Professional Documents
Culture Documents
Method+Results+Discussion AL
Method+Results+Discussion AL
https://doi.org/10.1093/applin/amac040
This study explores how lexical bundles link smaller functional units (steps) in
research article (RA) conclusions in hard and soft science knowledge fields by
using corpus-based and corpus-driven approaches. RA conclusion corpora from
the hard and soft sciences were compiled and four-word lexical bundles were
extracted from each corpus. Conclusions containing lexical bundles were then
imported into the UAMCorpus tool for manual tagging of functions, structures,
moves, and steps associated with lexical bundles. It has been found that the
four-word bundles play an important role in realizing the communicative pur-
pose of the functional units in RA conclusions in both the hard and soft sciences.
The results demonstrate that the resultative lexical bundle (e.g. the findings of the)
is a strong indicator of the conclusions in both soft and hard sciences. Specific
bundle structures and functions are identified as signalling different steps of the
conclusions. The findings of this study add to the growing body of knowledge
regarding the bundle–move connections and contribute to the construction of
the RA conclusions in different disciplinary cultures.
INTRODUCTION
Since the early 1980s, due to the importance of research articles (RAs) attached
to scholarly communication, numerous studies have examined the linguistic
characteristics of RAs across disciplines (e.g. Swales 1981, 1990, 2004; Biber
et al. 1999; Hyland 2008a, b; Cortes 2013; Lu and Deng 2019; Li et al. 2020).
A good number of the studies have contributed to our understanding of the
linguistic realization of different RA sections, including abstracts (Omidian et
al. 2018), introductions (Swales 1981, 1990, 2004; Lu and Deng 2019), meth-
ods (Martı́nez 2003), results (Lim 2010) and discussions (Brett 1994). Swales’
(1990) Create-a-Research-Space (CARS) model has generated considerable
research concerning the introductions of RAs. However, relatively few studies
© The Author(s) (2022). Published by Oxford University Press. All rights reserved. For permissions,
please email: journals.permissions@oup.com.
2 L. DENG AND J. LIU
conclusion from the hard sciences published in the journal Biosystem in 2007.
The Sublime Text software was used to check the spellings, collapse broken sen-
tences, and find non-ASCII characters. The hard science conclusion corpus
(CCHS) contains 1,500 texts and 345,139 word tokens. The soft science conclu-
sion corpus (CCSS) includes 1,500 texts and 536,134 word tokens.
Adapted from Biber et al. (2004), Durrant (2017), and Hyland (2008a, b)
M1S1: Summarizing the overall findings As shown in Table 6, the M1S1 of con-
clusions in both corpora is characterized by a heavy reliance on NP-based and
PP-based bundles, reflecting the highly informational condensed nature of jour-
nal articles (Gray 2015). Particularly, soft science conclusions utilize noun phrases
with embedded of bundles more frequently (LL = 22.78, p < 0.00) in Step 1. On
the other hand, VP-based and Clause-based bundles (Example 6) tend to be the
major indicators of M1S1 in the soft science conclusions. Clausal bundles are
frequently used in Step 1 (62.5 per cent) by hard science writers. Furthermore,
VP + passive verb (Example 7) and it + passive verb bundles were used to signal
M1S1 in hard science conclusions.
6) These findings suggest that the RAP not only has utility with Black youth
and adults, but with White youth…as well. (ConL.SS_2012RSP_003)
7) It can be concluded that this catalyst prepared via the CVAP technique
is assumed to be a promising ORR catalyst…(ConL.HS_2015ASS_0026)
(M1S1)
Moreover, writers from both science fields prefer using noun phrase + active verb
to trigger M1S1 in conclusions. The results indicate that 86.69 per cent of bun-
dle tokens in the noun phrase+ active verbs structure incorporate the verbs ‘suggest’
and ‘indicate’ in M1S1 by soft science writers, whereas 62.34 per cent of the
bundles in this structure incorporate the verb ‘show’ by hard science writers. The
use of ‘research’ verbs (e.g. show) in Example 8 from CCHS reflects the ‘scien-
tific ideology’ that regards ‘laboratory activity’ as ‘impersonal, cumulative, and
inductive’ in hard disciplines, which is in contrast to the interpretive operations
found in soft knowledge fields (Example 6) (Hyland 2004: 38).
