Lexical bundles and L1 transfer effects*

Magali Paquot
Université catholique de Louvain

This exploratory study makes use of Jarvis’s (2000) methodological framework

to investigate transfer effects on French EFL learners’ use of lexical bundles.
The study focuses on 3-word recurrent sequences that include a lexical verb in
the French component of the International Corpus of Learner English (ICLE)
as compared to nine other ICLE learner sub-corpora. Results are in line with
a usage-based view of language that recognizes the active role that the first
language (L1) may play in the acquisition of a foreign language. The different
manifestations of L1 influence displayed in the learners’ idiosyncratic use of
lexical bundles are traced back to various properties of French words, including
their collocational use, lexico-grammatical patterns, function, discourse con-
ventions, and frequency of use. Following Hoey (2005), these transfer effects are
subsumed under the general term of ‘transfer of primings’.

Keywords: phraseology, lexical bundles, lexical verbs, transfer, lexical primings

1. Introduction

The last ten years have witnessed a remarkable boom in the number of studies that
examine learners’ use of lexical bundles, i.e. “recurrent expressions, regardless of
their idiomaticity, and regardless of their structural status” (Biber et al. 1999: 990).
These repeated sequences of words may be grammatically complete (by contrast,
on the other hand) or incomplete (the nature of the, is based on the). They may be
composed of clause segments (e.g. I don’t know what) or be parts of phrases (e.g.
the use of). They are “conventionalized building blocks that are used as convenient
routines in language production” (Altenberg 1998: 122) and typically function as
referential markers (e.g. the end of the), text organizers (e.g. on the basis of, for
example), stance markers (e.g. it is possible to) or interactional discourse markers
(e.g. or something like that, thank you so much) (Biber et al. 2003). Corpus-query
tools often provide an option that retrieves repeated sequences of words of a given

International Journal of Corpus Linguistics 18:3 (2013), 391–417. doi 10.1075/ijcl.18.3.06paq

issn 1384–6655 / e-issn 1569–9811 © John Benjamins Publishing Company
392 Magali Paquot

length (e.g. two-word sequences, three-word sequences) fully automatically. This

method has been used extensively to compare the number of lexical bundles, their
structural characteristics and discourse functions in learner and native corpora.
As pointed out in Paquot & Granger (2012: 138), the results of such studies
are particularly difficult to compare. The lexical bundles investigated are of differ-
ent sizes (from two- to six-word bundles), and the settings used to extract them
may vary considerably. However, a number of general trends can be identified.
Learners tend to use more lexical bundles in writing when compared to native
speakers, but the overall number of recurrent word combinations tends to de-
crease as proficiency in the language (Reppen 2009) or the time spent in the tar-
get language environment (Groom 2009) increases. Most studies report a mixed
pattern of under- and overuse. For example, learner writing is often characterized
by an underuse of the most academic-like bundles, such as noun phrases with
postmodifier fragments (e.g. the idea that, the issue of), coupled with an overuse
of speech-like word sequences, such as and so, sort of and a lot of (De Cock 2003,
Juknevičienė 2009).
Some of the studies have put specific patterns of misuse, overuse and under-
use of lexical bundles down to the learners’ mother tongue. Allen (2011: 111), for
example, attributes Japanese learners’ overuse of it can be said that to the L1, as
its translational equivalent is repeatedly used in Japanese academic writing. Rica
(2010) notes that a large proportion of the multi-word connectors that Spanish
EFL writers overuse are very similar to the word sequences used in Spanish to
express similar meanings (e.g. I think, Sp. “Creo que”; for example, Sp. “por ejem-
plo”). However, no study has targeted transfer effects on EFL learners’ production
of recurrent word sequences as their primary object of investigation.
The main objective of the present work is to fill in this existing gap by con-
ducting a careful transfer study of lexical bundles in learner writing. The study
focuses on 3-word recurrent sequences that include a lexical verb in French EFL
learner writing and addresses the following research questions:
i. RQ1: How much of French learners’ idiosyncratic use of lexical bundles with
verbs can be attributed to L1 influence?
ii. RQ2: What type of transfer effect (e.g. transfer of form, transfer of function)
is most discernible?
It is hypothesized that lexical bundles are potentially transferable because they are
essentially semantically and syntactically compositional, thus typically unmarked
word combinations (cf. Kellerman 1978). It is also anticipated that transfer effects
will be particularly noticeable in the overuse of lexical bundles whose equivalent
forms fulfil specific discourse functions in French.
 Lexical bundles and L1 transfer effects 393

To answer the research questions, the study is grounded in Jarvis’s (2000) uni-
fied framework for the study of L1 influence (Section 2). Section 3 describes the
learner and native corpus data used. In Section 4, the different methodological
steps required are summarised. Section 5 offers the results of the analysis of trans-
fer effects on French EFL learners’ use of lexical bundles, and Section 6 provides
answers to the research questions in the light of the preceding sections. Section 7
contains concluding remarks.

2. Jarvis’s (2000) unified framework for the study of L1 influence

Transfer studies have too often fallen into the trap of making a case for L1 influ-
ence on the sole argument that the structure exists in the L1, thus relying exclusive-
ly on “‘shot-in-the-dark’ post hoc interpretive guesses which pass for explanations”
(Lightbown 1984: 245). To remedy this situation, Jarvis (2000) puts forward a uni-
fied framework for the study of L1 influence which is premised on the following
operationalizable definition of the construct of ‘transfer’:
L1 influence refers to any instance of learner data where a statistically significant
correlation (or probability-based relation) is shown to exist between some fea-
tures of learners’ IL performance and their L1 background. (Jarvis 2000: 252)

This definition of L1 influence translates into a list of at least three potential sourc-
es of evidence that transfer studies should consider altogether when presenting a
case for or against L1 influence:
i. Effect 1: Intra-L1-group homogeneity in learners’ IL performance is found
when learners who speak the same first language behave as a group with
respect to a specific second language (L2) feature. To illustrate this first L1
effect, Jarvis (2000) uses Selinker’s (1992) finding according to which Hebrew-
speaking­learners of English as a group tend to produce sentences in which
adverbs are placed before the object (e.g. I like very much movies).
ii. Effect 2: Inter-L1-group heterogeneity in learners’ IL performance is found
when “comparable learners of a common L2 who speak different L1s diverge
in their IL performance” (Jarvis 2000: 254). To illustrate, Jarvis (2000) refers
to a number of studies reported by Ringbom (1987) that have shown that
Finnish-speaking learners are more likely than their Swedish-speaking coun-
terparts to omit English articles and prepositions. Jarvis (2000) argues that
“this type of evidence strengthens the argument for L1 influence because it
essentially rules out developmental and universal factors as the cause of the
394 Magali Paquot

observed IL behaviour. In other words, it shows that the IL behaviour in ques-

tion (omission of function words) is not something that every learner does
(to the same degree or in the same way) regardless of L1 background” (Jarvis
2000: 254–255).
iii. Effect 3: Intra-L1-group congruity between learners’ L1 and IL performance
is found where “learners’ use of some L2 feature can be shown to parallel
their use of a corresponding L1 feature” (Jarvis 2000: 255). This is the type of
evidence that Selinker (1992) produced when showing that Hebrew-speaking
learners’ positioning of English adverbs parallels their use of adverbs in the
L1. The added value of this third effect is that it also has explanatory power by
showing what in the first language motivates the IL behaviour.
In a follow-up article, Jarvis (2010) acknowledges the existence of a fourth type of
evidence that was not accounted for in his original framework, viz. ‘intralingual
contrasts’, which he defines as “differences in learners’ performance on features of
the target language that vary with respect to how they correspond to features of
the source language” (Jarvis 2010: 175).

