The Modal Expression of Necessity in English for Tourism

This study focuses on the modal notion of necessity expressed by the central modal auxiliaries and other semantically related modal expressions in English for Tourism. The general aim is to offer detailed coverage of the investigated items by placing considerable emphasis on variability in terms of specialised registers. Specifically, I intend to explore the similarities and differences among necessity modals related to modal meaning and their distribution, and observe whether there are any register-induced variations. To this end, the analysis has been conducted on a small-scale specialised corpus which contains texts from one spoken, and two written registers (promotional website and online responses to consumers’ reviews). A conflation of relevant perspectives underpins the qualitative analysis. A descriptive account is based on the distinction between epistemic and root necessity, and the parameters of strength and source in case of root modal uses. Quantitative findings point to a low frequency of identified items as well as to marked variation in their distribution across the registers. Overall, the use of necessity expressions is contingent on the medium of communication, communicative purpose and interactants.


Motivation for the study
It has not been long since Lam (2007: 72) commented on the lack of linguistic analyses of English tourism industry texts. Due to a global tourism boom, on the one hand, and greater availability of new tools for research, particularly those of corpus linguistics, on the other, the explorations of the language of tourism peculiarities have been flourishing ever since. 1 Despite some prior views (see Calvi, 2016: 189;Suau-Jiménez, 2012a: 125), the specificity of the language of tourism -be it considered a register (Lam, 2007), a macro-genre (Calvi, 2010) or a (specialised) discourse (Jaworska, 2013;Manca, 2016;Suau-Jiménez, 2012a) -has been acknowledged and empirically proven. A designation 'tourism English' (Lam, 2007) neatly captures a widely held view that English used in tourism is different from general English and other specialised languages inasmuch as it is characterised by a set of distinctive stylistic, linguistic (lexical/semantic and syntactic) and functional features (Edo-Marzá, 2012;Lam, 2007;Ruiz-Garrido & Saorín-Iborra, 2013;Suau-Jiménez, 2012a, 2012bSulaiman & Wilson, 2019). As an alternative, the label English for Tourism (EfT) is used here for it seems closely reminiscent of the field of applied linguistics, and, by implication, of English for Specific Purposes (ESP).
A glance at relevant publications, particularly collections of articles (Maci, Sala, & Godnič Vičič, 2018;Raţă, 2013), reveals the abundance of research strands and considerable diversity of the aspects of the language use which attract scholarly attention. Linguistic enterprise potentially yields important implications pertinent to ESP instruction since the learners need to know "the lexis and grammar institutionalised in the English currently and authentically used in the tourism industry before they can communicate appropriately and effectively" (Lam, 2007: 72). Hence, there seems to be a valid reason for conducting research into complex areas of English grammar, such as modal verbs, or modals for short, in EfT.
Given a great variety of communicative functions modals can perform and the interpersonal aspect related to their usage, they are an important ingredient of communicative competence and accordingly an indispensable part of any grammar or a textbook for general English or ESP learning. Still, modals are "particularly troublesome to those who wish to learn the language" (Palmer, 2003: 1), EfT students being no exception (Godnič Vičič, 2008), and thus are "among the more difficult structures" teachers have to deal with (Larsen-Freeman, Celce-Murcia, Frodesen, White, & Williams, 2016: 137). Since a substantial contribution to ESP Vol. 8(2)(2020):  work may come from the studies investigating how the choice among modality markers is influenced by the differences in the register/genre 2 categories in different domains of language use, the research on modality in specialised discourses/genres has established itself as an important field of scholarly research. Drawing largely on Hyland's (1996Hyland's ( , 2005aHyland's ( , 2005b work, a wealth of studies have delved into academic discourse, especially academic writing (e.g. MA dissertations [Gabrielatos & McEnery, 2005], research articles [Orta, 2010;Piqué-Angordans, Posteguillo, & Andreu-Besó, 2002], student essays [Godnič Vičič, 2008], to name a few), whereas less attention has been given to modals in professional discourses. As for tourism discourse, modals have fallen under the scope of researches into various semantic-pragmatic concepts, including those thoroughly explored in academic writing (hedges [Hyland, 1996], stance and engagement [Hyland, 2005a[Hyland, , 2005b), as discussed below. So far, surprisingly few authors have specifically focused on modality markers (e.g. Alonso-Almeida & González-Cruz, 2012), which provides further motivation for this study.

