72 .61 .68 .43 .59
English N 25 25 25 25 25 1,2,3,4,NS .025
Mean 5.61 5.66 5.70 5.73 5.85
SD .69 .55 .38 .39 .52
LenAS French N 25 25 25 25 25 1,2<3,4<NS .535
Mean 5.79 6.36 7.80 7.92 9.14
SD .92 .92 1.56 .97 1.18
English N 25 25 25 25 25 1,2<3,4,NS .393
Mean 6.55 7.22 8.45 8.77 8.81
SD 1.06 1.14 1.38 1.02 1.14
C/AS French N 25 25 25 25 25 1<2<3,4<NS .566
Mean 1.08 1.18 1.36 1.43 1.61
SD .11 .13 .20 .18 .20
English N 25 25 25 25 25 1,2<3,4,NS .350
Mean 1.18 1.31 1.49 1.53 1.51
SD .18 .18 .21 .16 .22
Note. LenNP = Length of noun phrase; LenC = Length of clause; LenAS = Length of AS-unit; C/AS = Number of
clauses per AS-unit; NS = Native speaker.
differed considerably between Level 1 (.68) and At the NP level (DivNP), a Games–Howell post-
Level 2 (1.24). As was the case for the syntac- hoc test found no differences in syntactic diversity
tic elaboration measures, the difference between in the English learner groups, although learners
learners and native speakers was always statisti- at Level 1 tended to use more diverse NPs (1.37)
cally significant (p < .05), except for DivC, for than at Level 2 (1.18). As in French, we found
which only the learners at Levels 1 and 2 did not higher increases of syntactic diversity than of elab-
attain scores comparable to those of the native oration at the clausal level even though these are
speakers (p < .05). As was the case for syntactic only statistically significant between Level 1 on the
elaboration, there were no statistically signifi- one hand and Levels 3 and 4 on the other hand
cant differences in NP diversity in the learner (p < .05). Finally, the two measures show that
groups, with group means varying between 1.03 the increase in diversity found for both measures
and 1.31, although the gap between the learn- is more pronounced when measured in terms of
ers and the native speakers was considerable clause-linking patterns (SDI), for which we found
and statistically significant (p < .05). Interest- statistically significant differences between Levels
ingly, while no systematic increase was observed 1 and 2 and Levels 3 and 4, than when measured
for syntactic elaboration at the clausal level in terms of word length (DivAS), for which only
(LenC), clausal diversity increased. We find the the differences between Levels 1 and 2 and Lev-
highest effect sizes for the AS-unit measures els 3 and 4 were statistically significant. While no
(η2p = .513 for DivAS and .596 for SDI). statistically significant differences were observed
The scores in English indicate more moderate between the English native speakers and the more
increases in syntactic diversity than in French. A advanced learners of Levels 3 and 4, these learners
one-way ANOVA revealed statistically significant actually tended to use more diverse syntax than
differences for DivC (F(4,120) = 4.83, p < 0.05), the native speakers for all measured units.
DivAS (F(4,120) = 13.92, p < 0.05), and SDI A comparison of the two languages reveals
(F(4,120) = 23.61, p < 0.05), but not for DivNP that effect sizes in English are typically lower
(F(4,120) = .863, p = 0.488) (η2p = .028 – .440) than those found for French
TABLE 3
Summary of Diversity Measures: Simple Indices
DivNP French N 25 25 25 25 25 1,2,3,4<NS .342

Mean 1.03 1.19 1.23 1.31 2.08
SD .48 .50 .38 .42 .75
English N 25 25 25 25 25 1,2,3,4,NS .028
Mean 1.37 1.18 1.42 1.38 1.28
SD .53 .49 .52 .58 .44
DivC French N 25 25 25 25 25 1,2<2,3,4<3,4,NS .179
Mean 2.04 2.13 2.37 2.44 2.51
SD .46 .45 .41 .29 .35
English N 25 25 25 25 25 1,2<2,3,4,NS .139
Mean 2.12 2.30 2.53 2.54 2.44
SD .37 .40 .46 .46 .32
DivAS French N 25 25 25 25 25 1,2<3,4<NS .513
Mean 2.27 2.37 3.36 3.60 4.75
SD .75 .59 1.15 .86 1.03
English N 25 25 25 25 25 1,2<3,4,NS .317
Mean 2.52 2.81 3.96 4.08 3.62
SD .81 .77 1.13 1.16 .70
SDI French N 19 25 25 25 25 1,2<3,4<NS .596
Mean .68 1.24 2.38 2.65 3.43
SD .67 .86 .93 .85 .67
English N 25 25 25 25 25 1<2<3,4,NS .44
Mean 1.21 1.95 2.92 3.28 3.15
SD .95 .94 .93 .76 1.00
Note. DivNP = Noun phrase diversity; DivC = Clause diversity; DivAS = AS-unit diversity; SDI = Syntactic diversity
index; NS = Native speaker.
(η2p = .179 – .596). Group averages for syntac- decrease in the percentage of matrix clauses, sta-
tic diversity are consistently higher, more so than tistically significant from Levels 1 to 3 (p < .05), is
for syntactic elaboration, in the English learner offset by a similar increase in subordinate clauses,
data than in the French learner data, indicating equally statistically significant from Levels 1 to 3
that the learners start out using more diverse syn- (p < .05). No statistically significant differences
tax in English than in French, and their English were observed between any of the learner groups
maintains this edge at more advanced stages as in both languages for the coordination measure,
well. Once more, the French native speakers re- even if considerable differences were nonetheless
verse this trend by achieving higher scores than found in the group averages. The proportion of
both the English native speakers and the more ad- coordinated clauses with ellipsis was generally
vanced English learners. higher in English than in French, and decreased
As for the distribution measure (Table 4), a considerably in French from Level 3 onward.
one-way ANOVA reveals statistically significant dif- The clause-type distribution in the French
ferences in French and English for the percentage native speaker data converges with the trend
of matrix clauses (F(4,120) = 55.28, p < .05 for observed in the learner data and indicates that
French; F(4,120) = 21.65, p < .05 for English) and French native speakers rely even more on sub-
the percentage of subordinate clauses (F(4,120) ordinate clauses than the advanced learners.
= 55.01, p < .05 for French; F(4,120) = 28.11, p In English, the native speaker scores typically
< .05 for English). Additionally, a Kruskal–Wallis fall between those of Levels 3 and 4 for the use
test found a statistically significant difference for of matrix or subordinate clauses. Interestingly,
the percentage of coordinate clauses in English the learners seem to underuse coordinated
(χ 2 (4) = 16.76, p < 0.05), but not in French clauses with ellipsis in French and English when
(χ 2 (4) = 3.56, p = .464). compared to the native speakers.
A closer look at the scores reveals a strong in- Turning our attention to the different cat-
verse tendency in both languages. A progressive egories of subordinate clauses, a one-way
TABLE 4
Summary of Diversity Measures: Distribution Measures
% Mat French N 25 25 25 25 25 1>2>3,4>NS .648