8) …our results show that the traditional models have lower predictive capac-
ity than the machine learning models (ConL.HS_2017ESA_0031) (M1S1).
science writers use only 2 tokens in M1S1 out of the total 14 bundle tokens in
the conclusions. The results suggest that hard science writers intend to present
themselves as researchers and promoters of their study.
First-person pronoun + active verb bundles are usually (Example 9) used to sig-
nal M1S1 in the hard science conclusions. Hyland (2002: 1104) notes that using
first-person pronouns would help readers establish a personal authority based on
the confidence and command of their arguments, therefore, creating a plausible
interpretation of the findings.
In M1S1, resultative bundles are found to take up 45.99 per cent and 42.37 per
cent in soft and hard science conclusions, respectively, indicating that the com-
municative function of drawing primary findings from research (Hyland 2008a)
can be realized by using resultative bundles in different disciplines.
It is suggested that ‘knowledge in the hard sciences is mainly achieved
through the correct use of prescribed procedures, and nature reveals itself
directly through the scientific method’ (Hyland 1999: 355). It may be for this
reason that hard science writers use more procedure and location bundles in
M1S1 to specify the research procedures by describing the instruments used
(Example 10) and the properties of the objects investigated (Example 11).
10) Further, it is noted that with the use of the feature selection, the classifica-
tion performance improved for all TPMs with a smaller number of features
(ConL.HS_2015ESA_0023) (M1S1).
11) For un-doped TiO2 thin films, the nanoparticles reveal a polyhedral struc-
ture, and the surface of the polyhedron was level and smooth (ConL.
HS_2014ASS_002) (M1S1).
12) Together with the finding that own-account knowledge workers are more
likely to be satisfied with their jobs than the permanent workers … (ConL.
SS_2016SIR_0024) (M1S1).
13) In this report, we show for the first time that this posttranslational
modification is also present in human embryonic stem cells… (ConL.
HS_2014SCR_0022) (M1S1).
M1S5: Describing the research procedure/method Table 6 illustrates that hard science
conclusions employ more bundle tokens in M1S5. Writers from both science
fields are found to rely on NP- and PP-based bundles to describe the research
procedures with hard science writers using slightly more NP + of and PP + of
bundles (Examples 14 and 15) in M1S5.
14) The performance of the proposed method is then compared with PSO ...
based methods (ConL.HS_2016ESA_0016) (M1S5).
15) Hence the proposed technique was tested in the presence of noise cor-
rupted signals as it was illustrated in…(ConL.HS_2016AI_002) (M1S5).
Hyland (2008a) also found that bundles used to describe research objects or
contexts were typically realized by NP + of structure. Furthermore, the incor-
poration of nouns in bundles can contribute to the informational purpose of
academic writing (Biber 1988) in the hard sciences. It has also been found that
NP- and PP-based bundles serve more often as lexical complements to M1S5
than as triggers to M1S1 in the hard sciences.
Disciplinary variations are found in employing structuring bundles (LL =
17.45, p < 0.00), location bundles, and intangible framing bundles in M1S5.
Framing bundles such as on the basis of (Example 16) and is based on the are
used distinctively by hard science writers to introduce the steps/instruments
employed. Soft science writers often use within the context of to refer to the gen-
eral framework applied in the study (Example 17).
16) In the next step, by using hybrid feature selection algorithm on the basis
of 9 different filter algorithms and function-based clustering method…
(ConL.HS_2015ESA_003) (M1S5).
17) Within the context of expanded econometric framework, a wide range of
variables are considered (ConL.SS_2013HE_004) (M1S5).
MOVE-BUNDLE CONNECTION IN CONCLUSION SECTIONS ACROSS
DISCIPLINES 17
In general, soft science writers use more NP + active verbs in M1S1, whereas hard
science writers prefer anticipatory it + passive verbs bundles, suggesting an objec-
tive and ‘impersonal way’ of communicating research findings across differ-
ent science fields (Li et al. 2020: 95). Moreover, intangible framing bundles are
found to be closely associated with M1S5 in hard science conclusions to modify
the abstract properties of the instruments.
also favoured by the writers in both science fields; however, stance bundles are
used more frequently in M2S1 than in other moves and steps (Figure 4). The
following sections discuss the bundle–move connection in M2S1 and M2S2 in
greater detail.