Tourism discourse and the English modals
Due to the complexity and heterogeneity of the industry the language serves (Ruiz-Garrido & Saorín-Iborra, 2013) and a revolutionary impact of the Internet leading to the new forms of tourist communication (so-called Tourism 2.0 or cybertourism [Calvi, 2016;Edo-Marzá, 2012;Suau-Jiménez, 2016, 2019a), tourism discourse is characterised by a remarkable genre variety. Following Calvi (2010: 18-19), textual genres in the field of tourism may be grouped into three main thematic blocks 3 and placed in a hierarchical model which includes both higher categories (genre families and macro-genres) and lower ones (sub-genres). Although the cases of pure genres are rare, each genre may be defined in terms of its specific uses and a series of characteristics (lexico-grammatical, discursive and pragmatic) (Calvi, 2016: 197-198).
Notwithstanding the complexity of genres, most tourism texts essentially fulfil two pragmatic communication functions: providing information and promotion (Calvi, 2010). It is the latter that lies behind the prominence of modality markers in the discourse. Tourism discourse relies on the combination of textual and visual elements to attract the attention of the public and direct their gaze, urging them to act, to consume a tourist product (Suau-Jiménez, 2012a). The ultimate goal of reaching the maximum number of potential visitors turned into Vol. 8(2)(2020):  actual consumers/tourists engenders persuasion as an inherent feature of the discourse, notably in the field of tourism description and promotion (Calvi, 2016;Fuster-Márquez & Gregori-Signes, 2018;Manca, 2016;Suau-Jiménez, 2011, 2012a, 2012bSulaiman & Wilson, 2019). Various linguistic means which can increase the persuasive power of tourism texts (e.g. content words: adjectives [Edo-Marzá, 2012], verbs, nouns [Manca, 2016]) are used to induce certain behaviours in "the implied consumer" (Fuster-Márquez & Gregori-Signes, 2018). Previous studies (Đurović and Silaški's [2015] analysis of directives in tourism advertising genre, Manca's [2016] extensive research on persuasion in institutional websites, Potočnik Topler's [2018] study of travelogues) give clear indications of a persuasive aim being pursued through the English modals.
Since tourism is also regarded as "a cross-cultural dialogic process in which both tourists and promoters of tourist destinations participate" (Sulaiman & Wilson, 2019: 3), persuasion as a discursive macro-function can be viewed as part of the interpersonal relationship between the sender and the receiver (Suau-Jiménez, 2011, 2012a, 2012b, 2014, 2019b. Hence, the strategic use of modals as specific markers modulating persuasion has also been explored from the viewpoint of interpersonality and interactional metadiscourse (Hyland, 2005b). Previous studies suggest that modal verbs are salient interactional features in English texts. Modals are frequent markers of author's stance, i.e. hedges, in tourism formal (promotional websites [Suau-Jiménez, 2016] and travel books [Alonso-Almeida & González-Cruz, 2012]) and informal genres (traveller forums [Suau-Jiménez, 2014) and moderately used markers of engagement employed to build the relationship with the reader/customer in online genres (Suau-Jiménez, 2011, 2012a, 2012b, 2019a. As Suau-Jiménez's (2014, 2019a research shows, metadiscursive markers of interpersonal meaning participate in the characterisation of the investigated web-based genres. Suau-Jiménez (2019b: 208) states that "(b)oth author's stance and reader's engagement are typically discursive voices used to achieve persuasion in tourism promotional genres." This understanding of stance and engagement as the two voices in the dialogic interaction between author and customer (Suau-Jiménez, 2014, 2019a embodies the idea of tourism discourse as the language of social control (see Jaworska, 2013). Calvi (2010: 26) posits that bona fide genres in tourism discourse are those related to the description and promotion of the tourist product since "it is here where social control is exercised and intercultural mediation done". Linguistically, social control is manifested in greater use of specific verbal categories, including modal verbs, especially must and should (Jaworska, 2013), which brings us to the topic of this paper.

