Unlocking the Mysteries of Academic Writing: A Corpus-based Analysis of Lexical Bundles in L2 English for Engineering Students

Abstract

This study examines the use of lexical bundles in reports produced by university-level L2 English writers in civil and environmental engineering (CEE). The corpus consists of 119 reports, and the study compares the use of different types of lexical bundles (3-word, 4-word, and 5-word) in high-scoring and low-scoring papers. The results show that high-scoring papers use a wider range of lexical bundles, and that the use of lexical bundles varies by discipline. The study also identifies specific features of the lexical bundles that contribute to score differences. The findings suggest that teaching lexical bundles and vocabulary can improve the academic writing of engineering students. The paper concludes by calling for further research on the impact of formulaic expressions on attainment scores in L2 writing.

Plain language summary

A corpus analysis of lexical bundles in L2 English for engineering students

Keywords

lexical bundles L2 writing civil and environmental engineering academic writing vocabulary instruction

Introduction

Effective writing skills are crucial for academic success and contributing to scholarly discourse (Kellogg & Raulerson, 2007). Lexical knowledge is a key aspect of second language (L2) writing that indexes the students’ development as writers. For instance, research has shown a positive correlation between L2 writing proficiency and lexical variation, such as type-token ratio. Essays containing a wide range of lexical items tend to receive higher scores. Additionally, research has also shown that lexical errors have an impact on the evaluations of papers since the raters’ evaluations tend to be based primarily on the extent to which the vocabulary was utilized accurately or the deployment of a gamut of lexical items within essays. This suggests that lexical errors play a key role in how essays are evaluated. Indeed, it would appear that the extent to which the writer uses an advanced vocabulary is a prime indicator of his or her writing ability. Lexical items which are deployed less frequently are considered as more advanced, while words used more frequently are perceived as being less sophisticated (Kyle & Crossley, 2015).

Lexical bundles have become a significant topic in language research, particularly in relation to L2 writing and English for Academic Purposes (EAP) studies since Biber et al. (1999) introduced the concept. These bundles refer to frequent and repetitive lexical sequences in a register, which may even form complete structural units (Biber et al., 1999). Meeting specific frequency and dispersion thresholds (e.g., appearing 20 times per million words in five or more texts), they are understood as recurring word sequences that reflect conventional pragmatic and discourse roles used in specific settings (Biber & Barbieri, 2007; Hyland, 2008a), according to a frequency-based approach. Biber et al. (2004) even suggest that such sequences represent “fundamental building blocks of speech” that transcend the boundaries between lexis and syntax.

It is important to emphasize that lexical bundles play a structural and functional role in discourse management. A number of taxonomies have been suggested as a means of assisting researchers in gaining a better understanding of those structures and the functions they serve. Certain words and lexical bundles are required to be used in order to communicate effectively due to the varied situational contexts, production circumstances, and communicative objectives. This linguistic variation is captured in terms of the versatility in the use of lexical items in different situational contexts. This versatility is a result of the constraints that different situations impose on the speakers or authors of the language (Biber, 2006; Biber & Conrad, 2019). According to Nattinger and DeCarrico’s research (1992), lexical phrases are defined as “chunks of language of varying lengths” that exhibit distinct formal, functional, and statistical characteristics, and numerous lexical bundles have been found to be relevant across various engineering disciplines. Developing proficiency in these bundles can help civil engineering students communicate effectively with professionals from other fields, fostering collaboration, and interdisciplinary learning. Moreover, proficiency in lexical bundles can significantly help civil engineering students improve their language skills, particularly in the context of their field. Therefore, this study merely focuses on the significance of learning lexical bundles for engineering students as students can expand their technical vocabulary, enabling them to communicate more effectively with peers, instructors, and professionals in the field.

Earlier corpus research has paid substantial attention to how lexical bundles are used. For example, a study by Chen and Baker (2016) examined 585 argumentative or expository texts from the Longman corpus, analyzing “criterial discourse features in L2 writing” within essays produced by Chinese CEFR B1, B2, and A-level learners. They identified that the discourse of learner writing at lower levels has a greater tendency to “share more features with conversation, whereas the discourse of learner writing at higher levels is closest” to that of academic prose (Chen & Baker, 2016). Huang (2015) used a corpus of argumentative essays to compare the senior and junior levels essays produced by learners at Chinese universities, concluding that while senior year learners used lexical bundles more frequently than younger peers, these items were not used more accurately than junior level learners. In other research, investigating writing assignments from EAP and ESP courses in years 1 and 2 at Chinese universities Ruan (2017) found that higher level learners used varied lexical bundles in their writing. Investigating the “frequency, function, and degree of fixedness” of lexical bundles in TOEFL iBT essays across levels of proficiency, Staples et al. (2013) concluded that while low proficiency writers used more bundles, these tended to run parallel to those in the essay prompts. Vo (2019) investigated argumentative/expository essays from English placement tests and from level 1, 2, and 3 ESP learners. Vo’s (2019) findings concluded that “those learners with a higher level of proficiency used a greater number of types, tokens, and word families than those learners with a lesser level of proficiency.”Kim and Kessler (2022, p. 1) examined the “relationship between learners’ bundle use and holistic writing quality” within 120 essays produced by L2 writers at a Chinese university. Other research investigating the use of lexical bundles (LBs) includes studies carried out by several researchers (e.g., Pan & Liu, 2019; Qi & Pan, 2020). Analyzing a corpus of master level theses and research articles within applied linguistics, Pan and Liu (2019) compared LB usage in the writings of native English academic writers and non-native L2 writers. The researchers found structural and functional variations in the writings produced by the two groups, highlighting that these could be attributed to the influence of L1 and expertise.

