SCL 104 04dir
SCL 104 04dir
SCL 104 04dir
pan
Development of L2 writing complexity
Com
Clause types, L1 influence and individual differences
ing
Hildegunn Dirdal
University of Oslo
lish
Pub
Recent research on complexity development has demonstrated the need to dis-
tinguish between specific types of subordinate clauses and consider L1 influence
and individual variation. The present study combines these factors and makes
finer distinctions between clause types than usual in complexity research. A case
ins
study following five Norwegian learners of English over four years shows that in-
dividual clause types have different developmental trajectories and that learners
am
may develop in syntactic diversity even when showing little change in amount of
subordination. A lack of specific clause types in the L1 can lead to late develop-
enj
ment of these in the L2, and higher uses of subordination in particular areas may
boost complexity in the L2. Individual differences were found mainly in pace.
nB
1. Introduction
-
Although the term has sometimes been used to signify the use of more difficult or
more mature features, such ‘relative complexity’ should be distinguished from the
pro
proficiency. In the area of syntax, then, this means “the expansion of the capacity to
use the additional language in ever more mature and skilful ways, tapping the full
cor
range of linguistic resources offered by the given grammar in order to fulfil various
communicative goals successfully” (Ortega 2015: 82).
Un
https://doi.org/10.1075/scl.104.04dir
© 2022 John Benjamins Publishing Company
82 Hildegunn Dirdal
y
Syntactic complexity can involve both coordination and subordination/embed-
pan
ding of elements, and this coordination or subordination/embedding can happen
on either the level of the clause or the level of the phrase. This study focuses on the
Com
use of different kinds of subordinate clauses, which constitute one set of the linguis-
tic resources referred to in the quote from Ortega above, and which are structures
that learners will need to use on both the clausal level and within phrases. The
focus on individual types of subordinate clauses comes as an answer to the call for
ing
more specific and detailed measures of subordination in complexity research, dis-
tinguishing between particular forms/structural types (Norris & Ortega 2009: 562;
lish
Lambert & Kormos 2014: 608; Biber et al. 2016: 664) and adding measures of di-
versification, i.e. describing not only how much subordination is used, but also the
range of structural types employed (De Clercq & Housen 2017: 317–318). The study
Pub
also aims to investigate individual variation and first language (L1) influence with
respect to the said development. Although there is some evidence of individual var-
iation (Casanave 1994; Larsen-Freeman 2006; Vyatkina 2012, 2013; Vyatkina et al.
ins
2015) and of L1 influence (Lu & Ai 2015; Ehret & Szmrecsanyi 2019) in complexity
development, few studies have investigated these issues systematically.
am
Against this background, Section 2 discusses the move towards more detailed
measures of syntactic complexity in the field and discusses findings related to
enj
sents and discusses the results, before Section 8 concludes the paper.
ofs
Research synthesis studies (Wolfe-Quintero et al. 1998; Ortega 2003; Norris &
Ortega 2009; Bulté & Housen 2012) have attested a wealth of measurements of
ted
syntactic complexity, but have also shown that many of them overlap in what they
measure. They have demonstrated that complexity studies most commonly use
rec
global measures of either length or subordination ratios. The former type of meas-
ures gives the mean length, usually in words, of various entities such as sentences,
cor
clauses, utterances or terminal units (T-units).1 The idea behind length as a measure
Un
1. The T-unit consists of a main clause and all its dependent clauses (Wolfe-Quintero et al.
1998: 70).
Development of L2 writing complexity 83
y
of complexity is that longer entities must be more complex. Subordination ratios, on
pan
the other hand, measure complexity by dividing the number of subordinate clauses
by the number of other units such as sentences, T-units, c-units, overall number of
Com
clauses or, for spoken language, analysis of speech units (AS-units).2
The reliance on such global measures could be linked to the fact that research
on syntactic complexity in an L2 has often been part of attempts to find yardsticks
of proficiency or indices of language development (Larsen-Freeman 2009: 580).
ing
However, the measures have received criticism for lumping together grammatical
features that have distinct functions (e.g. Biber et al. 2011: 29; Lambert & Kormos
lish
2014: 608; Biber et al. 2016: 649). In particular, it has been suggested that research-
ers need to measure coordination, clausal complexity and phrasal complexity
separately and that learners develop complexity in these areas at different times
Pub
(Ortega 2003: 514; Norris & Ortega 2009: 563–564). This had led to more studies
including coordination measures (besides measures of subordination) and focusing
on phrasal complexity (e.g. Parkinson & Musgrave 2014; Kreyer & Schaub 2018;
ins
Kuiken & Vedder 2019). It has also led some researchers to divide subordinate
clauses into different groups depending on their function as adverbial, nominal or
am
2017). These studies show that different clause types can have different develop-
mental trajectories. For example, Vercellotti and Packer (2016) found that inter-
nB
national students in an English for Academic Purposes course in the United States
increased their subordination rate over three semesters. However, the increase was
Joh
not equally distributed over different types of subordinate clauses, and students at
different proficiency levels showed increases in different types of clauses. Among
the low-intermediate students, there was a small increase in non-finite clauses and
-
relative clauses, although adverbial clauses remained the most common subordinate
clause type. Students at high-intermediate level increased their use of adverbial
ofs
from the start of the data collection (Vercellotti & Packer 2016: 187 [their Figure 1]).
Differences in the way that clauses contribute to overall complexity at differ-
ted
fine-grained distinctions are necessary, both for different types of phrasal constit-
uents and for specific clause types: “Measures of complexity should differentiate
cor
Un
2. The AS-unit is similar to the T-unit, but also comprises units smaller than a main clause and
the subordinate structures belonging to them (De Clercq & Housen 2017: 322). The AS-unit is
thus useful especially in analysing spoken language.
84 Hildegunn Dirdal
y
among the structural types and syntactic functions of dependent clauses and
pan
phrases” (2016: 664). Parkinson and Musgrave (2014) follow the call for such spec-
ificity in analyzing phrasal complexity and find clear differences between writers
Com
at different proficiencies in preferences with regard to specific modifiers, notably
attributive adjectives and nouns as premodifiers.
Instead of focusing on modification used at a specific level, such as the noun
phrase, this study will explore the development of the set of clause types that exists
ing
in English. I will use the clause types described in the Longman Grammar of Spoken
and Written English (Biber et al. 1999) with the aim of exploring the development
lish
of this sub-system of syntactic complexity in a linguistically motivated way.
This fine-grained division into clause types also has the benefit that it gives
us the possibility of measuring diversity in addition to amount of subordination.
Pub
Diversity is central to the concept of complexity, but has rarely been measured
in studies of syntactic complexity. Two notable exceptions are De Clercq and
Housen (2017) and Vercellotti (2019), both studies that investigate spoken lan-
ins
guage. De Clercq and Housen (2017) includes standard deviation in the length
of noun phrases, clauses and AS-units within the same text as measures of diver-
am
sity. They also use percentages of different clause types (matrix, coordinated and
subordinated, the latter divided into the three functional subtypes of adverbial,
enj
complement and relative clauses and the two formal types of finite and non-finite
clauses). The proportions of clause types showed a decrease of coordinated clauses
nB
and an increase in subordinate clauses over four proficiency levels of French and
English learners. At the lowest level, the learners of French used only complement
Joh
clauses (both finite and non-finite). Although all the other clause types were added
at higher levels, the increase in subordination was mainly due to increased use of
complement clauses and increased use of finite clauses. The English learners had all
-
clause types at the lowest level of the study, even if complement clauses continued
to be dominant throughout. Vercellotti (2019) scored each monologue in her oral
ofs
data base for the number of clauses it contained out of the six types she operated
with: independent, coordinated, adverbial, relative, complement-taking predicate
and non-finite. On average, the ESL students in the study used four different clause
pro
These studies have shown that learners complexify their speech not only by
using more subordination, but by adding to the variety of clause types they em-
rec
ploy. The present study will investigate such diversification in written language and
use an even more fine-grained division into clause types, in order to give added
cor
y
3. Individual variation and cross-linguistic influence
pan
in syntactic complexity development
Com
There are few studies focusing systematically on individual differences in syntac-
tic complexity development, but such differences have been found with respect
to both global (Casanave 1994; Larsen-Freeman 2006) and more specific meas-
ures (Vyatkina 2012, 2013; Vyatkina et al. 2015). Larsen-Freeman (2006: 599–600)
ing
found a steady increase in syntactic complexity (measured as clauses per T-unit)
over six months for a group of five Chinese learners of English, but very different
lish
trajectories for the individual learners, one exhibiting little change and another a
negative development. Vyatkina (2013) has shown how similar trends in global
complexity growth, such as length of sentence, can mask differences in the means
Pub
used by individual learners to achieve this growth. In a study of beginning learn-
ers of German, one of her focal students increased his sentence length mainly by
using more coordination of simplex clauses, whereas another student used more
ins
complex noun phrases and constructions with non-finite verbs (p. 24). With re-
spect to specific clause types, Vyatkina et al. (2015: 40–41, 48) have demonstrated
am
that beginning learners may start using adverbial and relative clauses3 at different
times and that the frequencies of these clauses may vary greatly between learners.
