This research investigates whether there is a relationship between the use of three- and four-word Lexical bundles and language proficiency. The study conducts both quantitative and qualitative analyses to see whether learners from different CEFR levels groups exhibit the same behaviour in the use of Lexical bundles. Therefore, in the first stage, it compares between two different levels B2 and C1 in terms of frequency, structures and functions of Lexical bundles to give an overview of some of the linguistic features to differentiate between the levels. In the second stage, a longitudinal study investigated the development of ESL learners use of Lexical bundles across the levels to give a picture of the increases of the proficiency levels. A major finding from the analysis shows that generally, ESL learners favoured using more signalling bundles in their writing, three-word bundles turned out to be the most frequent bundles in ESL sub-corpora. Moreover, significant progress has been found in the variability of the structures and functions of Lexical bundles, C1 writers are found to have used various structures and functions as professional writers in their academic writing. For the development of Lexical bundles in relation to the CEFR levels, the findings clearly indicate that there is no significant relationship between the increased use of Lexical bundles and academic performance. However, multiple regression analysis revealed that there is a direct proportionality between variations of the use of Lexical bundles and the CEFR levels, as (C1) students act as professional writers and used variant structures and functions than (B2) Students.
Lexical bundles are word combinations that can be defined as continuous multiword
sequences that recur frequently to satisfy specified frequency and dispersion
thresholds; for example, occurring at least 20-40 times per million words in five texts, or in
at least 10% of texts [
Copyright ©2020 for this paper by its authors.
Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
of fluency in linguistic production" [
The bundles are important elements by which to measure learners’ language
development, and both native and non-native speakers indicate their language proficiency
by using lexical bundles in their academic writing; the absence of these bundles
signals a novice writer. This idea has been supported with empirical evidence showing
that the competent use of lexical bundles contributes to fluent language production.
[
However, many studies have investigated the use of lexical bundles by non-native
speakers of different levels across a range of registers and academic disciplines.
According to the previous studies, although there has been an increase in the use of
lexical bundles by non-native speakers, their use is limited to specific bundles causing
them to overuse some expressions compared to others, making their writing appear
non-native [
To the best of the researcher’s knowledge, while most previous studies have paid considerable attention to the use of lexical bundles across different registers and a number of disciplines, little research has been done to investigate whether learners from different proficiency level groups exhibit the same behaviour in their use (or not) of lexical bundles. This research investigates whether there is a relationship between the use of three- and four-word lexical bundles and language competence. The study utilises both quantitative and qualitative analyses to determine whether learners from different CEFR (Common European Framework of Reference) level groups exhibit the same behaviour in the use of lexical bundles. Additionally, this study examines the development of lexical bundles across proficiency levels. Specifically, it compares between two different levels, B2 and C1, in terms of the frequency, structures, and functions of lexical bundles to give an overview of some of the linguistic features that differentiate between the levels. This study addresses the following questions: – What are the most frequently used three- and four-word lexical bundles in the
B2 and C1 sub-corpora? – –What does a keyness analysis reveal about lexical bundles identified in theB2 and C1 sub-corpora? – –How do lexical bundles in the B2 sub-corpus differ from C1 in terms of structure and function? – Is there any growth in the lexical bundles identified in the study between B2and
C1 learners? 1 1.1
This study is first interested in the relationship between the use of lexical bundles and
academic performance; thus, the author compared B2 and C1 sub-corpora (for the
frequency, structures, and functions of lexical bundle) of ESL learners and then
compared them with a reference corpus. The data used came from written essays
equivalent to the IELTS test in terms of the title, written by 42intermediate and advanced
ESL learners from different mother tongue who have studied in the UK who
contributing 130 essays. These learners write academic essays to test their progress and place
them at new levels if they meet the requirements at the English Language Centre
(ELC). Only argumentative or expository pieces written by L2 learners were chosen
for the sub-corpora. The decision to use learners’ sub-corpora was based on the
assumption that they are useful to explore and identify the similarities and differences in
the use of recurrent word combinations across L2 proficiencies of “actual language in
use” [
The second stage of the study consisted of second language development research, which compares learners’ language across proficiency levels (CEFR levels). A longitudinal study investigated the development over three months of two ESL learners use of lexical bundles in their academic essays across the levels to trace the increases in proficiency level. The participants were two ESL students (one male and one female) at the upper intermediate level that moved to advanced level after two months who contributed 36 essays to be used for the investigation. 1.2
The procedure for determining the CEFR level originates from the manual for
Relating Language Examinations to the CEFR for Languages [
The reference corpus used in this study was taken from the British Academic Writing
English (BAWE) corpus, which contains 2,761 texts of proficient assessed academic
works written at universities in the UK (6,506,995 words), ranging in length from
around 500 words to approximately 5,000 words. However, since the target
subcorpora used argumentative essays (equivalent to the IELTS task 2)written by ESL
learners, it was decided to use BAWE (linguistics and English disciplines) as a
reference corpus to avoid skewing the sample heavily toward one discipline. These two
disciplines are big enough to be used as a reference corpus as well as include relative
language that ESL learners use in their academic essays, using other disciplines such
as Philosophy or Biochemistry might effect the results. Therefore, linguistics and
English disciplines are suited to the goal of this study as they provide a wide range of
language representative of ESL students writing in an authentic academic context.