Generally, lexical bundles reflect a very authentic aspect of users’ communication experiences. There is a phraseological tendency in the use of language (Carter & Sinclair, 2004), known as the “idiom principle,” whereby speakers and writers conventionally co-select vocabulary. This tendency can be attributed to phraseology. As language users, we tend to be intellectually “primed” to anticipate co-occurring terms as we have encountered them in multiple pairings (Hoey, 2012, p. 5). Therefore, lexical bundles function as processing shortcuts that make language more predictable for readers, enabling pragmatically efficient and effective communication. They shorten processing time in academic discourse by linking concepts together and allowing writers to deploy recognized patterns that help readers navigate the text. Additionally, because they indicate the correct application of a disciplinary code, the accurate use of lexical bundles enables authors to demonstrate solidarity with their peers (Cortes, 2006) and develop a competent voice in communicating disciplinary knowledge (Hyland, 2008a).

Recent research has identified “criteria traits” pertaining to lexis and grammar for the Common European Framework of Reference (CEFR, Council of Europe 2001), including the English Profile project headed by Cambridge University researchers (for example, Hawkins & Buttery, 2010; Hawkins & Filipović, 2012). However, researchers have paid limited attention to formulaic language across the span of CEFR levels. Further, existing research has not investigated L2 English learners’ use of lexical bundles in terms of writing quality. Further, existing research has tended to focus on the characteristics of different lexical bundles produced by L2 English speakers, but there is limited information on the extent to which the deployment of such bundles by L2 writers influences evaluators’ perceptions of the quality of writing. Understanding the features of the writing produced by the writers and their impact on quality is important, as such findings have significant implications for L2 competence and instruction. Hence, the present study endeavored to investigate lexical bundles in consideration of criteria aspects across learner writing development according to CEFR competency. First, a learner corpus comprising papers written by civil and environmental engineering students was used to choose and rate the learner data in the present study. Essentially, the current research focused on phraseological discourse elements rather than lexical or syntactical ones, examining lexical bundle use across CEFR-defined competency categories. L2 developmental studies have extensively focused on the latter, and there is a dearth of literature on the connection between the deployment of lexical bundles and the overall quality of L2 writing as perceived by readers. As a result, this study viewed it as essential to collect more data in a situated testing context with established ratings while applying the same measures used in earlier research in order to better understand the possible linkage between the utilization of lexical bundles and the quality of academic writing. Therefore, this study focuses on lexical bundles because “using well-tried expressions in appropriate places” (Biber et al., 1999) is an essential factor in assisting L2 learners in sounding natural in their second language (Vidakovic & Barker, 2010). In light of this, the current research investigates how lexical frequency distributions, as well as the frequency, grades, and functions of lexical bundles, are utilized in written responses across a range of proficiency levels to gain insight into the lexical development of learners of a second language (L2).

The research questions (RQs) outlined below serve as the basis for this investigation:

RQ1. What are the most common lexical bundles found in a corpus of academic reports written by English as a second language learners?

RQ2. To what extent do the most frequently used bundles differ between high scoring and low-scoring essays?

RQ3. How EFL learners’ writing performance and language proficiency influence their use of lexical bundles?

Literature Review

Within the English for Academic Purposes (EAP) literature, scholars (e.g., Casal & Kessler, 2020) have recognized the need for empirical research and the urgent need to understand formulaic sequences in academic writing. Regarding concerns related to research and teaching, lexical bundles are considered an essential component of the system. This is primarily because learning how a student makes use of lexical bundles can contribute to L2 learning and linguistic ability (Chen & Baker, 2016; Staples et al., 2013). Furthermore, because recurring chunks of formulaic expressions can differ depending on the genre and/or a writer’s rhetorical aims (e.g., Du et al., 2022), drawing the attention of L2 learners to these features is critical. Doing this is likely to foster the process of L2 learning and impact students’ success in a particular domain.