enj
The results from these studies show that it is important to take individual differ-
ences into account when studying complexity development and, further, that more
nB
More knowledge is also needed about the way in which the first language may
influence syntactic complexity development. In one of the few studies to address
this issue, Lu and Ai (2015) compared essays written by learners from seven dif-
-
for all measures. With respect to subordination, Chinese, Japanese and Russian
learners had significantly lower scores than the native speakers, whereas Tswana
learners had higher scores. French, Bulgarian and German learners did not differ
pro
significantly from the native speakers. In another study, Ehret and Szmrecsanyi
ted
3. Adverbial clauses and relative clauses were the only two types of subordinate clauses included
rec
in the study. It looks as if only finite adverbial and relative clauses were considered, even though
this is not made explicit in the paper. It is a problem that the terms are used in different ways in
the literature, sometimes to indicate formal categories (finite adverbial clauses starting with a
cor
subordinating conjunction and finite relative clauses with an overt or covert relative pronoun)
and sometimes to indicate functional categories (clauses functioning as adverbials or as modifiers
Un
of nouns).
86 Hildegunn Dirdal
y
(2019) found that L2 Dutch essays written by German learners were more complex
pan
than essays written by Italian, French and Spanish learners.
It is probable that cross-linguistic influence will have an effect on more specific
Com
measures of complexity as well, such as the use of particular clause types. Kreyer
and Shaub (2018), who focused on specific types of phrasal complexity, suggest
that their learners used adnominal infinitive clauses earlier than adnominal -ing-
and past participle clauses because of influence from their L1 German, where such
ing
clauses are much more frequent than adnominal participial clauses. Learner cor-
pus studies without any focus on complexity as such also provide evidence which
lish
indicates that the L1 may influence the use of particular clause types. For example,
Granger (1997) found that the Dutch learners in the International Corpus of Learner
English (ICLE, version 1.1; cf. Granger et al. 2002) had significantly lower frequen-
Pub
cies of past participle clauses in their writing than French and Swedish learners,
although all of them had an even greater underuse of -ing-clauses compared to
their native peers in the Louvain Corpus of Native English Essays (LOCNESS; cf.
ins
Granger 1998). Another contrast has been found between French and Spanish
learners of English with respect to complement clauses, the French learners in
am
the Longman Learners’ Corpus (LLC; cf. Gillard & Gadsby 1998) producing more
infinitive clauses than that-clauses and the Spanish learners showing the opposite
enj
pattern.4 The researchers suggest that the difference might be due to transfer, but
that the languages in question will have to be investigated carefully before we can
nB
vidual clause types. In order to find out whether it is indeed the first language that
is responsible for the patterns found, the L2 learner data will be compared not just
ofs
with texts written by native speakers of the target language, but also with texts
written by the learners in their L1 Norwegian. The next section outlines contrastive
differences that can be expected to play a role for Norwegian learners of English.
pro
ted
rec
4. Note that the LLC receives new material all the time, and that the results reported here were
cor
based on an early collection. See the Longman Corpus Network webpages for more information
(global.longmandictionaries.com/longman/corpus).
Un
Development of L2 writing complexity 87
y
4. Contrastive differences between English and Norwegian
pan
The greatest difference between English and Norwegian subordinate clauses is argu-
Com
ably the difference between English -ing-clauses and Norwegian present participle
clauses. In English, gerund and present participle clauses have merged and are (in
most cases) indistinguishable in form (Huddleston 2002: 1220–1222). Norwegian
can also form nouns from verbs by adding the suffix -ing, but such verbal nouns
ing
have not developed a clausal syntax (Næs 1952: 243). The difference can be il-
lustrated with Example (1), where English has the option of using both a verbal
lish
noun and an -ing-clause, and (2), showing that Norwegian has no present participle
clause corresponding to the verbal noun.
Pub
(1) He enjoyed the wrapping of Christmas gifts.
He enjoyed wrapping Christmas gifts.
(2) Han nøt pakkingen av julegaver.
ins
Gloss: He enjoyed wrapping-def of Christmas-gifts-def
*Han nøt pakking julegaver.
am
Norwegian present participle clauses thus have a much narrower range of functions
than English -ing-clauses. They function only in a few particular constructions:
nB
(1) as complements of the auxiliary-like verbs bli (‘remain’) and komme (‘come’),
the first giving an ingressive meaning (Faarlund et al. 1997: 653; Kinn 2014: 76,
79); (2) in the construction V-NP/pron-present participle, corresponding to a con-
Joh
struction with -ing-clauses that Quirk et al. (1985: 1202) call a variant of complex
transitive complementation; (3) as adverbials; and (4) as postmodifiers of nouns.
These uses are exemplified in (3a–d).
-
(3) a. Hun kom gående mot ham, men han ble sittende på benken.
Gloss: She came walking towards him, but he remained sitting on
ofs
bench-def.
b. De fant ham liggende på gulvet.
pro
In all four constructions, there are restrictions on the internal syntax of present
cor
participle clauses, disallowing the inclusion of objects and reflexives, and restric-
tions on the verbs they can contain, which have to be stative or denote unbounded
Un
88 Hildegunn Dirdal
y
activities (Behrens et al. 2012: 223–224). In the construction with komme (‘come’),
pan
we only find manner-of-movement verbs, and bli (‘remain’) only selects hete (‘be
called’), være (‘be’) and posture/movement verbs (Lødrup 2016: 383). Norwegian
Com
present participle clauses are therefore very infrequent and rather different from
-ing-clauses. A previous study focussing on the acquisition of -ing-clauses by the
same Norwegian learners as considered in this study concluded that the learners
did not make a cross-linguistic identification between Norwegian present partici-
ing
ple clauses and English -ing-clauses (Dirdal 2022). The cross-linguistic effect may
therefore be the same as for a structure that is completely lacking from the L1.
lish
English and Norwegian also exhibit differences in the structure and use of
relative clauses. Adnominal relative clauses in Norwegian are usually introduced
with the word som, which is arguably a subjunction rather than a relative pronoun,
Pub
as it cannot be inflected and can only occur at the beginning of subordinate clauses
(Faarlund et al. 1997: 25). Relative pronouns (of the same form as interrogative
pronouns) are only used in a very formal style, and only in Bokmål,5 the written
ins
standard historically influenced by Danish (see Faarlund et al. 1997: 1056–1057).
Nominal and sentential relative clauses are introduced by relative/interrogative
am
Relative clauses with som are used in both clefts and pseudo-clefts, rather than
ofs
5. Norway was under Danish rule from 1537 to 1814. During this period, Danish became the
written standard and influenced the spoken language of the elite. After independence, two written
cor
standards are mutually intelligible and may be regarded as “written dialects” (Vikør 2015).
Development of L2 writing complexity 89
y
If there is L1 transfer, these facts may lead to a boost of adnominal relative clauses
pan
and a lower number of sentential relative clauses and wh-clauses in the data from
the Norwegian learners of English when compared to native speakers.
