Lexical co-occurrence of the modal verb “must” in business media discourse: A corpus-based study

DOI:

Among modal verbs of necessity, must occupies a distinctive position in English grammar. Its frequency in institutional and professional registers, combined with its gradual retreat in conversational speech, makes it an instructive indicator of how written discourse maintains normative force. While much research has addressed the semantics of modality, distinguishing epistemic and deontic readings, comparatively little attention has been given to the habitual lexical environments in which must occurs. This study does not attempt to disambiguate its meanings; rather, it investigates the lexical co-occurrence patterns that accompany the form itself. The aim of the study is to trace recurrent associations of the modal verb must with particular verbs, subjects, and nominal fields to describe how necessity is habitually formulated in the language of business reporting.

The choice of business news as material stems from its characteristic tension between factual reporting and prescriptive framing. Economic events are not only described but also evaluated through implicit judgments about what actions ought to be taken. In this register, must functions as an instrument of authority that turns observation into recommendation. Mapping its collocational profile therefore allows one to see how the language of business reporting distributes necessity across institutional actors, such as governments, corporations, and markets, and how it situates these actors in relation to policy, regulation, and risk.

Methodologically, a collocational approach enables the analysis of must without reliance on interpretive coding. Instead of deciding whether a token expresses obligation or inference, the study examines statistically salient lexical neighbours that reveal patterned co-selection. This shift from semantic classification to distributional description follows corpus-linguistic principles that treat meaning as a function of use. It also ensures transparency and replicability: every association can be traced to measurable frequencies rather than subjective judgment.

The relevance of such an inquiry is twofold. Linguistically, it contributes to the description of modal behaviour in a specialised register, as it documents how the grammar of necessity interacts with domain-specific lexis. From a discourse perspective, it provides empirical evidence of how business reporting frames economic activity through recurrent verbal and nominal choices surrounding must . In this way, the study bridges grammatical description and discourse analysis and ultimately shows how a single modal verb participates in constructing the rhetoric of obligation that underlies professional communication.

Theoretical background

“Must” and modality in business reporting

Modal verbs have long been central to studies of English grammar and discourse because they mediate between description and evaluation. Within this group, must has been examined as a marker of necessity [Coates, 2014; Collins, 2009; Piotrovskaya, Trushchelev, 2022; Van Linden, Verstraete, 2011], inference [Leech et al., 2009; Collins, 2009; Depraetere, Reed, 2020], and normative pressure [Fairclough, 2003; Hyland, 2005; Malyuga, McCarthy, 2021; Trnavac, Põldvere, 2024], yet most existing scholarship focuses on general English, academic prose, or conversation. Analyses of business reporting remain limited, despite the modal’s clear presence in texts that blend factual reporting with prescriptive commentary.

Corpus-based descriptions by Biber [2004] and Boginskaya [2022] demonstrated that must is markedly more frequent in formal written registers than in spontaneous speech, suggesting its role in institutional discourse where statements of obligation and certainty are stylistically acceptable. The work by Coates [2014] and Collins [2009] refined the classification of modal meanings, but these studies rarely address how must operates in media contexts. When modality has been considered in financial and economic reporting [Rantsudu, Bartlett, 2024; Ahmad, Mahmood, Farukh, 2020; Khramchenko, 2023], attention has usually turned to the persuasive potential of hedging and stance rather than to the formal behaviour of a single modal item.

In business reporting, must often surfaces at the juncture between observation and instruction. Phrases such as companies must adapt or governments must act impose necessity while maintaining the appearance of analytical neutrality. Studies on evaluative language in financial discourse [Bednarek, Caple, 2012; Chernyavskaya, 2023] note that such constructions enable journalists to convey policy recommendations indirectly, through grammatical necessity rather than over judgment. Yet no systematic examination has traced the lexical partners that sustain this usage or quantified their distribution. The present research therefore positions itself at the intersection of modality studies and discourse analysis, as its intention is to describe the regular lexical surroundings of must as a means of identifying how institutional obligation is routinely phrased in business news.