Aims, scope and outline of the study
Considering the above, the focal point of this study is one of the modal verb classes with the capacity to serve as "overt markers of persuasion" (Biber, 1988: 111). The current study is an attempt to probe deeper into the modal expression of necessity in EfT by conflating theoretical approaches from different fields (language typology, descriptive linguistics, pragmatics). Unlike most previous studies, it takes a broad view of the class, thereby not restricting it to the central modal auxiliaries only, but rather, including other semantically related verbs and periphrastic expressions, here dubbed 'necessity modals'. The general aim is to offer detailed coverage of necessity modals used by tourism professionals in authentic contexts and address variability in the choice of these linguistic items in terms of specialised registers. In place of the genre, the term register referring to "a variety associated with a particular situation of use (including particular communicative purposes)" (Biber & Conrad, 2019: 8) is adopted because it appears to be more strongly associated with the study of language use. More specifically, I intend to explore the similarities and differences among investigated modals related to their distribution and modal meanings, and observe whether there are any registerinduced variations. To this end, a small-scale, preliminary analysis was conducted on the incipient corpus which contains texts from three registers: one spoken register, and two written registers, which significantly differ in their underlying purpose and structure.
The next section sets the framework for the study. Section 3 is devoted to methods and the data used in the analysis. Key findings are presented and discussed in section 4. Following section 5, which discusses some pedagogical implications, the final section provides a brief overview of the conclusions and outlines the limitations of the study.

The modal system in English
While the fundamental notions underlying the definition of modality, together with classification, taxonomy, and the accompanying nomenclature of modal meanings, are still lying in the realm of scholars' subjective choices and theoretical backgrounds (Trbojević-Milošević, 2004: 19-39), the key exponents of this "notional" (Palmer, 1986) category are much easier to grasp. In English, it is a wellestablished set of modal forms that create a modal system (Palmer, 2003: 2). The principal members of this system are the modal auxiliary verbs which are a part of a larger closed class of auxiliary verbs formally distinguished on the basis of their inflectional and syntactic properties. Like primary auxiliaries be and have, they 167 279 Vol. 8(2)(2020):  exhibit the so-called NICE (negation, inversion, code, and emphatic affirmation) properties, i.e. they do not require do-support in four specific contexts (Depraetere, 2016: 373). In addition, the modal auxiliaries share a few more other distinctive features which, however, are not fully agreed on. The commonplace is that these verbs do not inflect for person and number, do not co-occur and do not have nonfinite forms (Palmer, 1990: 4). Besides, the feature of being followed by a bare infinitive may be taken as a distinguishing criterion (Collins, 2009: 13;Depraetere, 2016: 373), so that nine modals, including four pairs of present/preterit forms of single lexemes and must, qualify as the central or core modal auxiliaries (Biber, Johansson, Leech, Conrad, & Finegan, 1999;Depraetere, 2016;Depraetere & Langford, 2020;Quirk, Greenbaum, Leech, & Svartvik, 1985). 4 Alongside the central modals are numerous other verbs and periphrastic expressions capable of conveying modal meanings, which have been variously termed and grouped. This study draws on a division proposed by Quirk et al. (1985: 137). Accordingly, need to, occasionally grouped into semi-modals (Biber et al., 1999;Depraetere & Verhulst, 2008;Smith, 2003), is omitted from the analysis, and have got to is treated as a separate item from have to, although they may be taken as variants (Palmer, 1990). The following then fall within the scope of my analysis: central modals (must, shall, should), marginal modals (need, ought to), modal idioms (had better, be to, have got to) and semi-modals (have to, be bound to, be obliged to, be supposed to).