3. Data

The learner corpus data used for the present study come from the first version of
the International Corpus of Learner English (ICLE) (Granger et al. 2002). ICLE
texts share a number of learner and task variables, which were used as corpus-
design criteria. All the learners are young adults who study English as a Foreign
Language (EFL) at university. They are all in their second, third, or fourth year
and their proficiency level has commonly been described as advanced although
learner groups differ in proficiency (Granger et al. 2009: 12). Learner productions
share many task variables, notably for medium (writing), genre (academic es-
say), field (general English rather than English for Specific Purposes) and length
(between 500 and 1,000 words). Other variables differ. A majority of the learner
texts are argumentative, but the essays cover a wide range of topics (e.g. the death
penalty, euthanasia). Learner texts also differ in task conditions.
The focus of this study is on French learner writing but other ICLE sub-
corpora­were also used as comparable corpora to test for inter-L1-group het-
erogeneity in learners’ interlanguage (IL) performance (cf. Section 2). Table 1
provides a breakdown of the ten ICLE sub-corpora used. Learner essays in each
sub-corpus were carefully selected in an attempt to control for a number of task
variables which may affect learner productions (cf. Kroll 1990, Ädel 2008): all
 Lexical bundles and L1 transfer effects 395

the texts are untimed argumentative essays, potentially written with the help of
reference tools. Although essays written without the help of reference tools would
arguably have been more representative of what advanced EFL learners can pro-
duce, untimed essays with reference tools are used as they represent the majority
of learner texts in ICLE.

Table 1. Breakdown of ICLE essays

No. of essays No. of words Average no. of words per essay
Corpus under analysis
French (ICLE-FR) 228 136,343 598
Comparable corpora
Czech (ICLE-CZ) 147 130,768 890
Dutch (ICLE-DU) 196 162,243 828
Finnish (ICLE-FI) 167 125,292 750
German (ICLE-GE) 179 109,556 612
Italian (ICLE-IT) 79 47,739 604
Polish (ICLE-PO) 221 140,521 636
Russian (ICLE-RU) 194 165,937 855
Spanish (ICLE-SP) 149 99,119 665
Swedish (ICLE-SW) 81 48,060 593
TOTAL 1,641 1,165,524 697

To evaluate Effect 3, several corpora of French writing were used. The 1.6 billion
word frWaC was consulted via the Sketch Engine (Kilgarriff & Kosem 2012) to
assess the frequency and prototypicality of specific word combinations in French
for general purposes. It was also deemed necessary to query smaller but more
comparable corpora of expert and student writing to control for text type and
levels of writing expertise:
i. The humanities component of the online Scientext corpus, i.e. a 3,431,531
word corpus of French published articles, theses and proceedings in linguis-
tics, psychology, education and natural language processing.
ii. The Corpus de Dissertations Françaises (CODIF), i.e. a 92,832 word corpus of
argumentative essays written by French-speaking students on similar topics
to ICLE-FR.
Spot-checks were also sometimes made in the 100 million word British National
Corpus and the 2 billion word English corpus ukWaC to check the lexicogram-
matical and distributional properties of English word combinations and hence
identify possible intralingual contrasts (Jarvis 2010). The two corpora were que-
ried via the Sketch Engine (Kilgarriff & Kosem 2012).
396 Magali Paquot

4. Methodology

The methodology used involves several steps which are described here. Sec-
tion 4.1 covers the extraction of lexical bundles from ICLE texts. Section 4.2
provides the procedures and statistical tests used to operationalize Jarvis’s (2000)
unified framework on learner corpus data. Section 4.3 describes the method used
to rule out topic influence and the rationale behind this extra step.

4.1 Extraction of lexical bundles

The focus of the study is on potential transfer effects on French EFL learners’ use
of bundles with lexical verbs. Lexical bundles of 3 words were first extracted from
the ICLE French sub-corpus with the help of the computer software WordSmith
Tools 5 (Scott 2008). A minimum frequency threshold of 5 occurrences was ad-
opted. The resulting list was filtered manually and the bundles that included a
lexical verb were selected for further analysis. A Perl program was then used to
retrieve relative frequencies per 100 words for each of the selected bundles in the
1,641 learner texts that make up the ten learner corpora.

4.2 Applying Jarvis’s (2000) unified framework to learner corpus data

Intra-L1-group homogeneity is most evident when directly compared with in-

ter-L1-group heterogeneity (Jarvis 2000), and I therefore make use of comparison
of means tests to operationalize Jarvis’s (2000) unified framework for the study of
L1 influence on learner corpus data. As more than two learner populations are be-
ing compared, one-way between-groups analysis of variance (ANOVA) tests are
used to measure the first two potential L1 effects, i.e. intra-L1-group homogeneity
and inter-L1-group heterogeneity. An ANOVA examines two sources of variance:
the variance between the groups (i.e. inter-L1-group heterogeneity between the
different ICLE sub-corpora) and the variance between individuals or texts within
each group (i.e. intra-L1-group homogeneity as displayed in each ICLE sub-cor-
pus). The two types of variance are then compared with one another. If the vari-
ance between the learner corpora is significantly higher than the variance within
each learner corpus, the interpretation is that the corpora are not taken from the
same population. The result of an ANOVA is an F ratio which tells us whether at
least one group in the set is different from the other groups. The level of risk or
level of significance used in this study is p < 0.01.
Importantly, while an F ratio indicates whether a significant difference exists
somewhere between the learner populations, it does not identify precisely where
 Lexical bundles and L1 transfer effects 397

the difference is. A post-hoc test must then be conducted to pinpoint the learner
population(s) responsible for the significant difference. As the objective here is to
evaluate Effects 1 and 2, the comparisons of interest are those between the French
learner corpus and the other ICLE sub-corpora. The Dunnett’s test is considered
the most powerful post-hoc test whenever one group is compared with each of the
other groups (Howell 1997: 380–381) and is therefore used in this study.1 When
lexical bundles display significant differences in use between the French learner
group and at least half of the other learner populations as revealed by Dunnett’s
tests, there is a strong case for intra-L1-group homogeneity and inter-L1-group
heterogeneity. The criterion used according to which over half of the comparisons
need to be significant is arbitrary and probably a relatively conservative estimate.
It is, however, used in this exploratory study to validate the methodology. All sta-
tistical tests were performed with R (R Core Team 2012).
While the first two effects readily lend themselves to automatic and quanti-
tative evaluation, intra-L1-group congruity between French learners’ L1 and IL
performance does not. Assessing this third effect requires a more qualitative ap-
proach. First, the use of each lexical bundle was carefully analysed in ICLE-FR.
The next steps consisted in identifying the French potential “equivalent” of each
lexical bundle in context, describing its use in French L1 and comparing learners’
L1 and IL patterns of use.

4.3 Addressing the issue of topic variability in ICLE

Learner texts in ICLE are varied in topic, and there is no single topic that is evenly
distributed across the 10 sub-corpora used in this study. Topic variability must
however be addressed as lexical bundles are particularly prone to this factor
(­Cortes 2004) and the ICLE French sub-corpus is characterised by a strong bias
towards just one topic (“Europe 92: loss of sovereignty or birth of a nation?”). This
topic was selected by c. 40% of all the French learners, and more than 70% of all
the texts about Europe 92 in ICLE are to be found in the French component. As
the issue of topic variability could not be addressed a priori, it is dealt with just
before intra-L1-group congruity between French learners’ L1 and IL performance
(Effect 3) is tested. To rule out topic influence, the ICLE in-built corpus query tool
is used to analyse the distribution by essay prompt of all the bundles that display
intra-L1-group homogeneity and inter-L1-group heterogeneity (Effects 1 and 2).
If a lexical bundle only appears in French learners’ essays discussing the creation
and future of Europe and in no other ICLE text, this provides a strong indication
that topic is a much more likely explanation than L1 influence.
398 Magali Paquot

5. Results

This section presents the results obtained from the transfer study. The extrac-
tion procedure outlined in Section 4.1 made it possible to identify 273 bundles
with a lexical verb in the French learner corpus, which were submitted to further
analysis.