Resultative bundles in M2S1 are used together with verb phrases can contribute,
may serve, and provide to indicate the contribution and the value of the research
in soft science conclusions (Example 21).
21) The findings of this study can contribute to the discussions around pro-
fessional identity formation (ConL.SS_2009TTE_009) (M2S1).
Distinctive stance bundles are found in M2S1 in the two clusters. For example,
for the first time and to the best of are distinctively adopted and used together in
MOVE-BUNDLE CONNECTION IN CONCLUSION SECTIONS ACROSS
DISCIPLINES 19
22) This paper shows for the first time the hierarchy of different stimuli, as well
as the combined effect of … on stimuli type (ConL.HS_2014BioS_006)
(M2S1).
23) In addition, to the best of our knowledge, this is the first report show-
ing the effect(s) of AMPs on the membrane permeability (ConL.
HS_2013CLMS_005) (M2S1).
According to the examples above, writers from both fields use the evaluative
adjective ‘first’ to emphasize the uniqueness and the significance of their contri-
bution. By adding the word ‘first’, the argument sounds less intrusive, fulfilling
the communicative function of convincing the readers that the study is worth-
while (Loi et al. 2016).
M2S2: Indicating the limitations It is found in this study that soft science writers
mainly use noun phrase with embedded of bundles to construct M2S2. In this step,
triggering bundles are found in soft science conclusions, such as the limitation of
this (study) and limitation of the current (study) (Example 24).
24) There are numbers of potential limitations of this study. First, the level of
specificity of the competencies…(ConL.SS_2015IHE_002) (M2S2).
It is interesting to observe that the * of the and the * of this bundle frames are
typically used by soft science writers to highlight the current findings in M2S2.
Usually, the two slots are filled by results and findings, along with negative
words (e.g. cannot in Example 25, and be cautious in Example 26) to indicate the
limitations.
25) Because the characteristics of students in this course may differed from
other …, the results of this study cannot be generalized beyond the educa-
tion program…(ConL.SS_2011IHE_002) (M2S2).
26) We would need to be cautious in generalizing the findings of this study to
the development of English listening strategy… (ConL.SS_2016Sys_0019)
(M2S2).
Framing bundles, such as beyond the scope of and in the absence of, contribute to a
negative evaluation of the study (Example 27), which may indicate M2S2 in the
soft sciences. Often, hard science writers use negative evaluation words instead
to indicate limitations (‘drawbacks’ and ‘disadvantages’ in the examples below).
It can be concluded that the noun phrase with embedded of structure is preferred
in M2S1 and M2S2 of the soft science conclusions. Specifically, the contribution of
the can be viewed as a trigger to M2S1, while limitations of the study is an indica-
tor of M2S2 which is used exclusively by soft science writers. Additionally, VP +
active verbs and NP + active verbs were found to be closely associated with M2S1
in the soft sciences.
Both corpora use evaluative words to help ‘assert the writer’s position and
perspective’ in Move 2 which is consistent with Loi et al.’s study (2016: 12).
Moreover, the results indicate that M2S1 is characterized by stance bundles in
both knowledge fields, and structuring bundles are connected to M2S2 in the
soft sciences.
purpose. Only soft science conclusions are found to employ bundles in Step 2,
indicating that soft science writers tend to be more focused on seeking pedagog-
ical implications.
29) Clearly, further research is needed on how MOOCs can better reach, retain
and support under-served populations (ConL.SS_2017IHE_005) (M3S1).
In the case of bundle functions, writers from both science fields utilize framing
bundles to convey the aim of the future study, such as for the purpose of in the
hard sciences (Example 30) and with the aim of (Example 31) in the soft sciences.
30) For the purpose of future testing, trade-off issues could be addressed by
placing a greater emphasis on the service…(ConL.HS_2017ESA_0041)
(M3S1).
31) Additional analysis, both quantitative and qualitative, needs to be car-
ried out, with the aim of revealing the mechanism of the change…(ConL.
SS_2016TTE_0013) (M3S1).
32) Hence, one of the major future directions for the current research is to
reduce the computational complexity of the proposed method… (ConL.
HS_2016ESA_0029) (M3S1).
33) To validate the findings of the present study, future research should
engage a larger number of participants in different EFL…contexts (ConL.