Mean 93.59% 85.58% 74.76% 70.75% 61.97%
SD 7.79% 9.33% 9.13% 8.67% 6.58%
English N 25 25 25 25 25 1>2>3,4,NS .419
Mean 86.30% 77.62% 68.25% 65.99% 67.37%
SD 11.27% 10.28% 8.98% 6.41% 9.00%
%Coor French N 25 25 25 25 25 1,2,3,4,NS np
Mean 3.05% 3.17% 1.70% 1.63% 3.91%
SD 4.04% 4.69% 2.38% 1.82% 5.62%
English N 25 25 25 25 25 1,2,3,4<2,NS np
Mean 4.72% 6.29% 3.52% 4.10% 8.46%
SD 5.10% 7.51% 2.96% 5.66% 4.95%
%Sub French N 25 25 25 25 25 1<2<3,4<4,NS .647
Mean 3.36% 11.25% 23.54% 27.62% 34.12%
SD 6.43% 10.02% 9.25% 8.51% 7.35%
English N 25 25 25 25 25 1<2<3,4,NS .484
Mean 8.98% 16.10% 28.23% 29.91% 24.18%
SD 9.43% 8.74% 8.88% 6.70% 7.40%
Note. %Mat = Percentage of matrix clauses; %Coor = Percentage of coordinated clauses; %Sub = Percentage of
subordinate clauses; NS = Native speaker.
ANOVA of the various formal and functional proportion of subordinate clauses in the French
subtypes (Table 5) reveals statistically significant native speaker data was mostly due to the high
differences in both languages for the percentage number of relative clauses in the L1 data, while
of complement clauses (F(4,120) = 25.92, p < .05, the more advanced learners scored similarly to
for French, F(4,120) = 13.26, p < .05, for English) the native speakers with regard to complement
and finite subordinate clauses (F(4,120) = 41.30, clause use.
p < .05, for French, F(4,120) = 25.04, p < .05, In the English data, the various subcategories
for English), as well as for the percentage of non- also increased overall throughout development,
finite subordinate clauses in English (F(4,120) especially from Levels 1 to 3. Compared to the
= 12.44, p < .05). An additional Kruskal–Wallis French data, however, all clause types already
test revealed statistically significant differences in appear at Level 1, even if complement clauses
both languages for the percentage of adverbial remain the dominant subordinate clause type
clauses (χ 2 (4) = 70.98, p < .05 for French; χ 2 (4) throughout development. The data also revealed
= 55.30, p < .05 for English) and relative clauses that the learners at Level 4 frequently surpassed
(χ 2 (4) = 62.94, p < .05 for French; χ 2 (4) = 32.82, the English native speakers in their use of most
p < .05 for English), as well as a statistically signif- subordinate clause types and that this difference
icant difference for the percentage of nonfinite was statistically significant for adverbial and finite
clauses in French (χ 2 (4) = 48.48, p < .05). subordinate clauses (p < .05).
These data reveal that, in French, the observed In sum, the analyses of syntactic diversity
increase in subordinate clauses can largely be suggest that this dimension develops similarly
explained by an increased use of complement to syntactic elaboration in that we found an
clauses and finite clauses. If all subordinate clause overall diversification with increased proficiency
types tended to increase with proficiency in the particularly at the supraclausal level. Moderate
French learner data, relative clauses remained a increases were observed as well for clausal diver-
minor and, compared to the native speakers, un- sity in both languages. The data also foreground
derused category. Note that the French learners at coordination as a more prominent category in
Level 1 only relied on complement clauses, both English than in French, although in both lan-
finite and nonfinite, but not adverbials or rela- guages its role throughout development seems
tives. The data equally revealed that the higher limited.
TABLE 5
Summary of Diversity Measures: Functional and Formal Subtypes
%Com French N 25 25 25 25 25 1,2<3,4,NS .464