As stated by Leech [
The analysis used to answer the above research questions was carried out using
Wordsmith computer software. [
After retrieving the corpus and applying the frequency and distribution criteria,
Wordsmith provided lists of three- and four-word lexical bundles for both
subcorpora. Hyponyms were checked and cleared from all the bundles found. In order to
narrow down the included lexical bundles, all content-based bundles were discarded,
as they do not reflect the use of general academic language, such as The United
Kingdom or The University of Liverpool. In addition, overlap-ping bundles were
combined as one bundle to avoid duplication in the counting of high-frequency bundles.
For example, the bundle can be used to and it can be used to were counted as one
bundle, adding a word between the brackets such as, (it) can be used to [
The results revealed that the B2 sub-corpus accounted for 102 (type) three- and
fourword lexical bundles, which occurred 458 times, making up 9.2 % of the total number
of words in the sub-corpus. The C1 essays contained 45 (type) three-and four-word
lexical bundles, which occurred 204 times in the sub-corpus and made up 5 % of the
total words in the sub-corpus. What stands out is that the lower-level students used a
larger stock of lexical bundles than the higher-level students as presented in table 2.
In addition, the three-word bundles were revealed to be the most common bundles at
both levels. Therefore, it can be concluded that ESL learners have a tendency to
employ a higher number of three-word than four-word bundles with an increase in
lowlevel students. A possible explanation might be related to the complexity of their
production, which language learners avoid in their writing, as it requires more effort and
time for students to produce longer sequences than shorter ones. The result was not
surprising; Biber et al. [
Surprisingly, few of the most frequent bundles in the BAWE corpus were found in the
ESL learners’ corpora: only eight out of the 50 most frequent lexical bundles in B2
and C1 sub-corpora were identified in the BAWE corpus. According to that, although
the B2 level students used more lexical bundles than C1 students, certain bundles
were new and used by only a few learners with repeated the same bundle more than
once in their essays. For example, the bundle on the other was identified 19 times in
the B2 sub-corpora (although one student used it three times in one text). A possible
explanation for this might be that ESL learners tend to use certain lexical bundles
more frequently to reflect a high level of formality and demonstrate their language
competence; alternatively, they may still be in the process of learning additional
lexical bundles. This result conflicts with those presented by Chen and Baker [
To determine the ’key’ bundles in B2 and C1,WordSmithsoftware was used to generate a list of ‘key’ bundles that occur unusually frequently in the target sub-corpora when compared with a reference corpus (i.e. BAWE) by means of statistical tests (e.g. chi-square or log-likelihood). A ‘keyness’ value is given for each bundle that has statistically significant, the higher the keyness score, the more the key bundle is statistically significant. The WordSmith software provides a list of lexical bundles which are positively and negatively key. However, as the main focus only on the positive keyness, the WordSmith tool was sitting to ignore all the negative results as provided in Table 3 and 4.
The results provided some evidence for the common assertion in the previous
studies that ESL learners favour particular bundles and overuse them in their writing [
Structurally, Biber et al. [
The results showed that EFL learners used more phrase bundles than clausal
bundles in their writing. More specifically, verb-based bundles were the most frequent
three- and four-word bundles found in the B2 and C1 sub-corpora. Among the two
CEFR levels, the C1 level had the highest proportions of verb-based bundles, at
53.4%, while the B2 level had a lower percentage, 40.5%. These results conflict with
the idea of the rarity of verb-based bundles in academic discourse [
It can be concluded that the three groups employed a different percentage of most of the structural sub-categories, except the ‘preposition-based’ category. The chisquare test results of the correlational analysis revealed a significant difference among the corpora. The standardised residuals in a chi-square contingency table for the distribution of structural types revealed that greater differences occurred in the ‘verbbased’, ‘noun-based’ and ‘other’ categories. For instance, the test shows that C1 writers overused verb-based bundles compared to B2writers, which supports the idea that C1 students rely more on spoken language in their writing. In regard to Noun-based, it appeared that B2 students underused these bundles in their writing. On the other hand, B2 writers overused ‘other’ bundles not related to any sub-category (e.g., as adverbial or modal bundles).
As the standardised residuals in a chi-square did not show any significant difference in the use of ‘prepositions-based’ bundles, the result reflects the similarity of the proportion of preposition-based bundles in both levels and BAWE, at 15% of total bundles. The ‘PP expressions’ subcategory is typically used to show the logical relationship between prepositional elements, which means that EFL learners could use this type of lexical bundle to link between the ideas of the argumentation. The difference in frequency of the use of different structural categories across the levels suggests that as their level increases the students are able to recognise and use the adverbial meaning of the bundles.