Lexical co-occurrence and distributional approaches

Lexical co-occurrence analysis rests on the premise that meaning is observable in patterns of proximity between words. Early formulations in corpus linguistics [Sinclair, 2004; Stubbs, 2005] argued that collocation reveals the habitual combinability of items and thereby their contextual function. In more recent models of distributional semantics, a word’s profile is defined by the company it keeps [Boleda, 2020; Grishechko, Tomalin, 2025], quantified through association measures such as Mutual Information or logDice. Collocational analysis has become a standard method for identifying lexical preferences, semantic prosody, and register-specific phraseology without recourse to interpretive coding.

A related line of research by Malyuga and Rimmer [2021] applied co-occurrence analysis to the term buzzword , demonstrating how lexical surroundings reveal shared attitudes and evaluative tendencies. Their method of using frequency and contextual association to expose discourse meaning illustrates how collocational data can serve as evidence of pragmatic orientation without the need for subjective coding.

Applications of this method to modality have shown that modal verbs display distinct collocational fingerprints depending on register and communicative purpose. Studies by Millar [2009] and Leech [2014] documented diachronic changes in the distribution of must , noting its gradual decline in conversation but persistence in written institutional genres. Other research has applied collocational statistics to stance expressions [Granger, Paquot, 2008; Römer, 2009], demonstrating that recurring lexical associations can serve as reliable indicators of pragmatic orientation. Within this framework, must can be approached not as a polysemous item requiring semantic disambiguation but as a node whose neighbouring words reveal the lexical-grammatical environment of necessity in a particular field.

Together, these works exemplify an empirical tradition that treats professional language as a patterned system of lexical and grammatical choices reflecting institutional behaviour.

For business reporting, a collocational analysis provides an empirically grounded means of capturing how necessity statements are routinely constructed. Rather than classifying each instance by meaning, the study identifies which verbs, subjects, and nominal fields most frequently co-occur with must . These associations, when aggregated across a large corpus, may offer a structural account of how professional media discourse encodes obligation, certainty, and normative stance through recurrent lexical patterning.

Material and methods

Corpus and data retrieval

The present study draws upon the British National Corpus (BNC), particularly its Newspapers subcorpus, which provides a broad sample of British press writing across genres and publication types. The corpus contains material from both national broadsheets and regional outlets, thus reflecting a variety of editorial voices and communicative purposes. Its textual range makes it a suitable source for examining the linguistic patterns through which obligation and necessity are expressed in the public domain.

The Newspapers component was accessed through The Prime Machine, a corpus analysis environment providing lemmatisation, part-of-speech tagging, and keyword-in-context functionality. The search term was the lemma must , retrieved case-insensitively to include all positional variants.

The corpus query yielded a very large number of occurrences of must, and a sample of six hundred concordance lines was selected for analysis. This number was determined intentionally rather than by technical limitation. The sample size was designed to provide a body of data that would be large enough to display stable lexical and grammatical tendencies, yet sufficiently compact to allow systematic manual examination and contextual verification. The decision follows the principle of manageable representativeness discussed in corpus-linguistic methodology [Sinclair, 2004; Tognini-Bonelli, 2001; McEnery, Brookes, 2024], which holds that the analytical adequacy of a corpus does not depend solely on its absolute size but on the interpretive stability of recurrent patterns within a well-defined subset of data. As Sinclair [2004] observes, representativeness is relative to purpose: a smaller, carefully controlled corpus may yield more consistent and replicable findings than extensive but unexamined material. This reasoning also aligns with Mizumoto, Plonsky, and Egbert [2021] view that sampling in descriptive studies should balance coverage and analytical tractability, ensuring that findings can be both empirically grounded and replicable.

Although the entire Newspapers subcorpus comprises several million words, the restricted sample reflects its internal proportionality: it includes texts from multiple outlets, subgenres, and publication periods. The size of the sample ensures that frequent collocational tendencies emerge clearly while rare or idiosyncratic cases do not distort the overall distribution.