5.1 Testing Effects 1 and 2

An R script was written to assess Effects 1 and 2 for the 273 lexical bundles under
study. The ANOVA test identified 87 lexical bundles that present significant dif-
ferences in use among the ten learner corpora. Among these, 34 bundles (12.45%)
display significant differences in use between the French learner group and at least
half of the other learner populations as revealed by Dunnett’s tests, thus showing
both intra-L1-group homogeneity and inter-L1-group heterogeneity. Table 2 lists
the 34 bundles, their F ratio and p value, as well as the number of learner popula-
tions from which the French learner group differs significantly in its use of each
lexical bundle.

Table 2. The 34 bundles that show Effects 1 and 2

Bundle F p Number of significant
learner corpus comparisons
be considered as 3.075 0.00116 6
be tempted to 4.534 6.45e-06 9
considered as a 4.947 1.4e-06 9
considered as the 2.876 0.00226 7
deeply rooted in 3.101 0.00106 8
does it mean 2.99 0.00154 8
going to become 2.813 0.00278 8
I would say 3.142 0.000919 6
is to know 3.195 0.000767 9
keep its own 3.839 8.03e-05 9
keep their own 3.822 8.54e-05 9
*loose their identity 2.463 0.00867 7
not forget that 6.457 4.59e-09 9
role to play 2.947 0.00178 9
say that Europe 4.723 3.21e-06 9
speak of a 2.737 0.00357 8
take the example 5.121 7.3e-07 9
 Lexical bundles and L1 transfer effects 399

Table 2. (continued)
Bundle F p Number of significant
learner corpus comparisons
to be found 5.206 5.32e-07 7
to build a 4.274 1.67e-05 9
to create a 2.788 0.00302 6
to go further 2.485 0.00809 6
to know whether 2.85 0.00246 8
wait and see 4.699 3.52e-06 9
want to create 3.011 0.00143 8
was considered as 2.421 0.00991 6
we can say 3.192 0.000774 6
we can wonder 2.669 0.00446 6
we may wonder 3.338 0.000469 9
we must not 2.606 0.00549 8
will be allowed 3.261 0.000612 8
will be needed 3.299 0.000536 9
will be united 3.328 0.000484 9
will keep its 3.309 0.000518 9
would say that 3.696 0.000134 8

5.2 The influence of the topic

An analysis of the 34 significant bundles in the 1,641 learner texts and their dis-
tribution by essay prompt reveals that 14 lexical bundles only appear in ICLE-
FR essays that discuss the creation and future of Europe. These bundles are keep
its own, keep their own, say that Europe, to build a, wait and see, will be needed,
will be united, will keep its, will be allowed, does it mean, going to become, want
to create, *loose their identity, and to create a. The influence of topic is visible in
the selection of content words (e.g. say that Europe, want to create) as well as in
tense preferences (e.g. will be allowed, will be united, will keep its) (Examples (1)
and (2)).
(1) Europe will be united against USA and Japan.  (ICLE-FR)
(2) Each country will keep its own identity, currency, institutions and constitu-
tion.  (ICLE-FR)

The influence of topic was ruled out for the remaining 20 lexical bundles as they
were found in essays covering a range of prompts (cf. Table 3).
400 Magali Paquot

Table 3. 20 lexical bundles for which topic influence is ruled out

(ordered by decreasing frequency in ICLE-FR)
Lexical bundle Freq. Rel. freq. Texts
(100,000 words)
we can say 22 16.1 16
I would say 20 14.7 16
would say that 19 13.9 15
not forget that 19 13.9 18
considered as a 18 13.2 17
be considered as 18 13.2 17
to be found 17 12.5 17
we must not 12 8.8 11
take the example 10 7.3 9
considered as the 8 5.9 8
was considered as 7 5.1 6
deeply rooted in 7 5.1 6
be tempted to 7 5.1 7
is to know 7 5.1 6
speak of a 6 4.4 5
to know whether 6 4.4 6
we may wonder 6 4.4 5
we can wonder 5 3.7 5
to go further 5 3.7 5
role to play 5 3.7 5

5.3 Testing Effect 3

The simplest way to test Effect 3 is to check whether there are equivalent lexical
bundles in French. Before doing so, however, a quick scan of concordance lines
for the 20 remaining lexical bundles (Table 3) showed that some regrouping of
embedded word sequences was possible (sometimes making up longer and more
syntactically complete bundles such as I would say that or pinpointing shorter but
more salient word combinations, e.g. considered as). Intra-L1-group congruity
between learners’ L1 and IL performance was consequently evaluated for fifteen
lexical bundles (see Table 4). L1/IL equivalence in form was found for a majority
of the English lexical bundles; equivalence in meaning or function was established
for the four lexical bundles involving the first person plural pronoun we. Table 4
also provides the most frequent corresponding bundles in French as identified in
frWaC for each of the fifteen longer, syntactically complete or more salient lexical
 Lexical bundles and L1 transfer effects 401

bundles. Small capitals are used to represent lemmas rather than word forms. The
extent of the correspondence between the English and French lexical bundles is
discussed in Section 6.

Table 4. Lexical bundles and their most frequent equivalent forms in French
English lexical bundles Most frequent equivalent bundles in French
be tempted to être tenté/es de
considered as considéré/es comme
deeply rooted in profondément enraciné/es dans
I would say that je dirais que
is to know whether est de savoir si
not forget that pas oublier que
role to play rôle à jouer
speak of parler de
take the example prendre l’exemple
to be found être trouvé/es
to go further aller plus loin
we can say on peut dire
we can wonder on peut se demander
we must not il ne faut pas
we may wonder on peut se demander

6. Discussion

This section addresses the research questions guiding the study by discussing the
results provided in Section 5. The combination of the three effects investigated in
Section 5 points to a firm conclusion of L1 transfer for the twenty lexical bundles
for which topic influence was ruled out (Table 3). This represents as much as
58.8% of the lexical bundles that set the French learners apart from at least 5 other
learner populations (Section 5.1). Thus, to answer RQ1, over a half of French
learners’ idiosyncratic use of lexical bundles with verbs can be attributed to L1
influence.
A close look at the lexical bundles and their equivalent forms in French helps
identify four major types of transfer effect found in French EFL learners’ use of
recurrent word sequences, thus addressing RQ2: (i) transfer of collocational and
colligational preferences, (ii) transfer of syntactic constructions, (iii) transfer of
functions and discourse conventions and (iv) transfer of L1 frequency.
402 Magali Paquot

6.1 Transfer of collocational and colligational preferences

In a collocational study of amplifiers in French EFL learner writing, Granger

(1998) already interpreted French learners’ use of deeply rooted as a manifestation
of French influence: the collocation has a direct translation equivalent in French,
i.e. profondément enraciné. Not only is this word combination congruent with
deeply rooted but the adjective profondément is also the most frequent adjective
found to modify the past participle enraciné in the frWaC (318 occurrences; 0.2
per million). Interestingly, the collocation firmly rooted is as frequent as deeply
rooted in English (0.4 per million in ukWaC) but is not used by French EFL learn-
ers. It also has a congruent form in French, i.e. fermement enraciné, but this com-
bination is rare (14 occurrences in frWaC; 0.008 per million).
The first language may also prompt learners to use lexical bundles that display
untypical colligational patterns in English such as considered as. As shown in Ex-
amples (3) and (4), French EFL learners mostly use the verb consider followed by
the preposition as to introduce an adjective or a noun phrase (52 occurrences per
100,000 words in ICLE-FR).
(3) Why is this easiness an asset for the EEC to be considered as a nation?
(ICLE-FR)
(4) Besides childhood is often considered as the happiest period in one’s life.
(ICLE-FR)