Com
5. Previous studies of L2 English complexity in the Norwegian context
ing
There are few studies on complexity in the written L2 English of Norwegian learn-
ers. Most of them are cross-sectional and investigate texts collected at only one
lish
point in time, as the focus is on differences between teaching programmes (Larsen
2016), whether high- and low-grade achievement corresponds with various meas-
ures of complexity (Tjerandsen 1995) or how complexity in the L2 compares to
Pub
complexity in the pupils’ L1. However, there are also a couple of longitudinal
studies (Drew 2010; Raaen & Guldal 2012), which (together with findings from the
other studies) provide some information about complexity development over time
ins
in the written English of Norwegian pupils. Drew (2010) studied young learners
in years 4–6 (age 9–11), whereas Raaen and Guldal (2012) investigated pupils in
am
years 7 and 10.6 In the following, I will focus on the findings related to syntactic
complexity involving the use of subordinate clauses, although the studies also
enj
cover aspects such as fluency (Drew 2010; Larsen 2016), phrasal complexity (Drew
2003), lexical complexity (Drew 2010; Larsen 2016) and orthography (Raaen &
nB
Guldal 2012).
With respect to overall complexity, the studies show an increase in the use
Joh
of subordination over time. The youngest learners in Drew (2010), from year 4
in primary school, used hardly any subordination at all. There were a total of six
subordinate clauses appearing in only four of the 33 texts. Since Drew also reports
-
the mean number of T-units, this can be translated to a ratio of 0.02 subordinate
clauses per T-unit. The ratio increased to 0.19 in year 5 and 0.26 in year 6. The
ofs
learners in Larsen (2016) were from year 7. The ones following the Early Years
Literacy Program (with a focus on differentiated reading in combination with
literacy-promoting activities) had a rate of 0.29 subordinate clauses per T-unit,
pro
complex sentences for students in year 7 and 10, and report an increase in complex
sentences from 28.7% to 53.3%.
rec
cor
6. After the education reform in 1997, pupils in Norwegian schools receive English instruction
from year 1, although the written work in the early years is limited to experimenting with writing
Un
y
The studies also tell us something about the development in the use of adverbial,
pan
nominal and adnominal clause types. The young learners from years 4–6 in Drew
(2010) used nominal clauses most frequently over all three years, and adnominal
Com
ones least frequently. Drew does not give numbers for subtypes of clauses, but men-
tions that the nominal clauses were mainly infinitive clauses and that-clauses used
as objects in year 5, and that the pupils had started to use what- and how-clauses
in year 6. He also says that nominal -ing-clauses became more frequent in year 6,
ing
especially as objects of verbs.
Raaen and Guldal (2012) is the study with the most detailed classification into
lish
different clause types, focusing on form rather than function: adverbial, nominal
that, relative, infinitive and -ing-clauses, as well as an ‘other’ category. They report
relative clauses as the most frequent type in the data from year 7, closely followed
Pub
by that-clauses, infinitive clauses and adverbial clauses. There were few -ing-clauses
(5.7%) and even fewer ‘other’ clauses (2.1%). In year 10, there was a more even
distribution. Infinitive clauses were now the most frequent (21.4%), followed by
ins
that-clauses (19.5%), relative clauses (19.4%), -ing-clauses (18.5%) and adverbial
clauses (15.5%). The use of “other” clauses had increased to 4.1% of the total (and
am
of subordination from year 4 to year 10 and that they also use an increasing variety
of subordinate clauses. After an initial reliance on nominal clauses, learners increase
nB
their use of adverbial and adnominal clauses. Within the more specific form-based
clause types, -ing-clauses and wh-clauses seem to come into use later than other
Joh
first seem to use predominantly the clause types that they are familiar with from
Norwegian, and then progressively more of the unfamiliar clause types. They base
ofs
this on the fact that -ing-clauses are less frequent for the younger learners, and
that -ing-clauses and the ‘other’ category, which includes nominal relative clauses,
show the largest increase from year 7 to 10. As seen in Section 3, cross-linguistic
pro
differences may lend support to their claim, but unless we compare the Norwegian
learners to L1 writers of a similar age, we cannot know whether this pattern is really
ted
a similar degree of subordination in their two languages, most students had a lower
degree of complexity in L2 English, but wrote understandable and coherent texts,
Un
and about 10% wrote an English that was much inferior to their Norwegian and
Development of L2 writing complexity 91
y
very difficult to understand. Apart from this, the studies of Norwegian learners
pan
say little about individual variation. Although Drew (2010) and Raaen and Guldal
(2012) conducted longitudinal studies following the same learners over time, they
Com
do not describe individual developmental paths, but analyse the data mainly on
the group level.
ing
6. Research questions and method
lish
On the background of the identified gaps, my aim was to investigate syntactic com-
plexity development with respect to the system of subordinate clauses, taking into
account both L1 influence and individual variation, and making a finer distinction
Pub
between clause types than what is normal in complexity research. This will give a
more complete picture of the paths that Norwegian learners traverse and provide
a better understanding of how different factors influence syntactic complexity de-
ins
velopment. The study addresses the following questions:
am
RQ1. How does Norwegian learners’ L2 English writing complexity develop over
time with respect to different types of subordinate clauses?
enj
Answering these questions required a longitudinal design where the same individu-
als were followed over time. I used data collected for the TRAWL Corpus (Tracking
-
Written Learner Language; Dirdal et al. 2017), a longitudinal corpus still under
development. The corpus contains texts written by Norwegian schoolchildren in
ofs
L2 English, French, German and Spanish, as well as L1 Norwegian for some of the
pupils. All the texts are written as part of regular schoolwork, and task specifications
are available for most texts. I chose to do a case study focusing on five students, for
pro
whom there is L2 English data spanning a four-year period: from the start of lower
secondary school (year 8, age 13) to the end of the first year of upper secondary
ted
school (year 11, age 17). They all have Norwegian as their L1 and had lived only in
Norway. The corpus material also comprises some L1 Norwegian data from lower
rec
secondary school for three of them. The five pupils will be referred to using the
individual pupil codes that they have received for the TRAWL Corpus: P01002,
cor
text types were found to be markedly different in subordination rate, with more
92 Hildegunn Dirdal
y
subordinate clauses in the non-narrative ones. Since there are no narrative texts
pan
from year 11 and few from year 10, I decided to focus only on non-narrative texts.
I chose to use texts written in test situations at school to make sure that the students
Com
had not received help from others. However, in Norwegian schools, students are
often allowed to bring books and notes to tests and may also get texts to read and re-
spond to. Most of the tests included in the material consist of several tasks, some of
which gave the students a choice between different options. Since narrative options
ing
were not included in the present study, this means that there may be more material
from some students than others from the same writing occasion. The nature and
lish
variety in tasks will be returned to in the discussion of the results. There are texts
from one test per semester from years 8 and 9 and two per semester for years 10
and 11, although texts are missing for one of the year 11 occasions for three of the
Pub
students. The writing occasions are fairly evenly distributed over the years, and the
texts are written in November and April/May, as well as September and January/
February in the final two years. Three of the students contributed Norwegian texts
ins
to the corpus. Unfortunately, there are few texts available from years 8 and 9. It was
therefore decided to focus on year 10 for the investigation of L1 effects.
am
Norwegian students with L1 English data from the Growth in Grammar Corpus
(GiG Corpus; Durrant & Brenchley 2018). This is a pseudo-longitudinal corpus
nB
with texts from British school children, written as part of regular schoolwork like
the texts collected for the TRAWL Corpus, but in L1 English. It contains data from
Joh
years 2, 4, 6, 9 and 11 in the British school system. It is debatable whether the year
10 data from the Norwegian learners should be compared with year 9 or 11 from the
GiG Corpus. Since children start school one year earlier in Britain than in Norway,
-
the students in year 11 would be of the same age. However, they would have had
an extra year of schooling. The data was initially compared with both years and
ofs
was found to be slightly closer to the GiG data from year 9 in overall subordination
rate. Since the point of the comparison was to disentangle L1 effects from general
developmental effects, it was thus deemed best to compare the Norwegian students
pro
the distinction between literary and non-literary texts used in the GiG Corpus.