Mean 3.12% 8.04% 14.90% 18.34% 18.59%
SD 6.44% 8.05% 6.05% 5.64% 6.92%
English N 25 25 25 25 25 1<2,NS<3,4,NS .307
Mean 6.48% 11.49% 16.39% 15.99% 14.91%
SD 6.13% 7.10% 5.66% 3.96% 5.13%
%Adv French N 25 25 25 25 25 1,2<3,4,NS np
Mean .00% 2.04% 6.18% 6.23% 8.17%
SD .00% 3.98% 5.04% 3.61% 4.30%
English N 25 25 25 25 25 1,2<2,NS<3,NS<3,4 np
Mean 1.75% 3.35% 8.03% 10.80% 5.92%
SD 3.62% 3.90% 3.92% 4.20% 4.52%
%Rel French N 25 25 25 25 25 1,2<2,3,4<NS np
Mean .00% 1.17% 2.46% 3.05% 7.35%
SD .00% 1.93% 2.95% 3.14% 4.56%
English N 25 25 25 25 25 1,2<2,NS<4,NS<3,4,NS np
Mean .75% 1.26% 3.81% 3.13% 3.35%
SD 2.60% 2.07% 3.28% 2.95% 4.28%
%Fin French N 25 25 25 25 25 1<2<3,4<4,NS .579
Mean .36% 5.07% 13.76% 14.83% 18.62%
SD 1.82% 5.93% 8.28% 6.45% 4.84%
English N 25 25 25 25 25 1,2<3,NS<3,4 .455
Mean 3.96% 4.98% 15.41% 16.23% 11.74%
SD 6.50% 4.92% 6.46% 4.45% 6.05%
%N-Fin French N 25 25 25 25 25 1,2<2,3<3,4,NS np
Mean 2.99% 6.18% 9.78% 12.79% 15.50%
SD 6.35% 6.67% 4.43% 4.95% 6.88%
English N 25 25 25 25 25 1<2,3,4,NS .293
Mean 5.02% 11.12% 12.82% 13.68% 12.44%
SD 5.42% 6.38% 4.04% 3.96% 4.39%
Note. %Com = Percentage of complement clauses; %Adv = Percentage of adverbial clauses; %Rel = Percentage of
relative clauses; %Fin = Percentage of finite subordinate clauses; %N-Fin = Percentage of nonfinite subordinate
clauses; NS = Native speaker.
DISCUSSION EXAMPLE 8
This section discusses the results in relation to (8a) le garçon et le chien cherche une
our two main research questions by focussing first grenouille. (L2: level 1)
on syntactic elaboration, then on syntactic diver- the boy and the frog search for a frog
(8b) and the boy see an uil [Dutch: ‘owl’]. (L2:
sity. Throughout our discussion we will consider
level 1)
the impact of our methodological choices.
Our first research question pertained to the
This does not imply that all clauses at Level 1
development of syntactic elaboration and poten-
are organized around finite verbs (9a), although
tial cross-linguistic differences. The data revealed
some complex structures, such as negatives with
that syntactic complexification mainly occurred
do-support in English (9b), already occur at Level
at the level of the AS-unit, where we found pro-
1 alongside simpler constructions (9c).
gressive complexification, especially from Levels
1 to 3, whereas no consistent differences were
observed at the clausal or NP level. These tenden- EXAMPLE 9
cies were observed for both languages and suggest (9a) mais ce soir le grenouille partir. (L2:
that learners at Level 1 are already past the basic level 1)
variety stage (cf. Klein & Perdue, 1997), and have but that evening the frog leave-INF
already mastered basic NP and clause structures (9b) but they do n(o)t find it. (L2: level 1)
(see Examples 8a and 8b). (9c) but they find him not. (L2: level 1)
While no statistically significant differences even more syntactically complex texts than in
were found between the learner groups for NP native English. The more advanced English
length in both languages, the French benchmark learners (and sometimes even beginning learn-
group scored significantly higher than the learner ers) achieved scores similar to the English native
groups, suggesting that the learners may have speakers’, which in turn tended to be lower than
mastered the basic NP structure but that devel- the scores of the French native speaker group.
opment at a more advanced level might involve Thus, if the English learners’ narratives tended
further NP complexification. This finding is in to be more syntactically complex than the French
line with previous findings that suggest a prefer- learners’ narratives, the inverse trend holds for
ence for phrasal elaboration in written French the native benchmark groups.
(cf. Cosme, 2007): It seems to hold for oral L2 As for our second research question, which
French as well. This higher complexity of the focuses on syntactic complexity as diversity, our
native speakers’ NPs is likely due to their higher simple diversity indices indicated a development
use of post-modification through relative clauses which parallels that of the syntactic elaboration
rather than through modification by preposi- measures. The strongest increases were observed
tional phrases and adjectives, which appear fre- for the two AS-unit measures (DivAS and SDI).
quently in the learner data as well: We also found a strong and statistically significant
difference in NP diversity between native speak-
EXAMPLE 10 ers and learners in French. In other words, na-
tive speakers seem to alternate more frequently
le petit garçon regarde en tout le maison. (L2 : between longer nominal structures (with post-
level 1) modifiers) and short NPs (determiner + noun):
the little boy looks in the whole house
il crie le nom de le grenouille. (L2: level 1)
EXAMPLE 11
he shouts the name of the frog
[mat] c’est {NP1: l’histoire d’un garcon
Turning to the supraclausal level, one may ex- it is the story of a boy
pect a considerable degree of collinearity between [fin-rel.] qui a reçu une grenouille}.
the length of AS-unit and the number of subordi- who got a frog.
nate clauses per AS-unit, in the sense that the use [mat] et alors avec {NP2: son chien} il la regarde.
of subordinate clauses will also affect the length (L1)
of the overarching AS-unit. Importantly, however, and he then looks at it with his dog.
AS-unit length may in theory also be affected by