In addition to the use of corpus-based methods to extract and analyze various (e.g., 3-, 4-, and 5-word) lexical bundles, many EAP studies have striven to describe the functional nature of these bundles by classifying them in terms of the taxonomy developed by different researchers, including Biber et al. (2003, 2004), Bychkovska and Lee (2017), Chen and Baker (2016), Nesi and Basturkmen (2006), (Afzaal & Xiangyi, 2020), and Staples et al. (2013). The framework developed by Biber et al. (2004) proposes three lexical bundles, including discourse organizers, referential expressions, and stance expressions. The authors state that their framework assists in differentiating bundles based on the functionality they contribute to the discourse. Using this taxonomy, researchers have provided significant insights into genre-specific speech and the parallels and differences in the use of the English language by L1 and L2 writers.

For instance, Bychkovska and Lee (2017) analyzed argumentative essays written in English by L1 and L2 speakers of the language and the lexical bundles (and functions) utilized in each set of texts. Their study found that the L2 writers utilized a substantially greater number of stance bundles in their writing, and they frequently used certain discourse organizers, whereas L1 writers of English used these infrequently. It is essential to understand such L2 writer strategies as the related insights can assist teachers in directing students’ attention to particular discoursal aspects.

When evaluating the written output of second language writers, research has primarily examined how various elements of their writing influence their evaluations (Todd et al., 2007). However, studies investigating the different types of lexical bundles produced by learners across a range of competency levels and settings have produced mixed results. Additionally, these studies have tended to overlook assessments of the functions of lexical bundles that depend on the content or prompt of the study. Nevertheless, some studies have examined the prevalence of these bundles. This is because lower-level learners’ written output may depend heavily on the prompt or topic. For example, Huang (2015) found that higher-level learners made more extensive use of content-dependent bundles, while studies by Staples et al. (2013) and Vo (2019) observed significantly less usage. Except for the study by Huang, the studies above have generally examined lexical bundles based on the framework offered by Biber et al. (2004). Chen and Baker (2016) and Ruan (2017) revealed that speakers of all competency levels extensively use discourse organizers. However, in the research conducted by Vo (2019), the group with the highest level of proficiency was the only one that made more frequent use of discourse organizers than other bundles. The lower groups tended to use referential or stance bundles more frequently than discourse organizers. According to Staples et al. (2013), learners regularly utilized stance bundles, accounting for over 50% of all bundles deployed across levels. Three of these investigations also examined the structural types of lexical bundles and found a consistent trend. They found that verb phrases (VPs) were commonly used, but as levels advanced, prepositional phrases (PPs) or noun phrases (NPs) were increasingly used instead. Researchers generally focus on the characteristics of the bundles produced by L2 writers, and research has yet to examine the extent to which learners’ use of lexical bundles may impact how well their writing is perceived. This is an essential area of investigation for several reasons, including the fact that the findings can assist in guiding aspects of classroom pedagogy, and lower or intermediate level students might benefit from such training.

In this study, we aim to investigate the use of vocabulary and lexical bundles in non-native writing samples across different skill levels. Our research departs from existing studies, which have not given sufficient attention to the specific type of lexical bundles created by various learner populations and their association with the overall quality of writing demonstrated by engineering learners. Our contribution to this field lies in focusing on three-word, four-word, and five-word lexical bundles, which have not received enough attention within the corpus-driven formulaic language spectrum compared to longer bundles. Following this, we will also explore the use of longer lexical bundles.

Data and Methods

This section of the paper provides theoretical insights and a detailed description of the corpus tools and methods used in the study. The first part serves as an introduction to the learner corpus, while the second section outlines the analysis procedure utilized in the study. Additionally, it describes the software and tools used for data management and analysis, such as Sketch Engine and LexiConc. The penultimate section presents the results of the analysis and a discussion of these results. Finally, the last section provides concluding remarks and highlights the pedagogical implications of the study.

Learner Corpus Description

The corpus used in the present study consists of 119 reports written by L2 learners of civil and environmental engineering (CEE). Before writing their reports, these L2 learners took English language courses. They came from different engineering disciplines, including Environmental, Geotechnical, Structural, Transportation, and Construction, and were enrolled at a university in Hong Kong where they were invited to participate in an English language workshop program. The program was taught entirely in English, and the language and communication team teaching at the CEE department designed the curriculum. As part of their assessment, students were required to submit academic papers. A holistic rubric was developed to evaluate the written reports, which looked at learners’ competence in presenting research problems, research design, results, and conclusions. The reports were graded according to their clarity, readability, and conformity to conventions of grammar, vocabulary, punctuation, and spelling. Reports were assigned grades of A, B, or C, with A indicating the highest level of mastery. Reports with a grade of C were excluded from the analysis, as they were written by students from only two of the disciplines included in the study.

Corpus of the study

Sections such as acknowledgments, table of contents, references, tables, and diagrams were excluded from the analysis. Only the abstract and content sections were used for lexical bundle analysis. The texts were uploaded to Sketch Engine for tokenization and lemmatization. The study utilized rubrics attached in the Appendix for scoring and grading of reports.