French EFL learners’ preference for the construction consider + as mirrors the
use of French considérer, which is typically followed by the preposition comme
when introducing adjective or noun phrases (Examples (5) and (6)). In frWaC,
for example, considérer + comme + ADJECTIVE has a relative frequency of
11.5 pmw while the structure without the preposition appears with a relative fre-
quency of 2 pmw.
(5) Il est également considéré comme le fondateur de l’abbaye de Malmédy en
Belgique.  (frWaC)
(“He is also considered the founder of the abbey of Malmedy in Belgium.”)
(6) La nature a longtemps été considérée comme une réserve plutôt que comme
un patrimoine.  (frWaC)
(“Nature has long been considered a reserve rather than a heritage.”)
 Lexical bundles and L1 transfer effects 403

6.2 Transfer of syntactic constructions

Among the lexical bundles that distinguish the French learner population from
the other learner groups, several include to-infinitive constructions. As illustrated
in Example (7), French learners use the lexical bundle to go further although it is
not very frequent in English (0.9 pmw in ukWaC). By contrast, the French con-
gruent bundle aller plus loin is relatively frequent (8.9 pmw in frWaC).
(7) Nevertheless the Americans decided to go further and were the first who
wanted to stop Hussein and his army.  (ICLE-FR)

The lexical bundle to be found appears in several ICLE sub-corpora but it is most
frequent in the French learner sub-corpus where it is almost always preceded by
a noun phrase (NP) + the verb be (Examples (8) and (9)). This larger frame cor-
responds to French NP + être + à trouver, which is itself a lexical realisation of
the frequent French structure NP + être + à + VERB (over 20 pmw in frWaC).
The meaning of this French construction is more commonly expressed with the
modal verb should in English and the most frequent bundles that exemplify this
structure in frWaC include dossiers sont à retirer (“forms should be picked up”),
candidatures sont à adresser (“applications should be sent to”), précautions sont à
prendre (“precautions should be taken”), règles sont à respecter (“rules should be
followed”), and supplément est à payer (“extra charge should be paid”).
(8) The real problem is to be found in the fact that women who wish to have a job,
also desire to have a family life.  (ICLE-FR)
(9) Another example is to be found between the French and the Italian vine grow-
ers: […].  (ICLE-FR)

There are only two sentences where to be found is not used with the verb be and
they both feature the combination a balance has to be found (Examples (10) and
(11)). Tellingly, the choice of have to in these two sentences is consistent with the
preferred expression of modality in the congruent phrase in French: un équilibre
doit/devra être trouvé is twice as frequent as un équilibre est à trouver in frWaC (41
vs. 23 occurrences).
(10) A balance has thus to be found.  (ICLE-FR)
(11) And a balance between the two orientations has to be found.  (ICLE-FR)

Similarly, the lexical bundle role to play is always introduced by the verb have
in ICLE-FR (Example (12)) and this larger word combination is congruent with
avoir un rôle à jouer, which is the most frequent lexical realisation of the French
404 Magali Paquot

construction avoir + NOUN + à + INFINITIVE VERB. This construction is rela-

tively frequent (2.2 pmw in frWaC) and lexicalised in a restricted set of recurrent
sequences such as avoir un équilibre à trouver (“have a balance to find”), avoir un
choix à faire (“have a choice to make”), avoir un effort à faire (“have an effort to
make”), avoir un conseil à donner (“have advice to give”), and avoir un défi à rele-
ver (“have a challenge to face”).
(12) The parents too have a role to play in the education of their children: […]
(ICLE-FR)

Transfer of a French lexicalised infinitive construction is also at play in the use of

the bundle is to know whether in ICLE-FR. As illustrated in Examples (13) and
(14), the sequence appears in a larger pattern, i.e. the question/problem is to know
whether.
Strikingly, the French lexicogrammatical pattern la question/le problème est
de + VERB, is relatively frequent (1.6 pmw in frWaC) and savoir is the verb that is
most often found in the free slot.
(13) The question is to know whether these various agreements will contribute to
form a new nation or […]  (ICLE-FR)
(14) […] and the problem is to know whether reality will be as good as the dream.
(ICLE-FR)

6.3 Transfer of functions and discourse conventions

As hypothesized, some of the lexical bundles used idiosyncratically in French EFL

learner writing have equivalent forms that fulfil specific discourse functions in
French. Quite a few include the first person plural pronoun we (we can say, we
can wonder, we may wonder, we must not) or are part of longer patterns that often
involve a personal pronoun subject (speak of, be tempted to, not forget that, take
the example). First, the ICLE French sub-component is the only L1 learner corpus
where the lexical bundle be tempted to is found. French EFL learners mostly use
the bundle with a modal verb, and with subject pronouns (we, you) or the generic
noun people (cf. Examples (15) and (16)).
(15) After all, we may be tempted to believe that this process may at times have been
beneficial to the cultural standards that prevail in our society.  (ICLE-FR)
(16) Even the most honest people can be tempted to satisfy their craving for money.
(ICLE-FR)

The larger pattern PRONOUN/GENERIC NOUN (+ MODAL VERB) + be tempt-

ed to found in ICLE-FR most probably corresponds to two French introductory
 Lexical bundles and L1 transfer effects 405

phrases, i.e. nous sommes/serions tentés de and on est/serait tenté de (see below for
a discussion of EFL learners’ use of modal verbs). The pronouns nous and on are
very frequent in French academic writing. The first person plural pronoun nous
(“we”) is commonly used to involve the reader in the argument or guide them
through the research process. Such cases of inclusive we are often the subjects
of procedural verbs (nous avons procédé à, “we conducted”; nous avons repéré,
“we identified”) and metadiscursive verbs (e.g. nous aborderons, “we will discuss”;
nous montrerons, “we will show”) (Tutin 2010: 38). It may also be found when
an argumentative dimension is introduced with an opinion verb (e.g. penser,
“think”) or a verb of questioning (e.g. se demander, “wonder”). With these verbs,
however, the indefinite pronoun on is much more frequent,2 especially with the
modal verb pouvoir (e.g. on peut admettre, “we can admit”; on peut se demander,
“we may wonder”) (Tutin 2010: 23).
In the Scientext corpus, the verb parler (“speak”) is often used in introductory
phrases but it is actually found three times as often with the indefinite pronoun
on as with the personal pronoun subject nous (“we”) and is modified by pouvoir
(“can”) in 10% of the cases (Example (17)). In the CODIF, by contrast, the two
patterns are equally frequent and the more frequent use of nous may perhaps be
interpreted as a feature of novice writing. When compared to expert writers, for
example, French doctoral students have been reported to use more instances of
the first plural pronoun subject nous in their published research articles (Fløttum
& Thue Vold 2010: 46).
(17) Dans ce cas, on peut parler d’ellipse métonymique.  (Scientext)
(“In this case we can speak of metonymic ellipsis.”)