Modal meanings and concepts
A highly characteristic property of central modals is their rich semantic content as they typically express the diversity of (un)related meanings. This study follows the polysemy-based approach embraced in the highly-influential works which are taken as a starting point for this study: Palmer's seminal typological works on modality (1990,1986,2001), descriptive (Huddleston & Pullum, 2002;Quirk et al., 1985) and corpus-based (Biber et al., 1999) grammars of the English language. Therefore, each modal is considered to convey at least two independent meanings which are sufficiently distinct and disambiguated by the context.
Linguists' accounts of modality have traditionally rested on possibility and necessity regarding them as "central to modality in English" (Palmer, 1990: 9) and accordingly the "core modal concepts" (Huddleston & Pullum, 2002: 175), which is mirrored in grammarians' accounts. In the Longman Grammar of Spoken and Written English (LGSWE), Biber and colleagues state that "(m)odals are used to express a speaker's or writer's stance, expressing either the degree of (un)certainty of the proposition, or meanings such as permission, obligation, or necessity" (Biber et al., 1999: 457) and distinguish three major functional Vol. 8(2)(2020): 275-296 categories of modal verbs: permission/possibility/ability, obligation/necessity, and volition/prediction.
To an extent, the descriptive accounts of modals are clear reflections of the theoretical positions laid by Palmer (1990Palmer ( , 1986. Palmer (1990) advocates for the account of the English modals based on the distinction between 'kinds' of modality, i.e. the types of modal meaning (epistemic, deontic and dynamic), and 'degrees' of modality (possibility and necessity). 5 Being the modality of propositions, epistemic modality is, in logical terms, concerned with the speaker's attitude to the (factuality or truth-value) status of the proposition (Palmer, 2001: 8). The function of epistemic modals is to make judgments that the proposition in question is possibly or necessarily true, or, put differently, "that something is or is not the case" (Palmer, 1990: 50). As such, epistemic modality is clearly distinguished from the other types. The distinction between the two types of non-epistemic modality can be seen in terms of orientedness (deontic modality is discourse-oriented, which makes deontic modals essentially performative, whereas dynamic modality is subject-oriented [Palmer, 1990: 36-45]) or conditioning factors (with deontic modality they are external to the relevant individual, whereas they are internal with dynamic modality [Palmer, 2001: 9]).
Although this widely-applied classification may suffice for present purposes, particularly for establishing the dividing line between the meanings of must and have (got) to (Palmer, 1990: 116), the analysis of must can prove rather problematic since there is no clear distinction between its uses for deontic and (neutral) dynamic necessity (Palmer, 1990: 69). For this reason, epistemic necessity and root necessity will be taken here, as elsewhere (e.g. Smith, 2003), as the main sub-categories, with the term root necessity denoting the non-epistemic side of the modality spectrum which covers the semantic area of necessity and obligation. In distinguishing between the two, some clues indicative of a modality type will be considered: the appropriate paraphrase (epistemic meanings are glossed with a that-clause, whereas it is a for-clause with root meanings [Palmer, 1990: 8]), and syntactic and grammatical distributional properties characteristic of epistemic modals (co-occurrence with the progressive or the perfect aspect, stative verbs, inanimate subjects, pleonastic/expletive it and existential there as a subject (Collins, 2009;Palmer, 1986)).

Strength and source of modality
The parameters of strength and source will serve as important benchmarks for comparing the meaning distinctions between the investigated items. Following Vol. 8(2)(2020): 275-296 Huddleston and Pullum (2002: 175), the former concerns "the strength of commitment (prototypically the speaker's commitment) to the factuality or actualisation of the situation". Typically, the markers of strong necessity (e.g. must, have to, have got to) are distinguished from the markers of weak necessity (e.g. should, ought to, be supposed to), based on "the difference in the severity of consequences if the obligation is not fulfilled" (Smith, 2003: 242). Additionally, semantic strength may be altered by pragmatic factors (Huddleston & Pullum, 2002: 176-177).