the use of adjuncts at the clausal level or post- Additionally, English learners showed a strong
modification at the phrasal level. Yet, the lack of increase in AS-unit diversity (SDI) at early levels,
systematic differences between learner groups at even if intra-group variation for this measure
the clausal and NP level, on the one hand, and was high in all groups. Thus, if the increase in
the simultaneous increase of AS-unit length and AS-unit length was minor and not statistically
the number of clauses per AS-unit, on the other significant between Levels 1 and 2, the English
hand, indicate that the progressive complexifica- learners nonetheless combined more different
tion observed at the supraclausal level is mostly clause types.
caused by an increased reliance on subordinate Interestingly, minor increases were found for
clauses (from 1.08 to 1.43 clauses per AS-unit in clausal diversity in both languages (statistically
French and from 1.18 to 1.53 clauses per AS-unit significant between Level 1 and Levels 3 and
in English) rather than by changes at lower syn- 4), contrary to our results for clause length.
tactic levels. Thus, even if same-level measures tended to
The two languages mainly differ in the ten- indicate converging trends (e.g., for LenNP and
dency of learners to attain higher scores in En- DivNP), diversity and elaboration measures can
glish than in French, which is most pronounced still display different tendencies. One possible
for AS-unit length and clauses/AS-unit and the explanation for the increased diversity at the
least clear for NP length. The lower scores in clausal level is the increased use of (finite and
French may partly be explained by the compara- nonfinite) complement clauses, which results
tively lower overall proficiency level of the French in AS-units consisting of shorter main clauses
learner groups, rather than by typological differ- and longer complement clauses (Example 12a).
ences, especially since the benchmark data indi- The clause-length measure does not necessarily
cate that native speaker productions tend toward distinguish between such utterances as those
in Example 12b, while AS-unit and diversity than most learners, but the native speakers’ use of
measures will pick up these differences. the different subordinate clause types is typically
situated between the scores of Levels 2 and 3, with
EXAMPLE 12 the difference between the native speakers and
the higher scores at Level 4 achieving statistical
(12a) LenC: 3.5 DivC: 2.12 LenAS:7 significance. Thus, while the syntactic elaboration
[mat] il pense measures indicate comparable levels of syntactic
he thinks
complexity in the more advanced learners’ pro-
[sub. comp. fin.] que c’est des branches.
(L2 : level 3)
ductions and the narratives from the L1 English
that they are branches. group, there is a strong tendency toward higher
(12b) LenC: 3.5 DivC: .71 LenAS:3.5 diversity in the learner group.
[mat] et il cherchE. A number of explanations could account for
and he search-INF. the lower diversity in the English benchmark
[mat] il tombe en arbre. (L2: level 1) data. It is possible that the English benchmark re-
he falls in tree. flects the fact that the narratives did not require
highly diverse language in order to be told effec-
Such examples may thus shed new light on so- tively. The increasing complexity in the learner
called ceiling effects in syntactic complexity at the groups would then be the result of a focus of
clausal level, since in certain cases clause length the language learning process on acquiring more
measures may conceal changes at the clausal level structures and using more complex language (cf.
that are captured by other measures. Lambert & Kormos, 2014), rather than an indi-
Turning to the distribution measures, these re- cation of increased effectiveness in performing
veal an important increase in subordinate clauses the task. In other cases, learners might rely on
in both languages, mainly of complement clauses, more syntactically complex paraphrases to over-
although all categories occur in the learner data come lexical problems, as can be seen in Exam-
from Level 2 onward. The measures also revealed ples 13a and 13b.
a number of differences between the two lan-
guages. The English learners relied more on
subordination than the French learners and this EXAMPLE 13
difference was most apparent for adverbial clauses (13a) [mat] There is a big animal with horns.
and nonfinite clauses. Moreover, while the pro- (L2: level 2)
portion of adverbial and relative clauses was low (13b) [mat] But there is a deer. (L2: level 4)
at Level 1 in English, they did not occur in any of
the French Level 1 narratives. The English learn- Alternatively, the age difference between the
ers also relied more extensively on coordinated native speaker benchmark and the advanced
clauses with ellipsis than the French learners, English learners might be such that the level
although high intra-group variation equally sug- of complexity in the L1 English group is not
gests considerable individual differences. Thus, representative of that of 17–18-year-old native
even within our restricted definition of clausal co- speakers of English.
ordination, our data seem to cautiously confirm In sum, the results of this study largely corrob-
previous findings (cf. Cosme, 2007) that both orate previous findings in that the French and
learners and native speakers rely more extensively English L2 learners both progressively use more