Search design

The study focuses on the modal verb must , selected for its high frequency and its capacity to convey both deontic and epistemic meanings. The search was carried out at the lemma level to include all inflectional variants, though in this case the paradigm is minimal. Queries were conducted case-insensitively to include sentence-initial occurrences.

Collocational data were obtained using the Lines function of The Prime Machine, which returns concordance lines with flexible context windows. The window was set to four words to the right of the node ( R1–R4 ), while the left-hand context was deliberately excluded. Only righthand collocates were subjected to quantitative examination. This methodological decision follows from the grammatical behaviour of English modal auxiliaries: must functions as an operator that determines the meaning of the clause element that follows it. The lexical material that completes its semantic structure lies to its right, in the form of a main verb, an auxiliary chain, or intervening adverbials.

Limiting the analysis to the right-hand environment allows for an accurate observation of the predicate field, that is, the linguistic area in which the modal verb imposes its semantic and pragmatic force. Left-hand collocates typically represent grammatical subjects or discourse referents (we, the government, companies), which, while relevant to agency, do not modify the modal meaning itself. From the perspective of modality, they provide structural rather than interpretive information. The restriction of the analytical scope therefore ensures that the results reflect the internal lexical realisation of obligation or inference, rather than topic or participant distribution.

Methodological rationale

The decision to confine the span to the righthand side also has a methodological grounding in corpus linguistics. Previous research on modals [Lieven, 2008; Hilpert, 2016; Flach, 2021] demonstrates that the principal collocational environment of auxiliaries is asymmetrical: the lexical complement, not the grammatical subject, carries the modal verb’s semantic load. Consequently, studies of English modality typically employ right-sided spans (R1–R4 or R1–R5) to capture the syntactic domain of dependency. This configuration avoids the inflation of collocation counts by predictable grammatical subjects and yields a more interpretable concentration of verbs, adjectives, and participles that realise modal meaning in text.

From a discourse-analytic perspective, this focus aligns with the nature of journalistic prose. In editorials and reports alike, must frequently introduces prescriptive or inferential assertions (formulations such as must act , must be addressed , must change ) where the rhetorical thrust of the sentence is carried by the lexical material following the modal. Concentrating on this segment therefore provides direct access to the linguistic devices through which business reporting encodes necessity, recommendation, or judgement.

Data processing

The concordance lines retrieved through The Prime Machine were exported as an Excel file for further processing. Each line contained three components: the left context, the node word, and the right context. The first lexical item appearing to the right of must was extracted automatically and normalised to lowercase. Non- lexical tokens, such as punctuation marks, numerals, and truncated strings, were excluded.

A frequency list of the extracted items was then produced, resulting in a ranked inventory of right-hand collocates. Verbs were examined both in their base and participial forms to account for variation between active and passive constructions. The resulting table constitutes the empirical foundation of the present analysis, providing a quantitative overview of the lexical environments of must in contemporary business reporting.

Results

The analysis retrieved 600 concordance lines containing the modal verb must from the British National Corpus (Newspapers subcorpus) accessed through The Prime Machine. Each line was automatically tagged and manually inspected to ensure accuracy of lemmatisation and part-of-speech identification. From each line, the lexical item immediately following the node was extracted. In cases where intervening function words (such as not , also , still ) preceded the first lexical element, the next content word was recorded as the collocate. This procedure yielded a dataset of 589 valid tokens, after excluding instances with incomplete or truncated right-hand context.

The frequency distribution of the collocates shows a clear concentration of a limited set of verbs and auxiliaries. The thirty most frequent right-hand collocates are presented in Table.

The distribution is heavily right-skewed, with be and have together accounting for nearly half of all right-hand collocates (275 of 589 tokens, or 46.7%). The remaining items occur at comparatively low frequencies, typically between two and fifteen instances. The majority of collocates belong to the verb class, with occasional adverbs ( also , now , not ) occupying the immediate post-modal position as intervening elements preceding a verb phrase.