There are 30 literary and 30 non-literary texts for each year in the annotated part
rec
of the corpus. However, about half of the non-literary texts are from science classes
(lab reports and similar). In order to have material that was as similar as possible
cor
to that from the TRAWL Corpus, I only included the texts from English classes,
yielding fourteen non-narrative texts altogether. Table 1 gives an overview of the
Un
size of the material used for the study in terms of the total number of words.
Development of L2 writing complexity 93
y
Table 1. Total numbers of words in the material
pan
(number of writing occasions in parenthesis)
Year 8 Year 9 Year 10 Year 11 Total
Com
P01002 English 1196 (2) 1438 (2) 3523 (4) 2634 (3) 8791 (11)
Norwegian 1625 (2) 1768 (2)
P01007 English 1016 (2) 310 (2) 2706 (4) 2375 (4) 6407 (12)
Norwegian 1363 (2) 1449 (2)
ing
P01015 English 319 (2) 1760 (2) 3402 (4) 2853 (4) 8334 (1)
Norwegian 1104 (2) 1261 (2)
lish
P01029 English 365 (2) 2308 (2) 2819 (4) 2510 (3) 8002 (11)
P01032 English 358 (2) 450 (2) 2140 (4) 2139 (3) 5087 (11)
Pub
GiG (n = 14) English 4983 4983
201). The main types of finite clauses listed are that-clauses, wh-clauses, adverbial
clauses, comparative clauses and relative clauses. Wh-clauses can be further di-
enj
vided into wh-interrogative clause and nominal relative clauses. The “clearest type of
wh-interrogative clause” (Biber et al. 1999: 683) follows verbs like ask and wonder.
nB
In these cases it is possible to paraphrase with a direct question (7), whereas nom-
inal relative clauses can be paraphrased using a noun phrase with an adnominal
Joh
in Examples (7) and (8)) are indistinguishable in form, and not all of them fall as
clearly into each category as these. For the wh-clause in (9), for instance, neither
pro
clauses: reporting clauses, comment clauses, question tags and declarative tags.
As reporting clauses have often been analysed as main clauses taking the reported
speech as a direct object (Biber et al. 1999: 196), and as both the reporting clause
Un
94 Hildegunn Dirdal
y
and the reported speech are similar in form to independent clauses, they have both
pan
been considered main clauses in this study. Comment clauses and tags were coded
but were infrequent in the data7 and will not be discussed here.
Com
Biber et al. list four types of non-finite clauses, based on the form of the verb:
infinitive clauses, -ing-clauses, -ed-clauses (past participle clauses) and verbless
clauses, the latter of which were not found in the material. The non-finite clauses
can be subordinate to both clauses and phrases. So can wh-clauses and that-clauses.
ing
Most relative clauses have an adnominal function (i.e. as modifiers of nouns), but
so-called sentential relative clauses modify the proposition as a whole and can be
lish
regarded as a type of adverbial (Biber et al. 1999: 867). Adverbial clauses function
at the clause level, whereas comparative clauses are parts of phrases. Table 2 sum-
marizes the clause types coded and gives examples from the material.
Pub
Table 2. Clause types with examples from the L2 English data
ins
Clauses with adverbial function
Adverbial She had lived in the same house since she was born (P01002, year 8)
am
clauses
Sent. rel. That set big limits for getting your dream job which is terrible. (P01002,
enj
Infinitive To get rid of stereotyped views of other people, I think we have to include
clauses more people in more stuff and get a bigger diversity. (P01007, year 9)
Joh
-ing-clauses One genuine good person can do something that looks bad without
becoming a bad person. (P01015, year 11)
Past part. Anthony and Peter do their actions based on prejudice and the thought that
-
clauses every white person is racist against the blacks. (P01015, year 11)
Clauses with nominal function / complements of verbs
That-clauses I think English will remain a global language in the future (P01007, year 9)
ofs
clauses
-ing-clauses Snowden himself do not regret doing it (P01015, year 11)
ted
7. There were altogether six comment clauses in the Norwegian material, five in the L2 English
material and none in the L1 English material. There were two tags in one of the L2 English essays.
Un
Development of L2 writing complexity 95
y
Table 2. (continued)
pan
That-clauses Farhad refuses to replace the door, despite the fact that the old one is no
longer safe (P01015, year 11)
Com
Wh-clauses You have no idea how much you mean to me (P01015, year 9)
Infinitive I don’t think there should be an obligation to go in the army at all (P01002,
clauses year 10)
-ing-clauses … maybe they are refugees fleeing from their homeland … (P01007, year 10)
ing
Past part. Furthermore, he is also a deductive genius who always catches every single
clauses detail presented to him … (P01007, year 11)
lish
Clauses modifying adjectives or adverbs
Comp. clauses … and without Arthur Phillips arrival in 1788, Australia would not be the
Pub
same as it are today. (P01032, year 10) From this documentary I got informed
more about the situation than we usually do from general news. (P01002,
year 11)
Relative clauses I’m happy because mommy is so much happier now that she doesn’t need to
ins
see dad every day (P01015, year 9)
That-clauses …but I am sure that I will get used to it eventually. (P01015, year 10)
am
Wh-clauses I’m not sure what I think of this way to get them in to society. (P01015,
year 9)
enj
-ing-clauses She tells me that working at a farm in Australia, is entirely different from
working on a farm in Norway. (P01029, year 11)
Joh
tion. If the preposition was part of a prepositional verb, the clause was regarded as
having a nominal function. If the preposition was selected by a noun, the clause
was regarded as adnominal, and if it was selected by an adjective or adverb the
ofs
The Norwegian data were coded according to the same categories, as there are
equivalent clause types in Norwegian, except that present participle clauses were
coded instead of -ing-clauses (see the description of cross-linguistic differences
ted
in Section 4). The annotated part of the GiG Corpus, which was the source of L1
data for this study, contains information about clause types and functions, and was
rec
kindly made available to me by the creators. However, for easier comparison and
data handling, it turned out to be useful to code the material in the same manner
cor
as the L2 texts.
Un
96 Hildegunn Dirdal
y
The frequencies of the different clause types were normalized per T-unit. The
pan
T-unit is defined as a main clause with all the subordinate clauses belonging to it
and with its own subject. The T-unit was originally suggested by Hunt (1965) as a
Com
useful unit for studies of children’s writing, since punctuation can be erratic, and
the name “terminal unit” indicates a unit that could function on its own. In the case
of subordinate clauses coordinated with other subordinate clauses, I only counted
them as separate clauses if they contained a new subject (or a new covert subject in
ing
the case of non-finite clauses). In a few cases, it was impossible to determine what
kind of clause a student had intended or what function it had. These instances were
lish
left out of the analysis and amounted to 18 clauses from the L1 English material, 33
clauses from the L2 English material and 6 clauses from the L1 Norwegian material.
An Excel file was programmed to pick out the codes and register the file name and
Pub
the sentence in which the code occurred so that the clauses could be inspected.
ins
7. Results
am
Figure 1 gives an overview of the number of subordinate clauses and T-units pro-
duced by each student at each writing occasion translated into overall subordina-
enj
tion rates in the graph. At the group level, there is a small increase in subordination
rate from the first to the last writing occasion (1.19–1.41 subordinate clauses per
nB
T-unit), but – as can be seen from Figure 1 – there is vacillation over time for all
students. There also seem to be individual differences in that one student in par-
Joh
ticular (P01032) uses fewer subordinate clauses over time. As will be shown in the
following sections, these overall numbers conceal clear differences in the trajecto-
ries of individual clause types, as well as similarities between P01032 and the other
-
types will be discussed and common trends as well as individual differences will
be explored. I will treat clauses with adverbial, nominal and adnominal func-
tion in separate sections (7.1–7.3). The less frequent group of clauses modifying
pro
y
.
pan
.
.
Com
.
.
ing
.