These findings help explain French EFL learners’ idiosyncratic use of the lexical
bundle speak of as an effect of their mother tongue. French learners often use the
verb with the first person plural pronoun we and a modal verb (cf. Example (18)),
a pattern that is not common in English academic writing (1.2 pmw in the aca-
demic component of the BNC).
(18) We cannot speak of a loss of national identity […]  (ICLE-FR)

French EFL learners’ overuse of lexical bundles including modal verbs is the re-
sult of a highly complex interplay of factors. This may, to some extent, simply be a
feature of novice writing: both L1 and L2 English student writers are reported to
rely extensively on modal verbs to convey statements with an appropriate degree
of doubt and certainty (Hyland & Milton 1997). L2 learners, however, appear to
depend far more heavily on these devices (e.g. Dagneaux 1995, Granger & Rayson
1998, Aijmer 2002, McKenny 2010) and to have incomplete mastery of the Eng-
lish modal system (Thewissen 2013).
406 Magali Paquot

The difficulties EFL learners face in using modal verbs may be reinforced by
interlingual factors as previously reported in the literature for other learner popu-
lations. Neff et al. (2003: 216), for example, attribute Spanish and Italian EFL learn-
ers’ erroneous use of the modal verb can in an epistemic sense to a mapping of the
more hypothetical meaning of the Spanish modal verb poder and the Italian modal
verb potere into their L2 English. An unnecessary use of modal verbs may also
be associated with transfer of writing conventions from the L1. Neff et al. (2004)
explain Spanish learners’ overuse of we must by the fact that the Spanish modal
verb deber can mean either must or should and that debemos (“we should” or “we
must”) + reporting verb is often used as a way of adding a further proposition to be
considered by the reader (e.g. debemos tener en cuenta, “we should/must take into
account”; debemos recorder, “we should/must remember”; debemos reconocer, “we
should/must recognize”; debemos aceptar, “we should/must accept”).
The data analysed for this study contained more examples of transfer of writ-
ing conventions. One of the most striking is French EFL learners’ overuse of we
can say (Example (19)), a lexical bundle which is not frequent in English academ-
ic writing (0.2 pmw) but is a translational equivalent of both nous pouvons dire
and on peut dire in French (0.3 and 1.3 pmw in Scientext). These two phrases are,
among other things, used to introduce the outcome of reasoning or put forward
a conclusion in French academic writing (Example (20)) and on peut dire is even
more frequent in French for general purposes (6 pmw in frWaC).
(19) In conclusion we can say that the birth of an economic nation would be
favourable.  (ICLE-FR)
(20) Dans cette optique, on peut dire qu’il existe des genres plus ou moins codi-
fiés…. (Scientext)
(“In this perspective, we can say that there are genres which are more or less
codified….”)

Similarly, the lexical bundle we may wonder is absent from the academic compo-
nent of the British National Corpus and the modal verb can is awkward in we can
wonder (Example (21)). However, both lexical bundles are used by French EFL
learners with the meaning and function of the French introductory phrase on
peut se demander (Example (22)).
(21) But we can wonder what a prison is and what its function is in our society.
(ICLE-FR)
(22) La question de la compositionnalité sémantique […] et on peut se demander si
elle présente un intérêt particulier pour le traitement automatique des langues.
 (Scientext)
(“The question of semantic compositionality […] and we can wonder whether
it is of particular significance for automatic language processing.”)
 Lexical bundles and L1 transfer effects 407

Learners also use modal verbs of obligation and necessity more than L1 writ-
ers and tend to adopt a more direct and emphatic style of persuasion (Hinkel
2002: 110). However, the use of must, should and have to seems to vary widely
across different L1 learner populations and reflects at least partly cultural conven-
tions (Hinkel 1995). The lexical bundle we must not sets the French learners apart
from all the other learner groups except the Swedes. It is used in sequences such
as we must not be pessimistic, we must not forget, we must not lose sight of, and we
must not neglect, to “influence the reader by emotional appeal” (Ädel 2006: 78),
persuade them that certain events are desirable, and present the writer and the
reader as a team in ICLE-FR (Example (23)).
(23) But we must not forget that books used to be written for only a small part of
the total population. (ICLE-FR)
(24) Cependant, il ne faut pas oublier que les données recueillies auprès des sta-
giaires sont uniquement déclaratives.  (Scientext)

The formally equivalent structure nous ne devons pas is not used in French aca-
demic writing; neither is the corresponding structure with indefinite on, i.e. on
ne doit pas. To express a negative obligation, French writers rather resort to the
impersonal structure il ne faut pas (Example (24)) but this discourse strategy is
more typical of general rather than academic language (20 pmw in frWaC vs. 1.2
in Scientext). It seems quite probable that French EFL learners’ use of the bundle
we must not is an attempt at expressing negative obligation and translating il
ne faut pas, a pattern which is also more frequent in French texts produced by
novice writers than expert writers. Interestingly, the larger bundle we must not
forget that is the only sequence that is repeated in ICLE-FR (5 occ.) and il ne faut
pas oublier que is also the only lexical bundle that is used repeatedly in Scientext.
As illustrated in Example (25), French EFL learners also made use of structures
involving the modal verb should as functional equivalent patterns to il ne faut
pas oublier.
(25) We should not forget that there are many sorts of criminals, ranging from the
accidental criminals and small fry to the hardened ones, the ones “beyond
redemption”.  (ICLE-FR)
The remaining occurrences of the lexical bundle not forget that in ICLE-FR are
used with the first plural imperative form let us, as are a majority of occurrences
of the bundle take the example. There is no lexically equivalent form to English
let us in French. Equivalence is however found at the morphological level as
French makes use of an inflectional suffix to mark the first imperative plural
form. Paquot (2008) compares the use of let us in ICLE-FR with that of first
person plural imperative verbs in CODIF and finds that the rhetorical and or-
ganisational functions fulfilled by let us in French EFL learner writing can be
408 Magali Paquot

paralleled with the very frequent use of first person plural imperative verbs in
French student writing to organize discourse and interact with the reader (see
also Paquot 2010: 189–191). Imperative forms that are repeated in ICLE-FR of-
ten have translational equivalents that are found in CODIF (e.g. let us take the
example of, “prenons l’exemple de”; let us consider, “considérons”; let us hope,
“espérons”; let us examine, “examinons”; let us take, “prenons”; let us (not/never)
forget, “oublions/n’oublions pas que”; let us think, “pensons”). This generalized
overuse of the first person plural imperative in EFL French learner writing as a
rhetorical strategy does not conform to English academic writing conventions
but rather to French academic style.
Lastly, the use of the lexical bundle I would say is also idiosyncratic in ICLE-
FR. As shown in Example (26), the bundle is most often used in phraseological
‘cascades’, “collocational patterns which extend from a node to a collocate and
on again to another node (in other words, chains of shared collocates)” (Gledhill
2000: 212), with an adverbial phrase such as in conclusion or to conclude to intro-
duce a conclusion.
(26) In conclusion, I would say that television has actually replaced religion in our
western civilization.  (ICLE-FR)

The French bundle je dirais appears in Scientext but it is not very frequent (0.38
pmw); the bundle, however, seems to be more typical of informal French and is
quite common in frWaC (5.3 pmw). The use of the first person pronoun je has
long been discouraged in French academic writing but it is used in disciplines
such as linguistics (cf. Fløttum 2003, Fløttum et al. 2006), where its use has in-
creased significantly between 1980 and 2000 in research articles (Gjesdal 2003,
quoted in Fløttum et al. 2006: 115). The lexical bundle je dirais is not found in
CODIF but it does occur with a relative frequency of 10 per 10,000 words in
the Corpus d’Apprenants du Français Langue Maternelle (CAFLaM), i.e. a newly
compiled corpus of argumentative texts produced by French-speaking first year
university students (Bolly 2008). Example (27) shows that EFL learners’ use of
longer sequences and phraseological cascades may also be transfer-related as the
bundle je dirais is also often introduced by discourse markers such as en conclu-
sion (“in conclusion”) or pour conclure (“to conclude”) in French-speaking novice
writing.
(27) En conclusion, je dirais qu’il existe un équilibre à trouver entre conformisme
et différence.  (CAFLaM)
(“In conclusion, I would say that a balance should be found between conform-
ism and difference.”)
 Lexical bundles and L1 transfer effects 409

It may thus be argued that French EFL learners’ use of I would say is the result of
a combination of L1-related factors, i.e. the relative and increasing tolerance of je
in the discipline they are studying, the high frequency of je dirais in general lan-
guage, and French-speaking novice writers’ reliance on phraseological cascades
including je dirais to conclude their argumentative essays.