Functionally, Hyland’s taxonomy was adopted, since the data used in the this study
was mainly academic prose (see Table 6) [
In order to be able to classify bundles into the correct sub-categories, it was
important to look at the concordance line to see the bundles in their context and to tackle
the issue of multi-functionality of the target bundles. There was similarity in the use
of functional categories between the levels. The most frequent functions of the
identified bundles across the levels were research-oriented followed by participant-oriented,
and then text-oriented. The increase in use of research-oriented bundles in the B2 and
C1 sub-corpora might be due to the fact that in argumentative essays, students need to
describe various aspects and provide different justifications of their ideas to the
reader. Bundles of this function accounted for more than 40% of all bundles identified in
the corpora. This result is similar to previous studies, which have found that academic
writing is dominated by research-oriented bundles over other categories [
Procedure Quantification Description Topic Total Transition signals Regulative signals Structuring signals Framing signals Total text-oriented Stance features Engagement features Total participantoriented 10 types In the comparison between the levels, it was seen that B2 writers used researchoriented bundles more often than C1 writers. By contrast, C1 writers employed more text-based and participant-based bundles than B2 writers. The study found a direct proportionality between the percentage of text-oriented and participant-oriented bundles as the level increased. In addition, chi-square unstandardized residuals statistical methods were used in the analysis of structural and functional type, to further support arguments in this study. Statistically, the study has failed to demonstrate any statistically significant difference in functional distributions between the levels. 2.4
For the development of lexical bundles across the levels in the second stage, the results were similar to the first stage, where three-word bundles were found to be the most frequent bundles in the EFL sub-corpora. However, the results provide some evidence that suggests there may be development of the use of lexical bundles across the levels, but not to a statistically significant degree. This might be due to the number of collected essays that made up the sub-corpus and the short period of time the learners were tracked over.
Structurally, there was much variability in terms of the structures and functions of lexical bundles across the levels. High-level EFL learners used a greater variety of structures and functions in their writing than low-level learners. The results showed that there were distinctive differences in terms of the greater use of ‘noun–based’, ‘preposition-based’ and ‘verb-based’ bundles by both levels and in the reference corpus. It should be noted that, across the four categories, the percentage of three structural categories in the C1 level seem closer to those in the reference corpus than did those at the B2 level. The B2 levels students used six out of 12 subcategories, while C1 and reference corpus students used 10 out of 12 subcategories. The chi-square revealed significant differences between the levels and the reference corpus, and the standardised residuals (R), which compared observed and expected counts in each cell, showed that greater differences occurred in all the categories, as the C1 and reference corpus used significantly more ‘verb-based’ and ‘noun-based’ bundles and fewer ‘preposition-based’ bundles than B2, except in the ‘other’ category, which did not show any significant difference between the levels, which reflected the frequent use of bundles such as I want to, a lot of, the fact that (the), and the development of.
By contrast, the overuse of preposition-based bundles in the B2 sub-corpus reflected the frequent use of bundles such as in order to and as well as. Functionally, while the density of text-oriented bundles appeared almost identical in the B2 sub-corpus, the use of research-oriented and participant-oriented bundles in the C1 sub-corpus seems to be more aligned with the reference corpus.
Further analysis of the functional sub-categories revealed the same results as for the structural sub-categories: the C1 level seemed closer to the reference corpus than the B2 level. The result of the chi-square test revealed a significant difference among the three groups. The standardised residuals (R), which compare observed and expected counts in each cell, showed that the greatest differences between the groups occurred in the ‘text-oriented’ and ‘participant-oriented ‘categories, as the C1 and reference corpora used significantly more participant-oriented bundles but fewer textoriented than the B2 level. This might be due to the wide range of topics that argumentative and expository essays covered. 3 3.1
A major finding from the analysis was that, generally, EFL learners favoured using more signaling bundles in their writing; three-word bundles were found to be the most frequent bundles in EFL sub-corpora. Moreover, significant progress was identified in terms of the variability of the structures and functions of lexical bundles, C1 writers were found to have used various structures and functions as professional writers in their academic writing. In terms of the development of lexical bundles in relation to CEFR level, the findings clearly indicated that there was no significant relationship between the increased use of lexical bundles and academic performance. However, multiple regression analysis revealed that there is a direct proportionality between variations in the use of lexical bundles and CEFR level, as higher-level students (C1) acted as professional writers and used more variant structures and functions than lower-level students (B2).
The results of this study show that there are specific lexical bundles that maybe considered to be the building blocks of ESL learners academic essays. These results might be interesting for English language teachers and instructors because it provides insights into the ESL learners community preferences in academic writing. 3.2
Like many other studies, the present investigation has its limitations. One of which is
the small corpora size. However, small corpora size can produce more lexical bundles
than the big corpus. [