.
lish
A S A S A1 A2 S1 S2 A1 A2 S1 S2
Year8 Year9 Year10 Year11
P01002 / / / / / / / / / / /
Pub
P01007 / / / / / / / / / / / /
P01015 / / / / / / / / / / / /
ins
P01029 / / / / / / / / / / /
P01032 / / / / / / / / / / /
am
Figure 1. Subordinate clauses per T-unit in the L2 English of the five learners
over the twelve writing occasions (A = autumn, S = spring)
enj
nB
As can be seen in Figure 2, P01032 and P01007 use a much higher number of clauses
Joh
with adverbial function than the rest at the beginning of the period. However, some
of the texts with the highest ratios are also among the shortest (7–10 T-units), which
makes it difficult to know how representative the numbers are. Apart from these
-
high points, there seems to be a slight increase over time in the use of clauses in
adverbial function, but with large fluctuations.
ofs
1.00
P01002
0.90
pro
P01007
0.80
0.70 P01015
0.60 P01029
ted
0.50 P01032
0.40
rec
0.30
0.20
cor
0.10
0.00
A S A S A1 A2 S1 S2 A1 A2 S1 S2
Un
y
The large differences between some of the students at the start do not seem to be
pan
due to task. They all answered exactly the same sets of questions in the autumn of
year 8 and in September and January of year 10, and at the writing occasions where
Com
there was a choice for one of the essays, the differences in choices do not align with
the differences in frequencies.
Figure 3 shows that finite adverbial clauses dominate among the clauses used in
adverbial function. The downward trend we saw for P01007 and P01032 is due to a
ing
decrease in the use of this type of clause only, and P01002 also produce fewer such
clauses over time. With respect to other clause types, there is generally an upwards
lish
trend, albeit slight, for all the five students.
P01002
Pub
.
Adverbial clauses
.
Sentential relative clauses
.
Wh-clauses
ins
. Infinitive clauses
. -ing-clauses
am
. .
. .
Joh
. .
. .
. .
-
A S A S A1 A2 S1 S2 A1 A2 S1 S2 A S A S A1 A2 S1 S2 A1 A2 S1 S2
. .
. .
pro
. .
. .
ted
. .
A S A S A1 A2 S1 S2 A1 A2 S1 S2 A S A S A1 A2 S1 S2 A1 A2 S1 S2
y
Because new clause types start to get used, there is an increase in diversity. The
pan
students begin with exclusive reliance on finite adverbial clauses in year 8 (except
for P01007, who uses one -ing-clause and one sentential relative clause in the second
Com
semester). Gradually, new clause types enter the scene, and as they increase and
the finite adverbial clauses decrease in frequency, there is a move towards greater
balance in the types of clauses used. However, for P01029 finite adverbial clauses
are still very dominant at the end of year 11.
ing
7.2 Clauses with nominal function
lish
Clauses used in nominal function also exhibit great fluctuations in frequency from
Pub
one writing occasion to the next, as shown in Figure 4. Just like we saw for clauses
with adverbial function, there is more variation between students at the beginning
of the period. Again, choice of task does not seem to explain why students diverge
ins
in frequencies at the same writing occasion.
1.20
am
P01002
1.00 P01007
enj
P01015
0.80
P01029
nB
0.60 P01032
0.40
Joh
0.20
0.00
-
A S A S A1 A2 S1 S2 A1 A2 S1 S2
The breakdown into more specific clause types8 in Figure 5, shows that that-clauses
and infinitive clauses dominate, but more so at the start of the period. In general,
there is a slight downward trend in the frequency with which these clause types
ted
are used.
Again, we see a move towards more diversity. Only one student uses nominal
rec
-ing-clauses in year 8 – P01002 has three in the spring semester. The other students
start using them in year 9, and P01032 as late as year 10. Wh-clauses also come in
cor
8. There were also three cases of clauses with the form of adverbial clauses functioning as subject
Un
y
P01002
pan
0.80
0.70 That-dauses
0.60 Wh-clauses
0.50
Infinitive clauses
Com
0.40
030 -ing-clauses
0.20
0.10
0.00
A S A S A1 A2 S1 S2 A1 A2 S1 S2
ing
YEAR 8 YEAR 9 YEAR 10 YEAR 11
P01007 P01015
0.80 0.80
lish
0.70 0.70
0.60 0.60
0.50 0.50
Pub
0.40 0.40
030 030
0.20 0.20
0.10 0.10
0.00 0.00
ins
A S A S A1 A2 S1 S2 A1 A2 S1 S2 A S A S A1 A2 S1 S2 A1 A2 S1 S2
0.50 0.50
0.40 0.40
030 030
nB
0.20 0.20
0.10 0.10
0.00 0.00
A S A S A1 A2 S1 S2 A1 A2 S1 S2 A S A S A1 A2 S1 S2 A1 A2 S1 S2
Joh
more gradually for most of the students, but are generally used more frequently
ofs
than -ing-clauses. With the new clause types and the decrease in that- and infinitive
clauses, there is more balance between different types towards the end of the period.
pro
Over time, there is a slight increase in clauses used to modify adjectives and adverbs
(Figure 6). However, there are not enough clauses for an analysis of the develop-
cor
ment of individual clause types (only 111 clauses altogether over the whole period).
Infinitive clauses are the most frequent in this function (44 instances), followed by
Un
y
0.25
pan
P01002
P01007
0.20
P01015
Com
0.15 P01029
P01032
0.10
ing
0.05
lish
0.00
A S A S A1 A2 S1 S2 A1 A2 S1 S2
YEAR 8 YEAR 9 YEAR 10 YEAR 11
Pub
Figure 6. Frequencies of clauses modifying adjectives/adverbs (clauses per T-unit)
ins
As shown in Figure 7, there is a clear upward trend in the frequencies with which
adnominal clauses are used (although P01007 has a high rate of subordinate clauses
am
at his first writing occasion, which does not fit the pattern).
enj
0.90
P01002
0.80
P01007
nB
0.70
P01015
0.60
P01029
0.50
Joh
P01032
0.40
0.30
-
0.20
0.10
0.00
ofs
A S A S A1 A2 S1 S2 A1 A2 S1 S2
YEAR 8 YEAR 9 YEAR 10 YEAR 11
pro
Similar observations can be made for the development of specific adnominal clause
ted
types as for those with adverbial and nominal functions. Figure 8 shows that one
clause type dominates, in this case relative clauses. For most students, other clause
rec
types increase slightly in use over time. However, in this case the dominant clause
type does not show any common decrease in frequency.
cor
Apart from relative clauses, other clause types used in adnominal function are
rare. Often a particular type features only once or twice on the same writing occa-
Un
sion. It is therefore not easy to spot clear trends. But although the increase in the use
102 Hildegunn Dirdal
y
of these clauses is not as obvious as for the new clause types that were introduced in
pan
adverbial and nominal functions, all the students exhibit more diversity towards the
end of the period than at the beginning. Two students use only relative clauses in
Com
year 8. Infinitive and past participle clauses seem to be introduced earlier than -ing-,
wh- and that-clauses, which the students use for the first time in either year 9 or 10.
. P01002
ing
Relative clauses
. That-clauses
Wh-clauses
lish
. Infinitive clauses
. -ing-clauses
Pub
. Past participle clauses
A S A S A1 A2 S1 S2 A1 A2 S1 S2
P01007 P01015
ins
. .
.
. .
.
am
.
. .
enj
. .
.
. .
A S A S A1 A2 S1 S2 A1 A2 S1 S2 A S A S A1 A2 S1 S2 A1 A2 S1 S2
nB
P01029 P01032
. .
Joh
. .
-
. .
. .
. .
ofs
A S A S A1 A2 S1 S2 A1 A2 S1 S2 A S A S A1 A2 S1 S2 A1 A2 S1 S2
Figure 8. Developmental trends for adnominal clause types (clauses per T-unit)9
Relative clauses dominate most clearly for P01029, who uses very few other clause
ted
types in adnominal function. She was the student with least diversity for clauses in
adverbial function too.
rec
cor
9. Note that the scale is different for student P01007 in order to display his high frequencies of
relative clauses in year 8. A scale up to 0.50 was chosen for the others so that the different clause
Un
y
7.4 Cross-linguistic influence
pan
In this section, the L1 and L2 production of the three students for whom Norwegian
Com
data is available will be compared to data from the GiG Corpus. The L1 Norwegian
and L2 English data come from the texts written by P01002, P01007 and P01015 in
year 10, and the L1 English data from texts written by 14 L1 users of English in year
9 in the British school system. As the texts from the GiG Corpus are quite short, the
ing
data from the L1 English users will be analysed jointly. However, the frequencies
derived from the Norwegian students’ compositions will be displayed individually,
lish
as it is important to demonstrate intra-group homogeneity in order to claim that a
particular linguistic behaviour is due to L1 influence.