Table 1 below provides the information of the learner data under investigation.

Table 1.

The Summary of Leaner Data.

Major	Grade	# of texts	# of tokens
Environmental	A	10	74,237
Environmental	B	12	1,55,707
Geotechnical	A	7	56,594
Geotechnical	B	18	1,26,847
Structural	A	18	1,53,654
Structural	B	31	3,05,497
Transportation and construction	A	9	90,508
Transportation and construction	B	14	96,548
In total		119	10,59,592

Tools of the Study: LexiConc: (A Corpus Tool for Lexical Bundles and Concordances)

This study uses LexiConc (www.lexiconc.info) software, purely designed for the lexical bundles, concordances, wordlist, word frequency, and visualization of words. It is a corpus tool that is used for analyzing language and examining patterns in texts. It is primarily used by linguists, lexicographers, and language teachers to study vocabulary and language usage. The program allows users to create and analyze large corpora of written or spoken language, including texts from multiple languages (Anthony, 2006). LexiConc offers various tools for text analysis, such as concordance, frequency lists, collocations, and clusters. These tools help researchers to identify patterns and trends in language usage, and to gain insights into the meaning and context of words and phrases. Overall, LexiConc is a powerful tool for analyzing language data and provides a valuable resource for language researchers and educators. Sinclair explains that the core of the corpus, in contrast to the text, is that you do not watch it directly; instead, you make use of instruments of indirect inquiry, such as query languages, concordancers, collocators, parsers, and aligners. Therefore, we used Sketch Engine and LexiConc to analsye the linguistic data. The CEE corpus is explored in two stages. Firstly, we identified 3-word, 4-word, and 5-word lexical bundles using Sketch Engine. In the second stage, we extracted the frequency of each lexical bundle using LexiConc. In addition, the software also provides facilities such as concordances, collocation, visualization, and Key Word in Context as shown in (Figures 1 and 2).

Figure 1.

Frequency and diversity of lexical bundles in LexiConc.

Figure 2.

Interface of LexiConc: frequency of lexical bundles by file.

Analysis Procedure

Identification of Lexical Bundles, Concordances, and Visualization

While lexical bundles are automatically determined based on how frequently and broadly associated words and phrases occur, researchers have employed a variety of criteria to establish what constitutes a bundle. The threshold frequency has been estimated to be anywhere from 10 (Biber, 2006) and 20 (Hyland, 2008a, 2008b) to 40 (Biber et al., 2004) occurrences per one million words. Additionally, raw rates have been taken into account (Chen & Baker, 2010). To eliminate the possibility of idiosyncratic applications, another identification criterion requires that sequences be present in a predetermined number of corpus files. For example, 3 to 5 files (Biber & Barbieri, 2007) or 10% of files (Hyland, 2008a) are acceptable options. The length of the strings chosen is another factor that analysts need to consider. Two-word bundles are highly prevalent, and they are of little utility for research purposes (Staples et al., 2013). On the other hand, 5- and 6-word bundles are relatively uncommon and frequently encompass lesser ones. Most research is focused on groups of four words, possibly because they are nearly ten times more common than sequences of five words and provide access to a wider variety of structural and functional patterns for investigation (Biber et al., 1999).

First of all, Sketch Engine was used in our study to generate the master list of lexical bundles used in learner corpus. We uploaded the essays written by students from all disciplines to Sketch Engine and used its N-GRAMS function to generated the list of 3, 4, and 5-word grams. N-grams used more than ten times and appearing in at least five papers were identified as lexical bundles in our analysis. In addition, we have manually scrutinized the master list to remove those that are sub-grams of another lexical bundle. For example, a better understanding and understanding of are both nested grams in a better understanding of, which should be ruled out of the analysis. As a result, 882 lexical bundles were identified from the learners’ English essay.

Secondly, the study uses a software called LexiConc to examine learners’ lexical bundle use in their written production. LexiConc is developed with two extraordinary functions. It allows users to feed in a list of lexical bundles and is able to calculate how many times each lexical bundle appears in the corpus and in how many different texts each bundle appears. Figure 1 shows the interface of LexiConc exhibiting each lexical bundle’s raw frequency in the learner corpus and the number of different texts it was used. Another novel function of LexiConc is that it can calculate the number of all lexical bundles and the number of different lexical bundle types used in a certain text. The next section exemplifies the interface of LexiConc displaying the frequency and diversity information of lexical bundle use in each essay. With the use of Skectch Engine and LexiConc, we managed to clarify for how many times and in how many different essays a certain lexical bundle was utilized by learners. In addition, the two research tools allowed us to analyze each essay concerning the number and types of lexical bundles included within it.