on coordination in English than in French. subordinate clauses and that this development is
The results from the benchmark groups are also accompanied by a diversification of formal
equally interesting. While the French learner and functional subordinate clause types, although
data suggest an increased use of subordinate few changes were observed at the level of the
clauses with proficiency, there nonetheless re- noun phrase and the clause in terms of syntactic
mains a considerable difference between the elaboration. The study revealed as well that learn-
most advanced learners (27.62%) and the native ers start using most clause types at relatively early
speakers (34.12%). This difference can mostly levels, although, especially in French, the main
be explained by the underuse of relative clauses differences between learners and native speakers
in all French learner groups, while the use of reside in the distribution of these categories.
complement clauses at Level 4 (18.34%) largely Since there was a tendency toward increased
coincided with native speaker usage (18.59%). elaboration and diversification at the group
In English, on the other hand, the benchmark level, regardless of the considerable overlap in
group relies considerably more on coordination the individual scores across all groups for most
measures, we were unable to confirm previously tic structures of different orders (e.g., clausal
formulated hypotheses that more advanced learn- embedding and negative structures) remains
ers use more syntactically simple language (cf. challenging and, if based on factors such as dif-
Lambert & Kormos, 2014). The generally lower ficulty or developmental timing, irreconcilable
complexity scores of the English benchmark with the absolute view on complexity advocated
group could be considered insightful in this re- by, for instance, Pallotti (2015).
gard, although it is debatable whether our group
of 9–11-year-old native speakers is representative CONCLUSION
of the ‘expert speakers’ discussed by Lambert
and Kormos (2014). This study aimed to enrich the current litera-
The results of this study have to be interpreted ture on linguistic and syntactic complexity in the
in light of a number of limitations. First, while L2 learning process by adopting a cross-linguistic
the learner groups covered a considerable part perspective. It proposed a systematic approach to
of the developmental spectrum, the study in- syntactic complexity that introduced a new set
cluded few if any absolute beginners, nor did it of measures tapping into complexity as syntactic
cover near-native learners. Hence, it is unclear diversity, alongside more traditional complexity
whether the learners of French in this study will measures that target complexity as syntactic elab-
continue along the path of syntactic elaboration oration, while at the same time tapping into the
and diversification to approach native speaker complexity of units at different levels of syntactic
levels. Moreover, the learners’ increased reliance analysis. While our results generally corroborate
on syntactically complex language could reflect those of previous studies, foregrounding the im-
their own cognitive and academic development portance of clause linking in the syntactic L2 de-
as the age of the participants varies from 12 to 18 velopment of both French and English, they also
years, and is thus not necessarily specific to the L2 pointed to the important role of the gradual di-
learning process. That is, an increase in complex- versification of structures. The study also opened
ity can be expected in the learners’ L1 (Dutch) up new research avenues, as future studies might
as well (Berman & Verhoeven, 2002), just like the extend the cross-linguistic comparison not only to
French and English benchmark groups will likely more typologically distant languages but also to
continue to develop in this respect. true beginner and very advanced stages of L2 de-
Second, care should also be taken when velopment and apply the notion of syntactic diver-
comparing the data across languages. While sity to longitudinal analyses.
we attempted to capture a similar range of
the developmental spectrum, no independent
cross-linguistically valid tool to assess proficiency NOTES
was used and future studies could rely on other
methods to capture the construct of proficiency 1 Length and subordination (or embedding) could
in a more nuanced or inclusive manner by, for arguably also be considered different forms of syntactic
example, not only focusing on productive skills complexity, but for the purpose of this article, we will re-
but also on receptive skills. Additionally, while fer to these as syntactic elaboration, which is contrasted
this study approached syntactic complexity using with syntactic diversification.
2 This measure thus represents the standard deviation
an arguably elaborate set of measures, concerns
of unit length in each individual performance and is not
about concept reductionism and the opacity of
to be confused with the more commonly used within-
complexity measures may still persist. Concerns group standard deviation. For a similar use of this mea-
about the validity of length-based measures, both sure, see Byrnes et al. (2010).
mean length of syntactic unit and the variation 3 While this tool was originally developed for morpho-
thereof, may also remain legitimate since these logical analyses, it equally functions with custom labels,