In an influential article, Jarvis (2000) puts forward three prerequisites for claim-
Pub
ing cross-linguistic influence: similarity between the L1 and the interlanguage (L2),
intra-group homogeneity and inter-group heterogeneity. He argues that any one of
these three types of evidence may be sufficient to point to L1 influence, but that the
ins
presence of two types of evidence strengthens the case. Testing for all three effects
at the same time gives the strongest evidence, although it is often not feasible (Jarvis
am
2000: 255). I will look for similarities between the learners’ L1 and L2 production
but will also check whether the three students all pattern in the same way. If not,
enj
the effect is probably not due to the L1, which they all share.
nB
.
L1 English
. P01002 Eng
Joh
P01007 Norw
. P01015 Eng
.
pro
.
Adverbial Sentential Wh-clauses Infinitive -ing-/pres. Past
clauses relative clauses participle participle
clauses clauses clauses
ted
Figure 9. Frequencies of clause types used in adverbial function (clauses per T-unit),
rec
y
Adverbial clauses dominate in the adverbial function in L1 English, L2 English and
pan
Norwegian alike. Overall, the L2 writers have a lower frequency of clauses used in
adverbial function: 0.29 clauses per T-units compared to 0.52 for the L1 users. On
Com
average, even adverbial clauses are less frequent, but the pattern is not consistent.
As can be seen in Figure 9, P01002 uses fewer adverbial clauses in his English than
his Norwegian and P01015 shows the opposite pattern. The patterns with respect to
infinitive clauses and -ing-clauses are more uniform and may indicate L1 influence.
ing
The frequencies of infinitive clauses in L2 English are close to those for Norwegian
for all three students, and much lower than for L1 English. All three students pro-
lish
duce only a very few -ing-clauses, which may be an effect of the lack of an equivalent
clause type in Norwegian. Sentential adverbial clauses and wh-clauses are very
infrequent even in the L1 English material, and there are no past participle clauses.
Pub
With such low frequencies, there is little scope for differences between the groups.
.
ins
L1 English
P01002 Norw
.
P01007 Eng
enj
P01015 Norw
.
Joh
.
That-clauses Wh-clauses Infinitive clauses -ing-/pres. part.
clauses
-
Figure 10. Frequencies of clause types used in nominal function (clauses per T-unit),
L1 English, L2 English and L1 Norwegian compared
ofs
However, the Norwegian students vary greatly in the frequencies with which they
use that-, wh- and infinitive clauses, and they also differ with respect to whether the
L2 English or L1 Norwegian text has more instances of a particular clause type (see
ted
Figure 10). This variation may be an effect of the fact that most nominal clauses are
grammatically selected by the verb, so that the clause type depends on the choice of
rec
lexis. With respect to -ing-clauses, however, we can see that the learners consistently
produce fewer of these than the L1 English group.
cor
It is interesting to note that the L2 English learners hardly lag behind the L1
English users in the area of adnominal subordination (Figure 11). The development
Un
Development of L2 writing complexity 105
y
.
pan
L1 English
. P01002 Eng
. P01002 Norw
Com
. P01007 Eng
. P01007 Norw
ing
P01015 Norw
.
.
lish
.
.
Relative That- Wh- Infinitive -ing-/pres. Past
Pub
clauses clauses clauses clauses participle participle
clauses clauses
Figure 11. Frequencies of clause types used in adnominal function (clauses per T-unit),
ins
L1 English, L2 English and L1 Norwegian compared
am
2011; Kyle & Crossley 2018). But whereas the L2 learners had lower subordina-
tion rates than the L1 English users with respect to both adverbial and nominal
nB
functions, the subordination rate for adnominal clauses in the L2 English texts is
0.25 clauses per T-unit, very close to the frequency in the L1 English texts of 0.28
Joh
adnominal clauses per T-unit. This is the only clause function where L1 Norwegian
has a higher subordination rate than L1 English, so the higher L2 rates may be an
effect of the learners’ L1.
-
English users, but the higher rates in Norwegian could be due to the use of relative
clauses in constructions that are not used in English, such as with singular demon-
pro
(10) … det man legger mest merke til er hvem som står for matlagingen i familien
og hvem som fikser det som er blitt ødelagt.
rec
Gloss: … that Ø one lies most mark to is who that stands for food-making-def
in family-def and who that fixes that that is become destroyed.
cor
‘…what one mainly notices is who deals with the cooking in the family and
who fixes what has been broken.’
Un
106 Hildegunn Dirdal
y
(11) Det er heller ikke store endringer som er gjort i oppbygningen av setningene…
pan
Gloss: There are neither not large changes which are done in construction-def
of sentences-def…
Com
‘Neither have any large changes been done to the construction of the sentences…’
ing
lish
7.5 Discussion
The first research question addressed in this study was how Norwegian learners’
Pub
L2 English writing complexity develops over time with respect to different types of
subordinate clauses. The data used came from texts written by five learners over a
period of four years, years 8–11 in the Norwegian school system. At the beginning
ins
of the period, clauses used in nominal function were the most frequent, followed
by clauses used in adverbial function, whereas adnominal clauses were less fre-
am
quent. This picture agrees with Drew’s (2010) findings that nominal clauses were
dominant in the data from even younger Norwegian learners of English from years
enj
4–6. De Clercq and Housen (2017) found the same for beginner German learners
of English.
nB
Over time, the five Norwegian learners evidenced no clear change in the com-
bined frequencies of subordinate clauses in nominal function and only a slight
Joh
growth for clauses in adverbial function for some of the students, whereas there
was a clearer upward trend for clauses used in adnominal function. These results
confirm previous findings that clausal subordination levels off or decreases at a
-
certain proficiency level and that phrasal complexity develops later (Ortega 2003;
Norris & Ortega 2009; Biber et al. 2011).
However, when we look at more specific clause types, continued development
ofs
is evident even on the clausal level (i.e. for clauses with adverbial and nominal
function) – development in terms of increased diversification. At the start, the
pro
learners relied on finite adverbial clauses in the adverbial function, and mainly on
infinitive clauses and that-clauses in the nominal function. Over time, new clause
ted
types were introduced in both functions. These new clause types increased in fre-
quency, whereas the dominant clause types became slightly less frequent for most
rec
ilar development in some respects, with a dominant clause type (relative clauses)
and an increase in diversity over time. However, there was no common decrease
in the use of the dominant clause type in this area.
Un
Development of L2 writing complexity 107
y
Diversification was not only evident in the introduction of new clause types,
pan
but also in the way that formally similar clause types were used. Infinitive clauses
and that-clauses were both frequent for the students from year 8, but mainly in
Com
nominal functions. Only from year 9 or 10 did the students start to use infinitive
clauses in adverbial function and that-clauses in adnominal function. -ing-clauses
also took up more diverse functions over time, being used in adverbial and nominal
functions before adnominal functions in this data set.
ing
The findings thus confirm the usefulness of distinguishing between different
clause functions (as done in e.g. Drew 2010; Verspoor et al. 2012; Vercellotti &
lish
Packer 2016; De Clercq & Housen 2017; Vercellotti 2019), but, more importantly,
they show that further detail in terms of formal clause types is required in order
to capture the development that is taking place in each functional domain. It is
Pub
positive that studies have started to pay attention to non-finite clauses, but these
cannot be lumped together and compared with finite adverbial, nominal and ad-
nominal clauses, or given a developmental order in relation to these. The results
ins
from the present study show clearly that infinitive clauses, past participle clauses
and -ing-clauses have different developmental trajectories and appear at different
am
times in the three functional domains. For example, infinitive clauses are used early
and frequently in nominal function, but appear later and is used less frequently in
enj
similarities than difference between the students. But there was more vacillation
and differences in frequencies in the earlier than the later years. Verspoor et al.