In addition, we calculated frequencies, normalized frequencies to answer RQ1. Moreover, a two-way Analysis of Variance (ANOVA) test was performed to analyze the effect of students’ majors and writing quality on the diversity and normalized frequency of lexical bundles used in their academic writing. Also, another two-way ANOVA test was conducted to examine where there was statistically significant interaction between the effects of major and grade concerning all different length of lexical bundles to answer RQ2 and RQ3.

Results and Discussion

The results of the study are presented in this section. The quantitative analysis includes both the structural point of view pertaining to the learners’ ability to incorporate lexical bundles in their reports as a means of enhancing fluency and discourse coherence.

882 lexical bundles have been identified from the learner corpus, including 757 3-word grams, 60 4-word grams, and 65 5-word grams. As shown in Table 2, on average, one academic paper contained 316.14 3-word grams (raw frequency, SD = 133.00, Range = 103–717), 140.73 variety of different number of lexical bundles (type, SD = 36.02, Range = 59–225) and 13.87 4 word grams (raw frequency, SD = 9.49, Range = 2–75) in 7.94 variety (type, SD = 3.32, Range = 219). Moreover, one academic report included 5.61 5-word grams (raw frequency, SD = 3.99, Range = 0–21) in 4.01 variety (type, SD = 2.46, Range = 0–13).

Table 2.

Frequencies and Variety of Lexical Bundles by Length: Mean (SD; Range).

Bundle type	Raw frequency	Normalized frequency (normalized per 10,000 words)	Type
3-word	316.14 (133.00;103.00–717.00)	413.66 (140.86;25.36–945.59)	140.73 (36.02;59.00–225.00)
4-word	13.87 (9.49;2.00–75.00)	18.18 (10.55;1.08–71.51)	7.94 (3.32;2.00–19.00)
5-word	5.61 (3.99;0.00–21.00)	7.48 (5.29;0.00–27.01)	4.01 (2.46;0.0–13.00)

To answer RQ1, the study found that the 3-word grams were found to be most frequently used by learners, and the 5-word grams were the least frequently used ones.

The 3-word, 4-word, and 5-word the most frequent lexical bundles that appeared in different academic papers are presented in Tables 3 to 5 respectively.

Table 3.

3-Word Lexical Bundles Used by Most Different Texts.

Lexical bundles	Raw frequency	# of texts
The degree of	208	105
Due to the	609	104
According to the	447	96
Based on the	431	90
It is a	228	86
In Civil Engineering	116	84
In order to	294	81
There is a	221	81
In Hong Kong	660	80
Which is a	174	78
As shown in	493	75
The number of	307	75
Most of the	184	74
The effect of	249	70
In this study	301	69
Part of the	186	68
To investigate the	167	65
It can be	186	63
To determine the	138	63
Compared to the	170	62
Is used to	133	60
There is no	110	60
Because of the	129	59
Related to the	120	59
The results of	173	59
In terms of	225	58
The result of	174	58
As well as	185	57
The relationship between	172	57
Which is the	128	56

Table 4.

The Four-Word Lexical Bundles Used in Different Papers.

Lexical bundles	Raw frequency	# of texts
Civil and Environmental Engineering	131	103
On the other hand	129	50
At the same time	78	44
Can be used to	55	32
One of the most	42	28
The effect of the	35	25
The size of the	34	24
It is necessary to	38	22
As a result of	36	21
Higher than that of	37	21
Be one of the	24	20
It is found that	67	20
It is important to	28	20
Can be divided into	29	19
Is similar to the	28	19
In the range of	31	18
It is observed that	31	18
One of the major	26	18
This is because the	37	17
A large number of	19	16

Table 5.

Five-Word Lexical Bundles Used in Different Papers.

Lexical bundles	# of texts	Raw frequency
Is one of the most	21	18
It can be concluded that	20	15
It was found that the	16	13
It should be noted that	23	12
Of this project is to	15	12
Is one of the major	11	11
Due to the fact that	14	10
It is assumed that the	17	10
Due to the lack of	12	9
Is larger than that of	9	9
Is to find out the	11	9
The lower part of the	15	9
The upper part of the	19	9
A better understanding of the	8	8
A significant impact on the	9	8
Due to the presence of	10	8
In the center of the	12	8
Is lower than that of	8	8
The purpose of this study	13	8

Differences Between High- and Low-Scoring Texts

In this section, we compare the overall the use of lexical bundles by writing quality, which will be determined by looking at the texts that received high scores and texts that received low scores.

Overall, the 4 (Disciplines: Environmental, Structural, Geotechnical and Transportation and Construction) × 2 (Grade: A vs. B) between-subjects ANOVA revealed that the lack of a statistically significant interaction between the effects of discipline and grade (F(3, 111) = 0.14, p = .94, =.004) on the diversity of lexical bundles. Simple main effects analysis demonstrated that both discipline (p < .05, =.12) and grade (p < .05, =.05) have a statistically significant effect on the types of lexical bundles. High-scoring groups across all four disciplines tend to use more diversified lexical bundles. In addition, the post hoc comparisons across disciplines showed that students majoring in Environmental studies used significantly less diversified lexical bundles than Structural majors (p < .05). No significant difference among the other majors was evident (as indicated in Figure 3).