6.4 Transfer of L1 frequency

Congruency is not a sufficient factor for cross-linguistic influence and distribu-

tional properties in the first language seem to play a significant role as well. An-
other way of approaching L1 frequency is to check whether the lexical bundles
used by the French learners have equivalent structures in another language rep-
resented in the ICLE corpus and if so, why these are not transferred into English
by the other learner population. Spanish is arguably a good candidate for this
purpose: French and Spanish are both Romance languages, and there are often
congruent sequences in Spanish for the French word combinations that were pin-
pointed in this study as responsible for transfer effects.
Spot-checks in the 100 million word Web corpus of Spanish available in the
Sketch Engine indeed strengthen the case for transfer of L1 frequency. While con-
siderar + como (“consider as”) exists in Spanish, for example, the verb is much
more frequently used without the preposition and the pattern considerar­+ AD-
JECTIVE is ten times as frequent as considerar como + ADJECTIVE (49.8 vs.
4.7 pmw). As a result, Spanish learners sometimes use the preposition as after the
verb consider (10 occurrences per 100,000 words in ICLE-SP) but they are much
less tempted to do so than their French counterparts. Similarly, el problema es de +
VERB (“the problem is to + VERB”) is extremely rare (0.008 pmw) in the Spanish
web-derived corpus and only two instances of the problem/question is to + VERB
are found in the Spanish learner corpus (Examples (28) and (29)).
(28) The second and greatest problem is to perform what they have learnt.
(ICLE-SP)
(29) Another important question is to analyze what can the goverment do about
this problem because they must fight to find a solution.  (ICLE-SP)

A comparison between French and Spanish also supports an L1 frequency-based

explanation for the overuse of the lexical bundle role to play in ICLE-FR and its
absence in ICLE-SP. There is a congruent form in Spanish (papel que desempe-
ñar) but it is rare (0.12 pmw), as is the larger construction tener + NP + que +
­INFINITIVE VERB (“have + NP + to + INFINITIVE VERB”) (0.4 pmw).
410 Magali Paquot

Spanish academic writing is characterized by a we-stance and as a result, EFL

Spanish learners also tend to overuse introductory phrases with we can (Neff et al.
2001). Patterns of overuse in ICLE-SP are however less marked when compared
to ICLE-FR. A likely explanation for this lies in the fact that first person plural
indicative forms compete with se impersonal passive phrases to perform similar
discourse functions in Spanish. If we look at the Spanish translational equiva-
lents of the English lexical bundles that are characterised by transfer of discourse
conventions in French learner writing, the prominent role of se impersonal pas-
sive structures appears clearly: podemos preguntarnos (“we may wonder”) and se
puede preguntar (“it can be wondered”) are equally frequent (0.4 and 0.3 pmw);
se puede decir (“it can be said”) is slightly more frequent than podemos decir (“we
can say”) (11.1 vs. 10.5) and se puede hablar de appears 251 times while podemos
hablar de (“we can speak of ”) only occurs twice in the corpus. The reason why
Spanish learners use fewer we can constructions than French learners is therefore
most probably because in Spanish they have the choice between first person plu-
ral structures and impersonal phrases with se. This mirrors Neff van Aertselaer’s
(2008) claim that Spanish expert and novice writers’ use of passive structures in
English probably reflects a transfer from Spanish discourse strategies.3
To sum up, congruency or formal equivalence is often pinpointed as the ex-
planatory factor for transfer effects. The results presented here show that congru-
ency is not sufficient in itself to ensure that a formal equivalent word combination
will be used in the foreign language. It is not because there is a formal equivalent of
an English lexical bundle in French and Spanish that the two learner populations
will use the English bundle in the same way. The frequency of word combinations
in the first language seems to play a crucial role: the more frequent a lexical bundle
is in the learners’ mother tongue, the more likely learners are to use its congruent
form in the foreign language. This seems to hold true for lexical bundles that ex-
emplify collocations (deeply rooted), colligations (consider as), syntactic structures
(NP + to-infinitive, e.g. role to play; NP + is + to-infinitive, e.g. the question is to
know whether) and discourse conventions (we-lexical bundles) alike.

7. Conclusion

Transfer effects on French learners’ use of 3-word sequences with lexical verbs
do not seem to generate obvious errors, at least at the intermediate to advanced
proficiency levels represented in the French component of the International
Corpus of Learner English. Rather, they are more visible in the learners’ selec-
tion of unmarked word combinations whose translational equivalents are deep-
ly entrenched in French speakers’ mental lexicon because these sequences are
 Lexical bundles and L1 transfer effects 411

particularly ­frequent or are directly anchored to important communicative or

metatextual functions. The word strings may be typical English sequences (e.g.
deeply rooted) or less favoured combinations (e.g. considered as). More interest-
ingly perhaps, they may be perfectly correct combinations in English but more
commonly used in less formal genres than that of academic writing: in the British
National Corpus, the lexical bundles I would say that, we can say, we must not, let
us not forget that, and let us take the example are generally more frequent in non-
academic texts and speech varieties including lectures and meetings.
All in all, results are in line with a usage-based view of language that recogniz-
es the active role that the L1 may play in the acquisition of a foreign language (e.g.
Bybee 2008). EFL learners bring knowledge of the L1 lexicon to the writing task
in the foreign language, including preferred collocations and lexicogrammatical
patterns of words, as well as their stylistic or register specificities, discourse func-
tions and frequency of use. As put by Hoey (2005),
As a word is acquired through encounters with it in speech and writing, it be-
comes cumulatively loaded with the contexts and co-texts in which it is encoun-
tered, and our knowledge of it includes the fact that it co-occurs with certain
other words in certain kinds of context. The same applies to word sequences built
out of these words; these too become loaded with the contexts and co-texts in
which they occur. (Hoey 2005: 8)

The transfer effects identified in this study are thus best described as “transfer of
primings” (Hoey 2005: 183). Mental primings for (at least frequent or core) L1
words and word strings are most probably superimposed on the primings for
their translation equivalent forms in the foreign language.
The direct pedagogical implication is that EFL teaching needs to counter the
default and sometimes misleading L1-related primings in EFL learners’ mental
lexicons. Awareness-raising activities focusing on similarities and differences
between the mother tongue and the foreign language are clearly needed. They
should not be restricted to “helping learners focus on errors typically committed
by learners from a particular L1” (Hegelheimer & Fisher 2006: 259) but should
also raise learners’ awareness of more subtle differences such as the collocational
preferences and distributional properties of similar words in the two languages.
This recommendation stands in sharp contrast to Bahns’s (1993: 56) claim that
collocations which are direct translation equivalents do not need to be taught.
Learners have no way of knowing which collocations are congruent in the mother
tongue and the foreign language; moreover, the differences between the colloca-
tions in L1 and L2 may lie in aspects of use rather than form or meaning.
Primings are also sensitive to the textual, generic and social contexts in
which a lexical item is encountered. Hoey (2005: 10) illustrates this with the word
412 Magali Paquot

r­ esearch, which is primed in the mind of academic language users to occur with
recent in academic discourse and news reports of research but is not primed to
occur in other text types or other contexts. A direct implication of Hoey’s theory
of lexical priming is that academic-like word combinations in the first language
cannot be assumed to be primed in the mental lexicon of novice native writers
who may have had little contact with academic texts in their L1. While many
of the French lexical bundles examined here proved to be relatively frequent in
French academic writing, some of them are indeed primed more strongly in gen-
eral language. This is particularly true of two sequences, i.e. on peut dire and il
ne faut pas, and calls for a more systematic deconstruction of the concept of L1
frequency in future research.
Many learner corpus-based studies, however, have fallen into the trap of
claiming L1 influence on the basis that the structure exists in the first language
without further investigation of L1 empirical data. In Douglas’s (2001: 451) words,
“the point here is not that these methods are faulty or that the interpretations are
invalid, but only that little or no evidence is provided for either quality [reliabil-
ity and validity]”. As shown in this study, formal similarity between L1 and L2
word combinations does not necessarily make the word combination in the first
language a strong candidate for transfer into the foreign language. Other factors
intervene and L1 frequency proved to contribute to transferability in a signifi-
cant way. The impact of L1 frequency is most apparent when different languages
are compared with the help of corpus data. As a consequence, this study also
brings support to the detection-based approach to transfer first outlined in Jarvis
(2010). The method is based on the premises that it is possible to identify the first
language of a learner on the basis of their use of specific features of the target
language and that these idiosyncrasies can serve as useful indicators of cross-
linguistic influence (Jarvis 2012).
Transfer effects were indeed pinpointed for twenty 3-word lexical bundles
which were further analysed as part of fifteen longer strings. This represents 7.3%
of all the 3-word sequences that appear at least 5 times in the French learner corpus
and c. 60% of the bundles that set the French learners apart from at least 5 other
learner populations. These figures are already quite high but they certainly under-
estimate the impact of the first language. The criterion according to which French
learners’ use of a given lexical bundle has to differ from that of at least five other
learner groups is very conservative. L1 influence may be obscured when the effects
of the mother tongue of different L1 learner populations coincide to produce the
same IL behaviour and this is certainly not a rare phenomenon (Jarvis 2000).
More generally, the study has also brought to light the considerable potential
of a corpus-driven approach to track L1 influence on learner language. Transfer
studies have often investigated “bits and pieces of learners’ language chosen for
 Lexical bundles and L1 transfer effects 413