Joh
(2012) suggest that learners progress towards a norm and that this explains why
there is less individual variation for more advanced learners.
In addition, there seems to be variation in pace, some learners starting to use
-
-ing-clauses before others and showing a greater increase in the use of these in year
11. The data from Drew (2010) reviewed in Section 5 indicate that some learners
ofs
may start diversifying their clause use even earlier than the five students included in
this study. He found some examples of nominal -ing-clauses from year-6 students.
The findings are in line with those from Vyatkina et al. (2015), where beginners
pro
were found to vary in the time at which they started using different clause types (see
Section 3). In the present study, student P01029 in particular showed less diversity
ted
than the others in both adverbial and adnominal functions, and may be a learner
who is developing more slowly in this area than the others.
rec
The third and final research question concerned possible L1 influence on com-
plexity development. For this part of the study, the focus was on data from the three
cor
students from whom there was also L1 Norwegian data and on comparisons with
L1 English data from students of a similar age. The clearest and most consistent
Un
y
in Section 3, -ing-clauses do not really have any equivalent in Norwegian, and
pan
the learners consistently produced very few -ing-clauses compared to L1 English
writers. It is obviously difficult to claim intra-group homogeneity based on data
Com
from only three learners. However, the other two learners also evidenced late use of
-ing-clauses, and in fact even lower frequencies than the other three, and Raaen and
Guldal (2012) have also remarked on the relatively late development in the use of
-ing-clauses by Norwegian learners compared to other clause types (see Section 5).
ing
Still, larger-scale statistical studies should follow up these findings.
The data pointed to further L1 influence in the use of adnominal clauses, where
lish
high subordination rates (chiefly for relative clauses) may have been the cause of a
more similar subordination rate to the L1 English writers than was the case on the
clausal level. This is an indication that the L1 may help the learners in the process
Pub
of phrasal complexification, and that children learning an L2 may therefore not be
very far behind novice L1 writers.
Indications of L1 influence on the use of particular clause types in the L2 are
ins
found in other studies as well, e.g. Kreyer and Schaub (2018), who report that their
German learners of English used adnominal infinitive clauses earlier than expected
am
on the basis of frequencies in L1 English writing. They suggest that this is due to
influence from German, where such clauses are more frequent. Granger (1997)
enj
and Biber and Reppen (1998) also speculate that frequency differences between
learner groups with respect to specific clause types may be due to influence from
nB
their L1s. In contrast to these studies, the present study compares the learners’
L2 English writing to their L1 (Norwegian) writing (as well as L1 English writing
Joh
from children of the same age) and thus gives firmer evidence for the claim of
cross-linguistic influence.
-
8. Conclusion
ofs
This study has demonstrated that group means and global measures, such as overall
subordination rate, may mask both individual differences between learners and
pro
disposal in year 8 but rely on a few frequent clause types. Although the amount of
Development of L2 writing complexity 109
y
subordination does not increase much over time (at least on the clausal level), there
pan
is a move towards more diversity in terms of a larger set of clause types, a more even
distribution of clause types and the use of the same clause forms in more functions.
Com
Individual differences were found mainly in pace. Some learners seem to be
slower in developing clausal diversity, and individuals may differ with respect to the
time when they start using particular clause types. Comparisons with L1 English
students of a similar age, as well as L1 Norwegian data from the learners themselves,
ing
revealed influence from the L1 in the late development of -ing-clauses as well as the
frequent use of relative clauses.
lish
The choice of a case study of a few students allowed for a detailed manual
annotation of data and close investigation of individual patterns. The data base
should be extended in the future, and larger-scale statistical studies used to find out
Pub
whether these focal students are representative of Norwegian learners of English. It
will also be important to look more closely at the influence of teaching and school-
books as well as to investigate how the texts students respond to in their tests may
ins
be reflected in their own writing. It was not within the scope of this study to look
at such materials in detail.
am
Acknowledgements
I would like to thank the editors, Agnieszka Leńko-Szymańska and Sandra Götz, and two anon-
-
ymous reviewers for their insightful comments on previous versions of this chapter. I am also
grateful to Bjarte Berntsen for programming the Excel file used in the analysis of the data.
ofs
References
pro
Behrens, Bergljot, Fabricius-Hansen, Cathrine & Solfjeld, Kåre. 2012. Competing structures: The
discourse perspective. In Big Events, Small Clauses: The Grammar of Elaboration, Cathrine
ted
Biber, Douglas, Gray, Bethany & Poonpon, Kornwipa. 2011. Should we use characteristics of
conversation to measure grammatical complexity in L2 writing development? TESOL Quar-
terly 45(1): 5–35. https://doi.org/10.5054/tq.2011.244483
cor
Biber, Douglas, Gray, Bethany & Staples, Shelley. 2016. Predicting patterns of grammatical com-
plexity across language exam task types and proficiency levels. Applied Linguistics 37(5):
Un
639–668. https://doi.org/10.1093/applin/amu059
110 Hildegunn Dirdal
y
Biber, Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan & Finegan, Edward. 1999. Long-
pan
man Grammar of Spoken and Written English. London: Longman. Also published as Biber,
Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan & Finegan, Edward. 2021. Gram-
mar of Spoken and Written English. Amsterdam: John Benjamins.
Com
https://doi.org/10.1075/z.232
Biber, Douglas & Reppen, Randi. 1998. Comparing native and learner perspectives on Eng-
lish grammar: A study of complement clauses. In Learner English on Computer, Sylviane
Granger (ed.), 145–158. London: Longman.
ing
Bulté, Bram & Housen, Alex. 2012. Defining and operationalising L2 complexity. In Dimensions
of L2 Performance and Proficiency: Complexity, Accuracy and Fluency in SLA [Language
Learning & Language Teaching Series 32], Alex Housen, Folkert Kuiken & Ineke Vedder
lish
(eds), 21–46. Amsterdam: John Benjamins. https://doi.org/10.1075/lllt.32.02bul
Bulté, Bram & Housen, Alex. 2014. Conceptualizing and measuring short-term changes in L2
writing complexity. Journal of Second Language Writing 26: 42–65.
Pub
https://doi.org/10.1016/j.jslw.2014.09.005
Casanave, Christine P. 1994. Language development in students’ journals. Journal of Second Lan-
guage Writing 3(3): 179–201. https://doi.org/10.1016/1060-3743(94)90016-7
ins
De Clercq, Bastien & Housen, Alex. 2017. A cross-linguistic perspective on syntactic complex-
ity in L2 development: Syntactic elaboration and diversity. The Modern Language Journal
am
Dirdal, Hildegunn, Danbolt Drange, Eli-Marie, Graedler, Anne-Line, Guldal, Tale M., Hasund,
Ingrid K., Nacey, Susan L. & Rørvik, Sylvi. 2017. Tracking Written Learner Language
(TRAWL): A longitudinal corpus of Norwegian pupils’ written texts in second/foreign
Joh
languages. Poster presented at the 4th Learner Corpus Research Conference, Eurac Re-
search, Bolzano, 4–7 Oct. 2017. <https://susannacey.hihm.no/wp-content/uploads/2017/09/
Abstract-LCR-4-Poster-Dirdal-et-al..pdf> (8 March 2022).
-
Drew, Ion. 2003. A comparison of early writing in Norwegian L1 and English L2. In Multilin-
gualism in Global and Local Perspectives: Selected Papers from the 8th Nordic Conference on
Bilingualism, November 1–3, 2001, Stockholm–Rinkeby, Kari Fraurud & Kenneth Hyltenstam
ofs
(eds), 341–355. Stockholm: Centre for Research on Bilingualism, City of Stockholm and
Stockholm University.
Drew, Ion. 2010. A longitudinal study of young language learners’ development in written Eng-
pro
ac.uk).
Ehret, Katharina & Szmrecsanyi, Benedikt. 2019. Compressing learner language: An informa-
rec
Universitetsforlaget.
Gillard, Patrick & Gadsby, Adam. 1998. Using a learners’ corpus in compiling ELT dictionaries.