measures tap holistically into linguistic surface such as our syntactic codes. For more details on the tool,
phenomena and reveal little about what exactly see Brezina & Pallotti (2015, 2017) and Pallotti (2015).
happens linguistically. In this study, these issues
were partly remedied by using simple complexity
indices alongside a more descriptive set of clause- REFERENCES
distribution measures. Other studies (typically in
L1 acquisition; cf. Botel, Dawkins, & Grabowsky, Ågren, M., Granfeldt, J., & Schlyter, S. (2012). The
1973; Scarborough, 1990) have considered syn- growth of complexity and accuracy in L2 French:
tactic complexity in greater detail, although Past observations and recent applications of de-
the attribution of complexity weights to syntac- velopmental stages. In A. Housen, F. Kuiken, &
I. Vedder (Eds.), Dimensions of L2 performance and Fleury (Eds.), Actes des 12es Journées Internationales
proficiency: Complexity, accuracy and fluency in SLA d’Analyse Statistique des Données Textuelles (pp. 529–
(pp. 95–120). Philadelphia/Amsterdam: John 538). Paris.
Benjamins. Buysse, M., Housen, A., & Pierrard, M. (2017).
Bartning, I., & Kirchmeyer, N. (2003). Le développe- The development of complex sentences in sec-
ment de la compétence textuelle à travers les ond language acquisition—A crosslinguistic study
stades acquisitionnels en français L2 [The devel- of French and English. Studia Linguistica. doi:
opment of textual competence across the acquisi- 10.2137/0269258316679371
tion stages of L2 French]. Acquisition et interaction Byrnes, H. (2009). Emergent L2 German writing abil-
en langue étrangère, 19, 9–39. ity in a curricular context: A longitudinal study of
Bartning, I., & Schlyter, S. (2004). Itinéraires acquisi- grammatical metaphor. Linguistics and Education,
tionnels et stades de développement en français 20, 50–66.
L2 [Acquisition routes and developmental stages Byrnes, H., Maxim, H., & Norris, J. M. (2010). Realizing
in L2 French]. Journal of French Language Studies, advanced FL writing development in collegiate ed-
14, 281–299. ucation: Curricular design, pedagogy, assessment.
Beaman, K. (1984). Coordination and subordination re- Modern Language Journal, 94 (Supplement 2010),
visited: Syntactic complexity in spoken and writ- iv–vi, 1–235.
ten narrative discourse. In D. Tannen (Ed.), Cosme, C. (2007). Clause linking across languages. A
Coherence in spoken and written discourse (pp. 45–80). corpus-based study of coordination and subordination in
Norwood, NJ: Ablex. English, French and Dutch. (Unpublished doctoral
Benazzo, S., & Starren, M. (2007). L’émergence de dissertation). Université Catholique de Louvain,
moyens grammaticaux pour exprimer les relations Louvain-la-Neuve, Belgium.
temporelles en L2 [The emergence of grammati- Crossley, S. A., Louwerse, M. M., McCarthy, P. M., & Mc-
cal resources to express temporal relations in the Namara, D. S. (2007). A linguistic analysis of sim-
L2]. Acquisition et interaction en langue étrangère, 25, plified and authentic texts. Modern Language Jour-
129–157. nal, 91, 15–30.
Berman, R. A., & Verhoeven, L. (2002). Cross- Dahl, Ö. (2009). Testing the assumption of complex-
linguistic perspectives on the development ity invariance: The case of Elfdalian and Swedish.
of text-production abilities: Speech and writing. In G. Sampson, D. Gil, & P. Trudgill (Eds.), Lan-
Written Language & Literacy, 5, 1–43. guage complexity as an evolving variable (pp. 50–63).
Biber, D., & Gray, B. (2010). Challenging stereotypes Oxford: Oxford University Press.
about academic writing: Complexity, elaboration, De Clercq, B. (2015). The development of lexical com-
explicitness. Journal of English for Academic Purposes, plexity in second language acquisition: A cross-
9, 2–20. linguistic study of L2 French and English. EuroSLA
Biber, D., Gray, B., & Poonpon, K. (2011). Should we use Yearbook, 15, 69–94.
characteristics of conversation to measure gram- De Clercq, B., & Housen, A. (2016). The development of
matical complexity in l2 writing development? morphological complexity: A comparative study of
TESOL Quarterly, 45, 5–35. L2 French and English. Second Language Research.
Botel, M., Dawkins, J., & Granowsky, A. (1973). A syntac- doi:10.1177/0267658316674506
tic complexity formula. In W. H. MacGinitie (Ed.), Doherty, M. (2005). Introduction: Language-specific
Assessment problems in reading (pp. 77–86). Newark, conditions for discourse linking and appropriate-
DE: International Reading Association. ness. Linguistics, 43, 1–15.
Brezina, V., & Pallotti, G. (2015). Morphological com- Ehret, K., & Szmrecsanyi, B. (2016). An information-
plexity tool. Accessed 20 June 2016 at http:// theoretic approach to assess linguistic complexity.
corpora.lancs.ac.uk/vocab/analyse_morph.php In R. Baechler & G. Seiler (Eds.), Complexity and
Brezina, V., & Pallotti, G. (2017). Morphological com- isolation (pp. 71–94). Berlin: De Gruyter Mouton.
plexity in written ESL texts. Second Language Ellis, R. (2009). The differential effects of three types
Research. of task planning on the fluency, complexity, and
Bulté, B., & Housen, A. (2012). Defining and op- accuracy in L2 oral production. Applied Linguistics,
erationalising L2 complexity. In A. Housen, 3, 474–509.
F. Kuiken, & I. Vedder (Eds.), Dimensions of Ellis, R., & Barkhuizen, G. P. (2005). Analysing learner lan-
L2 performance and proficiency: Complexity, accu- guage. Oxford: Oxford University Press.