analysis because they caught the researcher’s eye, seemed to exhibit some syste-
maticity, confirmed some intuition one had about SLA, or had been found in-
teresting in L1 acquisition” (Lightbown 1984: 245). As put by De Cock (2004),
the lexical bundle approach represents “corpus linguistic methodology at its most
heuristic, i.e. as a raw discovery procedure” (De Cock 2004: 227). Coupled with
Jarvis’s (2000) framework and appropriate statistical tests, it proves most useful to
extract fully automatically a number of word combinations that deserved further
analysis and consequently identify transfer effects that until now have been little
documented in the SLA literature. Lexical transfer has too often been narrowed
down to transfer of form/meaning mappings and the third aspect of word knowl-
edge, i.e. use, has rarely been investigated in all its complexity. Further research is
clearly needed. Lexical bundles of different sizes and built around different word
classes than just verbs should prove fascinating data types to start with.

Notes

* I would like to thank Sylviane Granger, Victoria Hasko and two anonymous reviewers for
their valuable comments and constructive suggestions for improvement. I acknowledge the
financial support of the Fonds de la Recherche Scientifique (FNRS).

1. The use of parametric tests may be criticized as the data used in this study is not normally
distributed. According to Howell (1997), those who argue in favour of using parametric tests
“argue, however, that the assumptions normally cited as being required of parametric tests are
overly restrictive in practice and that the parametric tests are remarkably unaffected by viola-
tions of distribution assumptions” (Howell 1997: 646, see also Rietveld et al. 2004: 360). More-
over, parametric tests are said to be more powerful than non-parametric tests: they require
fewer observations than do non-parametric tests and are more likely “to lead to rejection of a
false null hypothesis” (Howell 1997: 646) than are their corresponding non-parametric tests.
This advantage seems to be maintained “even when the distribution assumptions are violated
to a moderate degree” (ibid).

2. The French indefinite pronoun on is much more frequent and stylistically very different
from the English one: it can refer to one or more people, be substituted for all personal pro-
nouns and “has an unclear enunciative status (i.e. relation to speaker or locator and receiver)”
(Fløttum et al. 2006: 113).

3. Paquot (2008) has also shown that the distribution of let us in the interlanguage of French,
Spanish and Dutch learners parallels that of first person plural imperative structures in the
three languages.
414 Magali Paquot

References

Ädel, A. 2006. Metadiscourse in L1 and L2 English. Amsterdam: John Benjamins.

Ädel, A. 2008. “Involvement features in writing: Do time and interaction trump register aware-
ness”. In B. M. Diez-Bedmar, G. Gilquin & S. Papp (Eds.), Linking up Contrastive and
Learner Corpus Research. Amsterdam/New York: Rodopi, 35–53.
Aijmer, K. 2002. “Modality in advanced Swedish learners’ written interlanguage”. In S. Granger,
J. Hung & S. Petch-Tyson (Eds.), Computer Learner Corpora, Second Language Acquisition
and Foreign Language Teaching. Amsterdam: John Benjamins, 55–76.
Allen, D. 2011. “Lexical bundles in learner writing: An analysis of formulaic language in the
ALESS learner corpus”. Komaba Journal of English Education, 1, 105–127
Altenberg, B. 1998. “On the phraseology of spoken English: The evidence of recurrent word-
combinations”. In A. P. Cowie (Ed.), Phraseology: Theory, Analysis, and Applications.
­Oxford: Oxford University Press, 101–122.
Bahns, J. 1993. “Lexical collocations: A contrastive view”. ELT Journal, 47 (1), 56–63.
Biber, D., Conrad, S. & Cortes, V. 2003. “Lexical bundles in speech and writing: An initial
taxonomy”. In A. Wilson, P. Rayson & T. McEnery (Eds.), Corpus Linguistics by the Lune:
A Festschrift for Geoffrey Leech. Frankfurt: Peter Lang, 71–92.
Biber, D., Johansson, S., Leech, G., Conrad, S. & Finegan, E. 1999. Longman Grammar of Spoken
and Written English. Harlow: Longman.
Bolly, C. 2008. Corpus Argumentatif en Français Langue Maternelle (CAFLaM). Les Unités Phra-
séologiques: Un Phénomène Linguistique Complexe?, Unpublished Ph.D Thesis, Louvain-
la-Neuve: Université catholique de Louvain.
Bybee, J. 2008. “Usage-based grammar and second language acquisition”. In P. Robinson &
N. Ellis (Eds.), Handbook of Cognitive Linguistics and Second Language Acquisition.
­London: Routledge, 216–236.
Cortes, V. 2004. “Lexical bundles in published and student disciplinary writing: Examples from
history and biology”. English for Specific Purposes, 23 (4), 397–423.
Dagneaux, E. 1995. Expressions of Epistemic Modality in Native and Non-Native Essay-Writing.
MA dissertation. Louvain-la-Neuve: Université catholique de Louvain.
De Cock, S. 2003. Recurrent Sequences of Words in Native Speaker and Advanced Learner Spo-
ken and Written English: A Corpus-Driven Approach. Unpublished PhD thesis. Louvain-la-
Neuve: Université catholique de Louvain.
De Cock, S. 2004. “Preferred sequences of words in NS and NNS speech”. Belgian Journal of
English Language and Literatures (BELL), New Series, 2, 225–246.
Douglas, D. 2001. “Performance consistency in second language acquisition and language test-
ing”. Second Language Research, 17 (4), 442–456.
Fløttum, K. 2003. “Personal English, indefinite French and plural Norwegian scientific authors?
Pronominal author manifestation in research articles”. Norsk Lingvistisk Tidsskrift, 21 (1),
21–55.
Fløttum, K., Dahl, T. & Kinn, T. 2006. Academic Voices – Across Languages and Disciplines.
Amsterdam: John Benjamins.
Fløttum, K. & Thue Vold, E. 2010. “L’éthos auto-attribué d’auteurs-doctorants dans le discourse
scientifique”. Lidil, Revue de Linguistique et de Didactique des Langues, 41, 41–58. Available
at: http://lidil.revues.org/index3006.html (accessed June 2013).
 Lexical bundles and L1 transfer effects 415