Patterns of syntactic realisation indicate that must most frequently introduces verb phrases beginning with the auxiliary be followed by a past participle, forming passive or adjectival constructions. Combinations with have usually occur in perfect forms, typically followed by been and a past participle. Lexical verbs such as take, make, pay, and act mark active constructions denoting volitional or process-oriented actions. Less frequent verbs (decide, know, change, ensure) represent mental or institutional processes, reflecting the broader functional diversity of journalistic style.

Token distribution across the four-word window shows that the main lexical verb typically occurs within R1–R3, while items at R4 are predominantly complements or modifiers (e.g., must be held accountable ). The positional behaviour of adverbs confirms the flexibility of intervening elements: not and also most frequently occupy R1, preceding a lexical or auxiliary verb in R2.

No significant orthographic or tagging inconsistencies were found in the retrieved dataset. Cases containing embedded quotations, headlines, or incomplete clauses were removed to preserve grammatical comparability. The final working corpus therefore comprises 589 fully contextualised occurrences of must in the BNC Newspapers subcorpus, providing a stable empirical base for the subsequent discussion of collocational tendencies and their implications for the expression of necessity in business reporting.

Discussion

The collocational profile of must observed in the BNC: Newspapers subcorpus reflects a regular and internally coherent system of usage characteristic of British business reporting. The distribution of right-hand collocates confirms the strong grammatical asymmetry of must as an auxiliary, whose modal force is projected forward onto a narrow range of predicate structures. The findings show that, in the language of the press, necessity is rarely expressed through direct imperatives but through periphrastic, syntactically mediated forms that allow obligation to be presented as inference, evaluation, or shared assumption rather than instruction.

The most frequent collocates following must (R1–R4 position)

Rank	Collocate	Frequency	Example from corpus
1	be	186	must be accepted, must be considered
2	have	89	must have been, must have done
3	also	15	must also act, must also recognise
4	take	14	must take account, must take action
5	not	13	must not allow, must not fail
6	now	12	must now act, must now decide
7	make	9	must make sure, must make changes
8	pay	9	must pay attention, must pay for
9	come	8	must come from, must come under
10	go	7	must go through, must go on
11	start	5	must start again, must start from
12	decide	5	must decide whether
13	know	5	must know that, must know how
14	change	4	must change course, must change direction
15	get	4	must get rid, must get on
16	find	4	must find out, must find a way
17	wait	4	must wait and see
18	act	3	must act quickly
19	ensure	3	must ensure that
20	look	3	must look at, must look for
21	govern	3	must govern themselves
22	meet	3	must meet standards
23	do	3	must do more
24	face	2	must face up to
25	give	2	must give way
26	help	2	must help to
27	learn	2	must learn from
28	remember	2	must remember that
29	understand	2	must understand that
30	win	2	must win support, must win back

The predominance of be and have as the immediate right-hand collocates situates the modal within what Willer [2015] termed “epistemic expansion,” where must serves to mark inference or logical necessity rather than obligation in the interpersonal sense. Comparable tendencies have been reported in large-scale corpus studies by Suikkanen [2018] and Carr [2017], who note the gradual decline of deontic must in favour of other modal expressions, particularly in spoken and informal registers. The present results extend this observation to the written domain of business reporting, as they show that epistemic and impersonal patterns also dominate in texts that aim to instruct or evaluate institutional behaviour. The recurrent sequences must be and must have indicate that modal necessity in the press is realised predominantly through evaluative predicates ( must be understood , must be taken into account , must have been expected ), which frame obligation as a matter of reasoned necessity rather than directive force.

At the same time, the presence of a smaller but stable group of lexical collocates ( take , make , pay , act , change ) points to an active semantic core in which must continues to function as a marker of externally imposed necessity. These verbs, which denote tangible action or institutional compliance, correspond to the deontic subfield identified in earlier studies of professional discourse [Malyuga, McCarthy, 202 1 ; Sooriyaarachchi, 2025]. In business reporting, their use is typically associated with corporate or governmental actors ( companies must act , banks must hold capital , policymakers must take action ), illustrating how the modality of obligation is transferred from speaker-to-hearer interaction to an institutional plane. Such usage sustains the persuasive and evaluative potential of modal constructions without resorting to explicit commands, allowing writers to maintain an appearance of neutrality while articulating normative expectations.