Un
y
Granger, Sylviane. 1997. On identifying the syntactic and discourse features of participle clauses
pan
in academic English: Native and non-native writers compared. In Studies in English Lan-
guage and Teaching: In Honor of Flor Aarts, Jan Aarts, Inge de Mönnink & Herman Wekker
(eds), 185–198. Amsterdam: Rodopi.
Com
Granger, Sylviane. 1998. The computer learner corpus: A versatile new source of data for SLA
research. In Learner English on Computer, Sylviane Granger (ed.), 3–18. London: Addison
Wesley Longman.
Granger, Sylviane, Dagneaux, Estelle and Meunier, Fanny (eds). 2002. International Corpus of Learner
ing
I229
English, Version 1.1. Université catholique de Louvain: Centre for English Corpus Linguistics.
Housen, Alex, De Clercq, Bastien, Kuiken, Folkert & Vedder, Ineke. 2019. Multiple approaches
to complexity in second language research. Second Language Research 35(1): 3–21.
lish
https://doi.org/10.1177/0267658318809765
Huddleston, Rodney. 2002. Non-finite and verbless clauses. In The Cambridge Grammar of the Eng-
lish Language, Rodney Huddleston & Geoffrey K. Pullum (eds), 1171–1271. Cambridge: CUP.
Pub
https://doi.org/10.1017/9781316423530.015
Hunt, Kellogg W. 1965. Grammatical Structures Written at Three Grade Levels. Urbana IL: The
National Council of Teachers of English.
ins
Jarvis, Scott. 2000. Methodological rigor in the study of transfer: Identifying L1 influence in the
interlanguage lexicon. Language Learning 50(2): 245–309.
am
https://doi.org/10.1111/0023-8333.00118
Kinn, Torodd. 2014. Verbalt presens partisipp. Norsk Lingvistisk Tidsskrift 32(1): 62–99.
Kreyer, Rolf & Schaub, Steffen. 2018. The development of phrasal complexity in German inter-
enj
mediate learners of English. International Journal of Learner Corpus Research 4(1): 82–211.
https://doi.org/10.1075/ijlcr.16011.kre
nB
Kuiken, Folkert & Vedder, Ineke. 2019. Syntactic complexity across proficiency and languages:
L2 and L1 writing in Dutch, Italian and Spanish. International Journal of Applied Linguistics
29(2): 192–210. https://doi.org/10.1111/ijal.12256
Joh
Kyle, Kristopher & Crossley, Scott A. 2018. Measuring syntactic complexity in L2 writing using
fine-grained clausal and phrasal indices. The Modern Language Journal 102(2): 333–349.
https://doi.org/10.1111/modl.12468
-
Lambert, Craig & Kormos, Judit. 2014. Complexity, accuracy, and fluency in task-based L2 re-
search: Toward more developmentally based measures of second language acquisition. Ap-
plied Linguistics 35(5): 607–614. https://doi.org/10.1093/applin/amu047
ofs
Larsen, Line. 2016. Fluency and Complexity in the Written English of Norwegian 7th Graders:
A Comparative Study of Pupils in the Early Years Literacy Program (EYLP) with Those in a
Textbook-Based Approach. Master’s thesis, University of Stavanger.
pro
Larsen-Freeman, Diane. 2006. The emergence of complexity, fluency, and accuracy in the oral and
written production of five Chinese learners of English. Applied Linguistics 27(4): 590–619.
https://doi.org/10.1093/applin/aml029
ted
Larsen-Freeman, Diane. 2009. Adjusting expectation: The study of complexity, accuracy, and
fluency in second language acquisition. Applied Linguistics 30(4): 579–589.
rec
https://doi.org/10.1093/applin/amp043
Lødrup, Helge. 2016. Verbal present participles in Norwegian: Controlled complements or parts
of complex predicates. In Proceedings of the Joint 2016 Conference of Head-driven Phrase
cor
Structure Grammar and Lexical Functional Grammar, Polish Academy of Sciences, Warsaw,
Poland, Doug Arnold, Miriam Butt, Berthold Crysmann, Tracy Holloway King & Stefan
Un
y
Lu, Xiaofei & Ai, Haiyang. 2015. Syntactic complexity in college-level English writing: Differences
pan
among writers with diverse L1 backgrounds. Journal of Second Language Writing 29: 16–27.
https://doi.org/10.1016/j.jslw.2015.06.003
Næs, Olav. 1952. Norsk grammatikk. 1. Ordlære: Bokmål og nynorsk på bakgrunn av språkhistorie
Com
og dialekter. Oslo: Fabritius.
Norris, John M. & Ortega, Lourdes. 2009. Towards an organic approach to investigating CAF in
instructed SLA: The case of complexity. Applied Linguistics 30(4): 555–578.
https://doi.org/10.1093/applin/amp044
ing
Ortega, Lourdes. 2003. Syntactic complexity measures and their relationship to L2 proficiency: A
research synthesis of college-level L2 writing. Applied Linguistics 24(4): 492–518.
https://doi.org/10.1093/applin/24.4.492
lish
Ortega, Lourdes. 2015. Syntactic complexity in L2 writing: Progress and expansion. Journal of
Second Language Writing 29: 82–94. https://doi.org/10.1016/j.jslw.2015.06.008
Pallotti, Gabriele. 2015. A simple view of linguistic complexity. Second Language Research 31(1):
Pub
117–134. https://doi.org/10.1177/0267658314536435
Parkinson, Jean & Musgrave, Jill. 2014. Development of noun phrase complexity in the writing of
English for Academic Purposes students. Journal of English for Academic Purposes 14: 48–59.
ins
https://doi.org/10.1016/j.jeap.2013.12.001
Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey & Svartvik, Jan. 1985. A Comprehensive
am
Based Insights into Teaching and Learning, Angela Hasselgren, Ion Drew and Bjørn Sørheim
(eds), 101–117. Bergen: Fagbokforlaget.
nB
Svenonius, Peter. 1998. Clefts in Scandinavian: An investigation. ZAS Papers in Linguistics 10:
163–190.
Tjerandsen, Eli. 1995. ‘Stories You May Think is Boring, Can She Make Funny’: Sentence Com-
Joh
plexity in the Written Work of Norwegian 8th Graders. Master’s thesis, University of Bergen.
Vercellotti, Mary Lou. 2019. Finding variation: Assessing the development of syntactic complex-
ity in ESL speech. International Journal of Applied Linguistics 29(2): 233–247.
-
https://doi.org/10.1111/ijal.12225
Vercellotti, Mary L. & Packer, Jessica. 2016. Shifting structural complexity: The production of
clause types in speeches given by English for academic purposes students. Journal of English
ofs
https://doi.org/10.1016/j.jslw.2012.03.007
Vikør, Lars S. 2015. Norwegian: Bokmål vs. Nynorsk. The Language Council of Norway. <https://
www.sprakradet.no/Vi-og-vart/Om-oss/English-and-other-languages/English/norwegian-
ted
dividuals: A longitudinal leaner corpus study. The Modern Language Journal 96(4): 576–598.
https://doi.org/10.1111/j.1540-4781.2012.01401.x
Vyatkina, Nina. 2013. Specific syntactic complexity: Developmental profiling of individuals
cor
based on an annotated learner corpus. The Modern Language Journal 97(1): 11–30.
https://doi.org/10.1111/j.1540-4781.2012.01421.x
Un
Development of L2 writing complexity 113
y
Vyatkina, Nina, Hirschmann, Hagen & Golcher, Felix. 2015. Syntactic modification at early
pan
stages of L2 German writing development: A longitudinal learner corpus study. Journal of
Second Language Writing 29: 28–50. https://doi.org/10.1016/j.jslw.2015.06.006
Wolfe-Quintero, Kate, Inagaki, Shunji & Kim, Hae-Young. 1998. Second Language Development
Com
in Writing: Measures of Fluency, Accuracy and Complexity. Honolulu HI: University of Ha-
wai‘i at Manoa, Second Language Teaching & Curriculum Center.
ing
lish
Pub
ins
am
enj
nB
Joh
-
ofs
pro
ted
rec
cor
Un
Un
cor
rec
ted
pro
ofs
-
Joh
nB
enj
am
ins
Pub
lish
ing
Com
pan
y