Gjesdal, A. M. 2003. L’emploi du Pronom ‘on’ dans les Articles de Recherche. Une étude diachro-
nique et qualitative. MA dissertation. Bergen: University of Bergen.
Gledhill, C. 2000. Collocations in Science Writing. Tuebingen: Gunter Narr Verlag.
Granger, S. 1998. “Prefabricated patterns in advanced EFL writing: Collocations and formulae”.
In A. Cowie (Ed.), Phraseology: Theory, Analysis and Applications. Oxford: Oxford Univer-
sity Press, 145–160.
Granger S., Dagneaux E. & Meunier F. 2002. The International Corpus of Learner English.
Handbook and CD-ROM. Louvain-la-Neuve: Presses Universitaires de Louvain.
Granger, S., Dagneaux, E., Meunier, F. & Paquot, M. 2009. The International Corpus of Learner
English. Handbook and CD-ROM (Version 2). Louvain-la-Neuve: Presses Universitaires
de Louvain.
Granger, S. & Rayson, P. 1998. “Automatic profiling of learner texts”. In S. Granger (Ed.), Learn-
er English on Computer. London/New York: Addison Wesley Longman, 119–131.
Groom, N. 2009. “Effects of second language immersion on second language collocational
­development”. In A. Barfield & H. Gyllstad (Eds.), Researching Collocations in Another
Language. Basingstoke: Palgrave Macmillan, 21–33.
Hegelheimer, V. & Fisher, D. 2006. “Grammar, writing, and technology: A sample technol-
ogy-supported approach to teaching grammar and improving writing for ESL learners”.
­CALICO Journal, 23 (2), 257–279.
Hinkel, E. 1995. “The use of modal verbs as a reflection of cultural values”. TESOL Quarterly,
29 (2), 325–243.
Hinkel, E. 2002. Second Language Writers’ Text: Linguistic and Rhetorical Features. London:
Lawrence Erlbaum Associates.
Hoey, M. 2005. Lexical Priming: A New Theory of Words and Language. London/New York:
Routledge.
Howell, D. 1997. Statistical Methods for Psychology. Belmont: Wadsworth.
Hyland, K. & Milton, J. 1997. “Qualifications and certainty in L1 and L2 students’ writing”.
Journal of Second Language Writing, 6 (2), 183–205.
Jarvis, S. 2000. “Methodological rigor in the study of transfer: Identifying L1 influence in the
interlanguage lexicon”. Language Learning, 50 (2), 245–309.
Jarvis, S. 2010. “Comparison-based and detection-based approaches to transfer research”.
In L. Roberts, M. Howard, M. Ó Laoire & D. Singleton (Eds.), EUROSLA Yearbook 10.
­Amsterdam: John Benjamins, 169–192.
Jarvis, S. 2012. “The detection-based approach: An overview”. In S. Jarvis & S. Crossley (Eds.),
Approaching Language Transfer through Text Classification: Explorations in the Detection-
Based Approach. Bristol: Multilingual Matters, 1–33.
Juknevičienė, R. 2009. “Lexical bundles in learner language: Lithuanian learners vs. native
speakers”. Kalbotyra 61 (3), 61–72.
Kellerman, E. 1978. “Giving learners a break: Native language intuitions as a source of predic-
tions about transferability”. Working Papers on Bilingualism, 15, 59–92.
Kilgarriff, A. & Kosem, I. 2012. “Corpus tools for lexicographers”. In S. Granger & M. Paquot
(Eds.), Electronic Lexicography. Oxford University Press, 31–56.
Kroll, B. 1990. “What does time buy? ESL student performance on home vs. class composi-
tions”. In B. Kroll (Ed.), Second Language Writing. Cambridge: Cambridge University
Press, 140–154.
416 Magali Paquot

Lightbown, P. M. 1984. “The relationship between theory and method in second-language-

acquisition­research”. In A. Davies, C. Criper & A. Howatt (Eds.), Interlanguage. Edin-
burgh: Edinburgh University Press, 241–252.
McKenny, J. 2010. A Corpus Study of the Phraseology of Written Argumentative English. Saar-
brücken: Lambert Academic Publishing.
Neff, J., Ballesteros, F., Dafouz, E., Martínez, F. & Rica, J. P. 2004. “The expression of writer
stance in native and non-native argumentative texts”. In R. Facchinetti & F. Palmer (Eds.),
English Modality in Perspective. Frankfurt am Main: Peter Lang, 141–161.
Neff, J., Dafouz, E., Herrera, H., Martínez, F., Rica, J. P., Diez, M., Prieto, R. & Sancho, C. 2003.
“Contrasting learner corpora: The use of modal and reporting verbs in the expression of
writer stance”. In S. Granger & S. Petch-Tyson (Eds.), Extending the Scope of Corpus-Based
Research. New Applications, New Challenges. Amsterdam/New York: Rodopi, 211–230.
Neff, J., Martínez, F. & Rica, J. P. 2001. “A contrastive study of qualification devices in NS and
NNS argumentative texts in English”. In ERIC Clearing House on Language and Linguis-
tics (ERIC Document Reproduction Service, ED 465301). Washington, D.C.: Educational
Resource Information Center, U.S. Department of Education.
Neff van Aertselaer, J. 2008. “Contrasting English-Spanish interpersonal discourse phrases:
A corpus study”. In F. Meunier & S. Granger (Eds.), Phraseology in Foreign Language
Learning and Teaching. Amsterdam: Benjamins, 85–100.
Paquot, M. 2008. “Exemplification in learner writing: A cross-linguistic perspective”. In
F. Meunier & S. Granger (Eds.), Phraseology in Foreign Language Learning and Teaching.
Amsterdam: Benjamins, 101–119.
Paquot, M. 2010. Academic Vocabulary in Learner Writing: From Extraction to Analysis. Lon-
don/New York: Continuum.
Paquot, M. forthcoming. “Phraseology and lexicography”. In D. Biber & R. Reppen (Eds.), The
Cambridge Handbook of Corpus Linguistics. Cambridge: Cambridge University Press.
Paquot, M. & Granger, S. 2012. “Formulaic Language in Learner Corpora”. Annual Review of
Applied Linguistics, 32, 130–149.
R Core Team. 2012: online. R: A Language and Environment for Statistical Computing. Available
at: http://www.R-project.org (accessed June 2013).
Rayson, P. 2003. Matrix: A Statistical Method and Software Tool for Linguistic Analysis through
Corpus Comparison. Unpublished PhD thesis, Lancaster University.
Reppen, R. 2009. “Exploring L1 and L2 writing development through collocations: A corpus-
based look”. In A. Barfield & H. Gyllstad (Eds.), Researching Collocations in Another Lan-
guage. Basingstoke: Palgrave Macmillan, 49–59.
Rica, J. P. 2010. “Corpus analysis and phraseology: Transfer of multi-word units”. Linguistics
and the Human Sciences, 6, 321–343.
Rietveld, T., van Hout, R. & Ernestus, M. 2004. “Pitfalls in corpus research”. Computers and the
Humanities, 38 (4), 343–362.
Ringbom, H. 1987. The Role of the First Language in Foreign Language Learning. Clevedon/
Philadelphia: Multilingual Matters.
Scott, M. 2008. WordSmith Tools version 5. Liverpool: Lexical Analysis Software.
Selinker L. 1992. Rediscovering Interlanguage. London/New York: Longman.
Scientext corpus. Available at: http://scientext.msh-alpes.fr/ (accessed June 2013)
Thewissen, J. 2013. “Capturing L2 accuracy developmental patterns: Insights from an error-
tagged EFL learner corpus”. Modern Language Journal, 97 (Suppl. 1), 77–101.
 Lexical bundles and L1 transfer effects 417

Tutin, A. 2010. “Dans cet article, nous souhaitons montrer que… Lexique verbal et position-
nement de l’auteur dans les articles en sciences humaines. Enonciation et rhétorique dans
l’écrit scientifique”. Lidil, Revue de Linguistique et de didactique des langues, 41, 15–40.
Available at: http://lidil.revues.org/index3040.html (accessed June 2013)

Author’s address
Magali Paquot
Centre for English Corpus Linguistics
Institut Langage et Communication
Université catholique de Louvain
Place Blaise Pascal 1, bte L3.03.31
1348, Louvain-la-Neuve
Belgium
magali.paquot@uclouvain.be