racy and fluency in SLA (pp. 21–46). Philadel- Ellis, R., & Yuan, F. (2004). The effects of planning on
phia/Amsterdam: John Benjamins. fluency, complexity and accuracy in second lan-
Bulté, B., & Housen, A. (2014). Conceptualizing and guage narrative writing. Studies in Second Language
measuring short-term changes in L2 writing com- Acquisition, 28, 50–84.
plexity. Journal of Second Language Writing, 26, 42– Fenk–Oczlon, G., & Fenk, A. (2008). Complexity trade-
65. offs between the subsystems of language. In M. Mi-
Buysse, M., & De Clercq, B. (2014). Matching the data: estamo, K. Sinnemäki, & F. Karlsson (Eds.), Lan-
Developing a multilingual corpus of language de- guage complexity: Typology, contact, change (pp. 43–
velopment. In E. Née, J. M. Daube, M. Valette, & S. 66). Philadelphia/Amsterdam: John Benjamins.
Fenk–Oczlon, G., & Fenk, A. (2014). Complexity trade- task performance. Language Learning, 62, 439–
offs do not prove the equal complexity hypothesis. 472.
Poznán Studies in Contemporary Linguistics, 50, 145– Kuiken, F., & Vedder, I. (2007). Task complexity and
155. measures of linguistic performance in L2 writing?
Ferrari, S. (2012). A longitudinal study of complex- International Review of Applied Linguistics in Lan-
ity, accuracy and fluency variation in second language Teaching (IRAL), 45, 261–284.
guage development. In A. Housen, F. Kuiken, & Kuiken, F., & Vedder, I. (2012). Syntactic complexity,
I. Vedder (Eds.), Dimensions of L2 performance and lexical variation and accuracy as a function of task
proficiency: Complexity, accuracy and fluency in SLA complexity and proficiency level in L2 writing and
(pp. 277–298). Philadelphia/Amsterdam: John speaking. In A. Housen, F. Kuiken, & I. Vedder
Benjamins. (Eds.), Dimensions of L2 performance and proficiency:
Foster, P., Tonkyn, A., & Wigglesworth, G. (2000). Mea- Complexity, accuracy and fluency in SLA (pp. 143–
suring spoken language: A unit for all reasons. Ap- 170). Philadelphia/Amsterdam: John Benjamins.
plied Linguistics, 21, 354–375. Lambert, C., & Kormos, J. (2014). Complexity, accuracy,
Graesser, A. C., McNamara, D. S., Louwerse, M. M., & and fluency in task-based L2 research: Toward
Cai, Z. (2004). Coh-Metrix: Analysis of text on more developmentally based measures of second
cohesion and language. Behavior Research Methods, language acquisition. Applied Linguistic, 35, 606–
Instruments, & Computers, 36, 193–202. 614.
Gunnarsson, C. (2012). The development of complex- Larsen–Freeman, D. (2006). The emergence of com-
ity, accuracy and fluency in the written production plexity, fluency, and accuracy in the oral and writ-
of L2 French. In A. Housen, F. Kuiken, & I. Vedder ten production of five Chinese learners of English.
(Eds.), Dimensions of L2 performance and proficiency: Applied Linguistics, 27, 590–619.
Complexity, accuracy and fluency in SLA (pp. 247– Lu, X. (2010). Automatic analysis of syntactic complex-
276). Philadelphia/Amsterdam: John Benjamins. ity in second language writing. International Journal
Han, Z., & Lew, W. M. (2012). Acquisitional complexity: of Corpus Linguistics, 15, 474–496.
What defies complete acquisition in second lan- Lu, X. (2011). A corpus-based evaluation of syntactic
guage acquisition. In B. Kortmann & B. Szmrec- complexity measures as indices of college-level
sanyi (Eds.), Linguistic complexity: Second language ESL writers’ language development. TESOL Quar-
acquisition, indigenization, contact (pp. 192–217). terly, 45, 36–62.
Berlin: De Gruyter Mouton. MacWhinney, B. (2000). The CHILDES Project: Tools for
Housen, A., & Kuiken, F. (2009). Complexity, accuracy, analyzing talk. Mahwah, NJ: Lawrence Erlbaum.
and fluency in second language acquisition. Ap- Malvern, D., Richards, B. J., Chipere, N., & Duran, P.
plied Linguistics, 30, 461–473. (2004). Lexical diversity and language development:
Housen, A., Kuiken, F., & Vedder, I. (2012). Complexity, Quantification and assessment. Basingstoke, UK: Pal-
accuracy and fluency: Definitions, measurement grave Macmillan.
and research. In A. Housen, F. Kuiken, & I. Vedder Martin, M., Mustonen, S., Reiman, N., & Seilonen, M.
(Eds.), Dimensions of L2 performance and proficiency: (2010). On becoming an independent user. In I.
Complexity, accuracy and fluency in SLA (pp. 1–20). Bartning, M. Martin, & I. Vedder (Eds.), Commu-
Philadelphia/Amsterdam: John Benjamins. nicative proficiency and linguistic development: intersec-
Housen, A., & Simoens, H. (2016). Cognitive perspec- tions between SLA and language testing research (pp.
tives on difficulty and complexity in L2 acqui- 57–80). Amsterdam: EuroSLA.
sition. Studies in Second Language Acquisition, 38, Mayer, M. (1969). Frog, Where are you? New York: Dial
163–175. Press.
Hunt, K. (1965). Grammatical structures written at three Michel, M. C., Kuiken, F., & Vedder, I. (2007). The influ-
grade levels. NCTE Research report No. 3. Cham- ence of complexity in monologic versus dialogic
paign, IL: NCTE. tasks in Dutch L2. International Review of Applied
IBM. (2015). IBM SPSS Statistics (Version 23.0). Linguistics, 45, 241–259.
Iwashita, N., Brown, A., McNamara, T., & O’Hagan, S. Miestamo, M. (2008). Grammatical complexity in a
(2008). Assessed levels of second language speak- cross-linguistic perspective. In M. Miestamo, K.
ing proficiency: How distinct? Applied Linguistics, Sinnemäki, & F. Karlsson (Eds.), Language complex-
29, 24–49. ity: Typology, contact, change (pp. 23–41). Philadel-