Figure 3.

Lexical bundle use by different majors: diversity and normalized frequency.

As for the normalized frequency of lexical bundles, the ANOVA test did not reveal statistically significant interface between the effects of disciplines and grade (F(3, 111) = 1.90, p = .13, =.05). Simple main effects analysis revealed that grade did not have a statistically significant effect on the normalized frequency of lexical bundles (p = .78), while discipline had a statistically significant effect on it (p < .05, =.13). Post hoc tests indicated that Environmental majors used a significantly smaller number of lexical bundles than Geotechnical majors (p < .05) and Structural majors (p < .05). Meanwhile, Structural majors used significantly more lexical bundles than Transportation and Construction majors (p < .05). No significant difference among the other majors was evident (Figure 4).

Figure 4.

High- and low-scoring groups’ lexical bundle use: diversity and normalized frequency.

Comparison of 3-Word/4-Word/5-Word Lexical Bundles

Breaking down by bundle length, the interaction effects of independent variables on the use of lexical bundles obtained from the two-way ANOVA test are summarized in Table 6. To answer RQ1 and RQ2, the study revealed that there is no direct effect of interaction between the effects of discipline and grades pertaining to the usage of lexical bundles. Moreover, the two-way ANOVA revealed that there was no statistically significant interaction between the effects of discipline and grade concerning all different lengths and types of lexical bundles.

Table 6.

Interaction Effects of Discipline and Grade.

Lexical bundles	Discipline × grade
Lexical bundles	F	p	ηp2
3-word normalized frequency	1.77	.16	.05
3-word diversity	0.12	.951	.00
4-word normalized frequency	2.21	.09	.06
4-word diversity	2.43	.07	.06
5-word normalized frequency	1.01	.40	.03
5-word diversity	0.50	.68	.01

Table 7 as shown with * sign presents the main effects of discipline as well as grade on the deployment of different lengths of lexical bundles. Simple main effects analysis indicated that discipline only had a statistically significant effect on the diversity of 3-word lexical bundles and frequency of 3-word lexical bundles and 5-word lexical bundles. Papers of the Structural discipline were found to feature a statistically larger number of 3-word lexical bundles than those of Environmental and Transportation and Construction ones. Moreover, the main effects analysis revealed that discipline had statistically significant effects on the diversity of three-word lexical bundles. Academic papers of the structural discipline also included significantly more types of 3-word lexical bundles. In addition, papers in structural discipline contained a larger amount of 5-words than the other disciplines. It should also be noted that high-scoring groups contained significantly less types of 3-word lexical bundles. To answer the RQ3, the findings of the present study showed variations when compared to those reported by previous research in L2 settings. Our study showed many similarities with the studies by Staples et al., 2013) and Vo (2019) in terms of high scoring among the higher proficiency scoring group. Both high-scoring and low-scoring categories exhibited a tendency toward extensive use of 3-word lexical bundles, which was a notable trait observed in the study.

Table 7.

Main Effects of Discipline and Grade.

Lexical bundles	Discipline			Grade
Lexical bundles	F	p	ηp2	F	p	ηp2
3-word normalized frequency	5.53	<.05*	.13	0.10	.75	.02
3-word diversity	5.29	<.05*	.13	5.68	<.05*	.05
4-word normalized frequency	0.69	.56	.02	0.05	.82	4.65*e−4
4-word diversity	0.59	.63	.02	1.21	.28	.01
5-word normalized frequency	3.34	<.05*	.08	0.00	.97	1.11*e−5
5-word diversity	1.49	.22	.04	0.76	.39	.00

Overall, our study identified that writers using English as a second language (ESL) employ fewer bundles and a limited variety of bundle types in their writing, in addition to using more verbal and clausal bundles than CEE learners whose works received high scores (Tables 2 and 3). This suggests that there are still patterns utilized by CEE learners that have not been found in ESL writings, while some patterns appear much less frequently than in academic reports. The results of this study indicate that the texts written by CEE writers contain a relatively higher number of passive structures for packaging information. The writers tend to use more research-oriented bundles, focusing on the study subject matter rather than on presenting information to the audience and engaging them.

Moreover, the study found that the writers’ L2 backgrounds appear to influence the bundles they choose to use, with limited overlap in the favored forms and the degree of utilization. These findings diverge from those of the study conducted by Wei and Lei (2011), which focused on the use of lexical bundles by advanced Chinese academic writers of English as a foreign language (EFL). They used a corpus of Chinese EFL doctoral theses and a corpus of professional journal articles. However, the lexical bundles revealed in these corpora shared comparable structural characteristics, featuring prepositional phrases, noun phrases, and adjectival phrases. They observed that the Chinese EFL learners’ corpus featured more lexical bundles in terms of frequency and variety. It is important to note that as the results of the present study are also likely to be influenced by the learners’ disciplinary background and level of study, it is difficult to attribute patterns of usage to the writers’ L2 alone. Although CEE writers with grade A made more frequent and wider use of bundles in comparison to writers with grade B, variations in results when compared to the findings of the study by Wei and Lei (2011) may also be attributed to the greater number of texts within the corpus used in the current study. While the study by Wei and Lei found that students used fewer types and tokens than academics, in the present study, junior scholars were also found to use fewer bundles than scholars in other groups.