SS_2008JSLW_001).
Conclusion writers from both fields usually use structuring bundles together
with phrases ‘future research/studies/work’ as complementary elements to recom-
mend further research. However, soft science writers prefer procedure bundles
such as more research is needed and stance bundles (e.g. it would be interesting) to
help trigger M3S1.
M3S3: Drawing empirical implications Writers from both science fields attempt to
highlight the empirical and theoretical implications in M3S3. In hard sciences,
writers tend to rely heavily on VP + passive verbs and anticipatory it with passive
verb bundles (Example 34). However, VP with active verbs and anticipatory it +
copula be + adjectives structures are more frequent in the soft sciences (Example
35). Bundles with the passive structure are often based on the bigrams ‘be
used’ (Example 36) and ‘be applied’ found in M3S3 which is corroborated by
Durrant’s (2017) finding that bundles with ‘used to’ are common in science
and technology texts. It is also found that ‘be used’ occurs most often in the
bundle ‘can be used’ and is often followed by prepositions: in, to, for, and as
(Example 36).
The use of procedure bundles in M3S3 differs between the two fields. Hard
science writers often use the hedging word ‘could’ to collocate with ‘be used to’
(Example 37), which indicates writers’ intervention in the discourse to evaluate
(Hyland 2004) and provide readers with practical suggestions.
MOVE-BUNDLE CONNECTION IN CONCLUSION SECTIONS ACROSS
DISCIPLINES 23
Soft and hard science conclusion writers employ stance bundles we believe that
the and we were able to in M3S3 to express confidence in the implications of their
findings/approach and promote the study (Examples 38 and 39).
Overall, VP-based bundles with passive verbs are associated with M3S3 in the
hard sciences to draw empirical findings, whereas clausal bundles with passive
verbs are likely to trigger M3S1 (e.g. further research is needed) in the soft sci-
ences. Additionally, writers from both knowledge fields make extensive use of
stance bundles that include cognitive nouns in first-person pronouns + active verb
structure (e.g. believe) to promote the empirical implications.
CONCLUSIONS
Summary of findings
This paper attempts to analyse the disciplinary variations in the conclusion sec-
tions of RA as reflected in the use of four-word LBs relating to moves and steps.
Unlike previous studies, it focuses not only on bundle–move connections but
also on bundle–step and bundle–move relations by combining corpus-driven
and corpus-based methods as well as genre analysis. The findings of this study
shed light on how bundle structures and functions are linked with steps and
how specific four-word LBs trigger particular steps of conclusions. Although not
all the bundle–step connections can be discussed, the present study has gener-
ated some major findings.
First, regarding the overall distribution of LBs, writers from both soft and hard
sciences make extensive use of bundles in M1S1 (summarizing the findings).
Although hard science writers adopt a wider range of LBs than their counter-
parts in soft sciences, soft science writers tend to use more bundle tokens in
general. Moreover, soft science conclusion writers employ more bundle tokens
in M2S1 (indicating the significance/advantage of the study) and M3S1 (recom-
mending future research).
In terms of the grammatical structure of LBs across disciplines, writers from
both science fields still rely heavily on NP-based and PP-based LBs, especially
‘the noun phrases with embedded of’ phrase fragment bundles in conclusions,
24 L. DENG AND J. LIU
END NOTE
1 Since we set the cut-off point at 25 whole corpus. The equations are as fol-
times per million words when we lowing:For soft science corpus:The raw
retrieve the bundles in Antconc, we frequency = 25 × 536,134/1,000,000 ≈
pre-calculated the raw frequency 13For hard science corpus:The raw fre-
that the bundles should appear in the quency = 25 × 345,139/1,000,000 ≈ 9
ACKNOWLEDGEMENTS
We would like to extend our great gratitude to the editors and the blind review-
ers for the valuable comments and suggestions for the manuscript.
This manuscript is part of the research output of the National Social Sciences
research project entitled ‘A Genre-based Study of the Dynamic Interdiscursive
System in Chinese and Foreign Professional Discourse’ (China National Social
Sciences, NO. 17BYY033). It is also supported by Project No.19G002 from Hubei
Provincial Department of Education, China. This manuscript is also supported
by Hubei University of Economics under grant no. XJ20BS39.
26 L. DENG AND J. LIU