Jarvis, S. (2013). Capturing the diversity in lexical diver- phia/Amsterdam: John Benjamins.
sity. Language Learning, 63, 87–106. Miestamo, M. (2009). Implicational hierarchies and
Kerr–Barnes, B. (1998). The acquisition of connectors grammatical complexity. In G. Sampson, D. Gil, &
in French L2 narrative discourse. Journal of French P. Trudgill (Eds.), Language complexity as an evolv-
Language Studies, 8, 189–208. ing variable (pp. 80–97). Oxford: Oxford Univer-
Klein, W., & Perdue, C. (1997). The Basic Variety (or: sity Press.
Couldn’t natural languages be much simpler?). Neff, J., Dafouz, E., Diez, M., Prieto, R., & Chaudron,
Second Language Research, 13, 301–347. C. (2004). Constrastive discourse analysis: Argu-
Kormos, J., & Trebits, A. (2012). The role of task mentative text in English and Spanish. In C.
complexity, modality, and aptitude in narrative Moder & A. Martinovic–Zic (Eds.), Discourse across
languages and cultures (pp. 267–283). Philadel- B. Szmrecsanyi (Eds.), Linguistic complexity: Second
phia/Amsterdam: John Benjamins. language acquisition, indigenization, contact (pp. 6–
Norris, J. M., & Ortega, L. (2009). Towards an organic 34) Berlin: De Gruyter Mouton.
approach to investigating CAF in instructed SLA: Towell, R. (2012). Complexity, accuracy and fluency
The case of complexity. Applied Linguistics, 30, from the perspective of psycholinguistic second
555–578. language acquisition research. In A. Housen,
Ortega, L. (2003). Syntactic complexity measures and F. Kuiken, & I. Vedder (Eds.), Dimensions of L2
their relationship to L2 proficiency: A research performance and proficiency: Complexity, accu-
synthesis of college-level L2 writing. Applied Lin- racy and fluency in SLA (pp. 47–69). Philadel-
guistics, 24, 492–518. phia/Amsterdam: John Benjamins.
Ortega, L. (2012). Interlanguage complexity: A con- Véronique, D. (Ed.). (2009). L’acquisition de la grammaire
struct in search of theoretical renewal. In B. Kort- du français, langue étrangère. Paris: Didier.
mann & B. Szmrecsanyi (Eds.), Linguistic complex- Verspoor, M., Schmid, M. S., & Xu, X. (2012). A dynamic
ity: Second language acquisition, indigenization, con- usage-based perspective on L2 writing. Journal of
tact (pp. 127–155). Berlin: De Gruyter Mouton. Second Language Writing, 21, 239–263.
Pallotti, G. (2009). CAF: Defining, refining and differen- Vyatkina, N. (2012). The development of second lan-
tiating constructs. Applied Linguistics, 30, 590–601. guage writing complexity in groups and individ-
Pallotti, G. (2015). A simple view of linguistic complex- uals: A longitudinal learner corpus study. Modern
ity. Second Language Research, 31, 117–134. Language Journal, 96, 576–598.
Perdue, C. (Ed.), (1993). Adult language acquisition: Vyatkina, N. (2013). Specific syntactic complexity: De-
Cross-linguistic perspectives. Volume 2. The results. velopmental profiling of individuals based on an
Cambridge: Cambridge University Press. annotated learner corpus. Modern Language Jour-
Salamoura, A., & Saville, N. (2010). Exemplifying the nal, 97 (Supplement 2013), 11–30.
CEFR: Criterial features of written learner English Welcomme, A. (2013). Connecteurs et complexité syn-
from the English Profile Programme. In I. Bart- taxique dans les récits d’apprenants néerlando-
ning, M. Martin, & I. Vedder (Eds.), Communica- phones de FLE: Premiers résultats [Connectors
tive proficiency and linguistic development: Intersections and syntactic complexity in the narratives of
between SLA and language testing research (pp. 101– Dutch learners of L2 French]. In U. Paprocka–
131). Amsterdam: EuroSLA. Piotrowska, C. Martinot, & S. Gerolimich (Eds.),
Scarborough, H. S. (1990). Index of productive syntax. La complexité en langue et son acquisition (pp. 261–
Applied Psycholinguistics, 11, 1–22. 284). Lublin Poland: Towarzystwo Naukowe Kul
Schleppegrell, M. J. (1992). Subordination and linguis- Katolicki Uniwersytet Jana Pawla II.
tic complexity. Discourse Processes, 15, 117–131. Wolfe–Quintero, K., Inagaki, S., & Kim, H.–Y. (1998).
Szmrecsanyi, B. (2004, March). On operationalizing syn- Second language development in writing: Measures of
tactic complexity. Paper presented at the Journées fluency, accuracy, and complexity. Honolulu, HI: Na-
Internationales d’Analyse statistique des Données tional Foreign Language Resource Center.
textuelles, 2004. Louvain-la-Neuve, Belgium. Yoon, H.–J., & Polio, C. (2016). The linguistic develop-
Szmrecsanyi, B., & Kortmann, B. (2012). Introduction: ment of students of English as a second language
Linguistic complexity: Second Language acquisi- in two written genres. TESOL Quarterly. Early View
tion, indigenization, contact. In B. Kortmann & 6 May 2016; doi:10.1002/tesq.296

句法复杂性发展（2017）

Uploaded by

Copyright:

Available Formats

句法复杂性发展（2017）

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

句法复杂性发展（2017）

Uploaded by

Copyright:

Available Formats

A Cross-Linguistic Perspective on

A CONSIDERABLE BODY OF RESEARCH ON between these components (Miestamo, 2009).

Language English French

The high frequency of such connectives in our

Measure 1 2 3 4 NS Post-hoc η2p

LenNP French N 25 25 25 25 25 1,2,3,4<NS .233

Measure 1 2 3 4 NS Post-hoc η2p

DivNP French N 25 25 25 25 25 1,2,3,4<NS .342

Measure 1 2 3 4 NS Post-hoc η2p

% Mat French N 25 25 25 25 25 1>2>3,4>NS .648

Measure 1 2 3 4 NS Post-hoc η2p

%Com French N 25 25 25 25 25 1,2<3,4,NS .464

You might also like