The presence of intervening adverbs (also, now, not) is consistent with findings by Laliberté [2022], who notes that adverbial modification contributes to the pragmatic shading of modality in press prose. Their positioning immediately after must creates temporal or argumentative adjustments: must now act implies urgency; must also consider adds inclusiveness or concession; must not allow introduces moral or procedural interdiction. These recurrent patterns expose how the modal verb interacts with discourse-level operators to align readers with implied evaluations, a mechanism often described in pragmatic studies of stance-taking [Xiao-Desai, 2021; Kiesling, 2022; Sun, Yang, Chen, 2024].

The results portray must in business reporting as a modal form that has retained its grammatical prominence but shifted its pragmatic centre. It no longer functions primarily as a linguistic marker of authority but as a rhetorical instrument that encodes necessity through inference, convention, or institutional reasonableness. This finding reinforces the diachronic tendencies noted in previous corpusbased and general language surveys [Gisle, 2022; Grigoryeva, Zakirova, 2022] while refining them in an applied context where modal usage is tied to economic and policy discourse rather than general English. It also bears pedagogical relevance that has been noted in recent work emphasising the role of authentic data in advancing fostering language awareness [Grishechko, 2023].

The study has isolated and quantified the immediate right-hand collocates of must , and through this has contributed a replicable and data-driven account of how modal necessity is constructed in the professional register of business reporting. It offers an empirically delimited view of modal syntax and phraseology, avoiding reliance on impressionistic examples and grounding its observations in statistically supported lexical behaviour. The results will hopefully enrich current discussions on modality and evaluation in institutional discourse, as well as extend the descriptive basis for studies of stance, obligation, and epistemic authority in the language of the press.

Conclusion

The analysis of must in the BNC: Newspapers subcorpus provides a detailed picture of how necessity and obligation are articulated in contemporary business reporting. Working with a deliberately delimited corpus of six hundred concordance lines made it possible to observe recurrent lexical associations without losing sight of their local grammatical environments. The righthand collocate model, restricted to four positions following the node, has proven effective for isolating the predicate field through which modal meanings unfold. Within this narrow syntactic corridor, patterns emerged with striking regularity. These patterns have revealed how a limited number of verbs and auxiliaries carry the weight of modal expression across a wide textual spectrum.

The results indicate that must in business reporting functions as a pivot between obligation and inference. The dominance of be and have points to a preference for epistemic framing, while the smaller set of action verbs such as take , make , pay , and act keeps the deontic core alive, particularly in contexts describing institutional or policy responsibility. In this respect, the modal operates less as a command and more as a linguistic device that formalises consensus and transforms evaluation into necessity.

Methodologically, the study demonstrates that the description of modal verbs can benefit from fine-grained collocational observation rather than broad quantitative generalisation. The use of a moderate-sized, manually verifiable dataset ensures that findings rest on genuine textual evidence rather than statistical artefacts. This design, grounded in the principle of manageable representativeness, allows the results to be transparent, reproducible, and readily comparable to other corpora or registers.

The contribution of this study lies in its focus on the intersection of modal semantics and professional discourse. While earlier corpus-based analyses have described the general decline or grammatical drift of must in English, few have addressed its lexical surroundings in the specialised communicative domain of business reporting. Since this study has traced the recurrent syntagmatic companions of must , it has effectively added empirical texture to ongoing discussions of modality as a vehicle of stance and institutional persuasion.

Future research might build on these findings by extending the scope beyond British press material to other varieties of English or other professional registers where normative language plays a constitutive role, such as legal and policy documents. Diachronic comparison could also show how the modal patterns of business communication change in response to shifts in economic discourse and rhetorical convention.

The present results confirm that even a single auxiliary, when observed closely, can reflect a great deal about the communicative habits of a discourse community. In business reporting, must continues to serve as a subtle yet indispensable marker of necessity – less the voice of authority than the syntax of reason.