The present study provided some valuable insights but was limited in several ways. Firstly, the corpus used was relatively small, and future research could benefit from a larger corpus or replicate the study under similar conditions. Additionally, the data collected in this study was cross-sectional, and future research could investigate lexical bundle usage over time to identify developmental patterns and their relationship to improvements in writing quality. To ensure robust and consistent findings, researchers should examine formulaic expressions produced by students across various competence levels and assessment conditions.

Conclusion

The primary objective of this research was to gain insights into lexical frequency distributions and bundles in L2 writing in civil engineering across various disciplines and levels of language proficiency. The aim was to provide a more comprehensive understanding of lexical development in student writing. The results were based on three CEE sub-corpora, comprising 119 texts. Notably, the effect analysis indicated that both discipline and grade had a statistically significant impact on the use of lexical bundles. The study found that learners’ skills improved with an increase in the frequency of word types and tokens used. High-scoring writers used a broader range of word families than low-scoring writers. The study also found that high-performing writers across all disciplinary groups tended to use more diverse lexical bundles. Post hoc comparisons revealed that while Environmental studies majors used less diverse lexical bundles than Structural engineering majors, no significant difference was found among the remaining majors.

Broadly speaking, the study reveals how CEE writers, with different language backgrounds, bring their individual experiences and perspectives to their writing, infusing it with collocations that often differ from those found in published texts and papers written by L1 writers.

In conclusion, it is essential to emphasize that the study’s findings are descriptive and aim to provide insights into how L2 academic writers write, rather than prescribe how they should write. Academic writers’ texts often contain conventional combinations but are also notable for their use of unfamiliar combinations or the absence of familiar ones. This does not imply that the texts are “wrong,”“inappropriate,” or “non-native” (Afzaal, 2020; Kanglong & Afzaal, 2020). Instead, the writings of CEE scholars reflect contributions to an evolving code that is continuously appropriated and modified. Given that academic discourse is a melting pot of Englishes where different varieties of English intersect (Barbara et al., 2024; Kaibao & Afzaal, 2024; Siu et al., 2024), the ongoing participation of ESL authors in the global academic discourse community is likely to increase the range of bundles gaining visibility in research and academic texts over time.

We urge further investigation into engineering students’ L2 competence and language and replications studies in their pedagogical settings. While the present study attempted to capture CEE learners’ contributions to the formulaic patterns of academic discourse, it would be insightful to compare advanced L2 learners’ use of lexical features with those of L1 academic writers, considering the diversity, productivity, and appropriateness identified in the advanced level responses. Finally, the study concludes that lexical bundle proficiency can significantly enhance the language skills of civil engineering students, enabling them to understand and communicate complex ideas in their field. This proficiency can contribute to their academic success and professional development.

Footnotes

Appendix

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Barbara Wing Yee Siu

Muhammad Afzaal

Hessah Saleh Aldayel

Samantha Curle

Data Availability Statement

Data will be provided on request to the corresponding author.

References

Afzaal

Xiangyi

(2020). Book review: Ken Hyland and Feng (Kevin) Jiang, academic discourse and global publishing: Disciplinary persuasion in changing times. Discourse Studies, 22(3), 384–386. https://doi.org/10.1177/1461445620905135

Anthony

(2006). Concordancing with AntConc: An introduction to tools and techniques in corpus linguistics. JACET Newsletter, 155, 2085.

Barbara

S. W. Y.

Afzaal

Aldayel

H. S.

(2024). A corpus-based comparison of linguistic markers of stance and genre in the academic writing of novice and advanced engineering learners. Humanities and Social Sciences Communications, 11(1), 1–10.

Biber

Conrad

(2019). Register, genre, and style. Cambridge University Press.

Biber

(2006). University language: A corpus-based study of spoken and written registers. John Benjamins.

Biber

Barbieri

(2007). Lexical bundles in university spoken and written registers. English for Specific Purposes, 26, 263–286.

Biber

Conrad

Cortes

(2003). Lexical bundles in speech and writing: An initial taxonomy. na.

Biber

Conrad

Cortes

(2004). If you look at…: Lexical bundles in university teaching and textbooks. Applied Linguistics, 25, 371–405.

Biber

Johansson

Leech

Conrad

Finegan

(1999). Longman grammar of spoken and written English. Pearson.

10.

Bychkovska

Lee

J. J.

(2017). At the same time: Lexical bundles in L1 and L2 university student argumentative writing. Journal of English for Academic Purposes, 30, 38–52.

11.

Carter

Sinclair

(2004). Trust the text: Language, corpus and discourse. Routledge.

12.

Casal

J. E.

Kessler

(2020). Form and rhetorical function of phrase-frames in promotional writing: A corpus-and genre-based analysis. System, 95, 102370.

13.

Chen

Y.-H.

Baker

(2010). Lexical bundles in L1 and L2 student writing. Language, Learning and Technology, 14(2), 30–49.

14.

Chen

Y. H.

Baker

(2016). Investigating criterial discourse features across second language development: Lexical bundles in rated learner essays, CEFR B1, B2 and C1. Applied Linguistics, 37(6), 849–880.

15.

Council of Europe. (2001). Common European framework of reference for languages: Learning, teaching, assessment. Cambridge University Press.

16.

Cortes

(2006). Teaching lexical bundles in the disciplines: An example from a writing intensive history class. Linguistics and Education, 17, 391–406.

17.

Afzaal

Al Fadda

(2022). Collocation use in EFL learners’ writing across multiple language proficiencies: A corpus-driven study. Frontiers in Psychology, 13, 752134.

18.

Hawkins

J. A.

Buttery

(2010). Criterial features in learner corpora: Theory and illustrations. English Profile Journal, 1, e5.

19.

Hawkins

J. A.

Filipović

(2012). Criterial features in L2 English: Specifying the reference levels of the common European framework (Vol. 1). Cambridge University Press.

20.

Hoey

(2012). Lexical priming: A new theory of words and language. Routledge.

21.

Huang

(2015). More does not mean better: Frequency and accuracy analysis of lexical bundles in Chinese EFL learners’ essay writing. System, 53, 13–23.

22.

Hyland

(2008a). Academic clusters: Text patterning in published and postgraduate writing. International Journal of Applied Linguistics, 18, 41–62.

23.

Hyland

(2008b). As can be seen: Lexical bundles and disciplinary variation. English for Specific Purposes, 27, 4–21.

24.

Kaibao

Afzaal

(2024). The translation teaching platform based on multilingual corpora of Xi Jinping: The Governance of China: Design, resources and applications. Acta Psychologica, 242, 104110.

25.

Kanglong

Afzaal

(2020). Lexical bundles: A corpus -driven investigation of academic writing teaching to ESL undergraduates. International Journal on Emerging Technologies, 11(5): 476–482.

26.

Kellogg

R. T.

Raulerson

B. A.

(2007). Improving the writing skills of college students. Psychonomic Bulletin & Review, 14(2), 237–242.

27.

Kim

Kessler

(2022). Examining L2 English University students’ uses of lexical bundles and their relationship to writing quality. Assessing Writing, 51, 100589.

28.

Kyle

Crossley

S. A.

(2015). Automatically assessing lexical sophistication: Indices, tools, findings, and application. Tesol Quarterly, 49(4), 757

29.

Nattinger

J. R.

DeCarrico

J. S.

(1992). Lexical phrases and language teaching. Oxford University Press.

30.

Nesi

Basturkmen

(2006). Lexical bundles and discourse signalling in academic lectures. International Journal of Corpus Linguistics, 11(3), 283–304.

31.

Pan

Liu

(2019). Comparing L1–L2 differences in lexical bundles in student and expert writing. Southern African Linguistics and Applied Language Studies, 37(2), 142157.

32.

Pan

(2020). Lexical bundle variation across moves in abstracts of medical research articles. Southern African Linguistics and Applied Language Studies, 38(2), 109–128.

33.

Ruan

(2017). Lexical bundles in Chinese undergraduate academic writing at an English medium university. RELC Journal, 48(3), 327–340.

34.

Siu

B. W. Y.

Nok Yi

Afzaal

(2024). Programme-level English enhancement strategy in civil engineering undergraduate education: an application of innovative strategy. Innovation in Language Learning and Teaching, 1–18. https://doi.org/10.1080/17501229.2023.2298409

35.

Staples

Egbert

Biber

McClair

(2013). Formulaic sequences and EAP writing development: Lexical bundles in the TOEFL iBT writing section. Journal of English for academic purposes, 12(3), 214–225.

36.

Todd

R. W.

Khongput

Darasawang

(2007). Coherence, cohesion and comments on students’ academic essays. Assessing Writing, 12(1), 10–25.

37.

Vidakovic

Barker

(2010). Use of words and multi-word units in Skills for Life Writing examinations. Cambridge ESOL: Research Notes, 41(1), 7–14.

38.

(2019). Use of lexical features in non-native academic writing. Journal of Second Language Writing, 44, 1–12.

39.

Wei

Lei

(2011). Lexical bundles in the academic writing of advanced Chinese EFL learners. RELC Journal, 42(2), 155–166.