As regards the parameter of source, especially pertinent to the present study is the framework developed by Depraetere and Verhulst (2008) with subsequent modifications (Verhulst, Depraetere, & Heyvaert, 2013). The source, conceived of as the driving force behind the necessity (Depraetere & Verhulst, 2008: 3), which lies at the origin of root necessity meaning, may be of three main types: discourseinternal, discourse-external, and mixed sources. With discourse-internal sources, a necessity originates in the subjective source "which is either the speaker or some other person that is part of the discourse situation" (Verhulst et al., 2013: 213). A discourse-external source is further grouped into three subcategories: a rule, a condition, or circumstances. A rule, or a regulation and rule, is a cover term which refers to compelling situations resulting from more strongly binding forces such as laws and institutional rules to more weakly binding forces such as traditions, instructions for use, rules of a game and social patterns (Depraetere & Verhulst, 2008: 6). The circumstances behind the necessity are varied: arrangements or particular situations that necessitate the actualisation of a particular situation, the nature of things viewed either as the laws of nature or in rather broad terms (Depraetere & Verhulst, 2008: 6). In examples of a condition, "a particular situation is necessary in order to achieve a particular purpose" (Depraetere & Verhulst, 2008: 8).

METHODS AND DATA
To comply with the set aims, the study employed both quantitative and qualitative analyses. Also, the methods of manual and software-assisted data analyses were combined. The material used in this study is a compilation of texts illustrating different types of communication between tourism industry professionals and customers (written vs. oral, and direct vs. indirect, see Suau-Jiménez [2012a]). This compilation, which contains 88,020 words, can be called a "small-scale specialized corpus" (Flowerdew, 2004). The rationale for using the corpus of this size is found in Flowerdew's (2004: 2, 21) observations that small-scale specialised corpora (20,000-250,000 words) provide relevant contextual information about the communicative situation, which makes them useful for context-sensitive analysis. The corpus contains three sub-corpora focused on specific text types belonging to spoken register (conversation [CON]) and two written registers (the promotional website [VISL], and online responses to consumers' reviews [TAR]).
Vol. 8(2)(2020):  Given the lack of publicly available spoken corpus of tourism discourse, it was the collection of speech-like text samples (Listening scripts) from various sources (language teaching and thematically relevant testing materials), 6 that was used to create CON (28,806 words) on the grounds that these are sufficiently illustrative of real professional communication and based on authentic language production. The texts cover the typical workplace professional-tourist communication (e.g. during check-in, check-out, at a travel agency), professionalprofessional communication, and a few monologue-like extracts (commentaries on tours, delivering a speech, or a presentation). The sub-corpus VISL (35,623 words) was compiled in May 2019 by extracting promotional texts (on things to do, places to visit and things to see) from the official London website. To build TAR (23,591 words), a selection of 175 responses from 56 London-based hotels to guests' reviews were taken from TripAdvisor, "the largest and best known site for online reviews in the area of tourism" (Suau-Jiménez, 2019b: 208), in March 2020. Responses to the reviews of all ratings displayed on a 1-5 scale (from 'terrible' to 'excellent') were included. To provide a sufficient variety, not more than five responses, each responding to a review with a different bubble rating, signed by the same person, were selected. Review responses strongly resemble traditional business letters and e-mail messages in terms of structural and textual features. The business's representatives acknowledge customers' voices, validate or repudiate their remarks concerning the experience in the hotel with at least twofold goal: to foster and repair the relationship with the reviewer, and to maintain or improve the business's online reputation (Zhang & Vásquez, 2014: 55). These management-related texts then partially share a promotional purpose with the destination's website.
After the texts were manually downloaded into Word files and converted into plain text files, AntConc (v 3.5.2) (Anthony, 2018) was applied as a word counter and analysis software. Concordance searches of each necessity modal, including the present-tense forms of the semi-auxiliaries and modal idioms, were conducted in each sub-corpus. The generated results were manually checked for irrelevant occurrences. Besides examples with no discernible modal (necessity) meaning, such as those with must-used as a combining form (must-see, must-eat) 7 or 'conditional' should (Huddleston & Pullum, 2002), 8 the 'syntactically-motivated' occurrences of have to 9 were also discarded. The results were then combined to obtain the data on the raw and normalised frequencies (per 1,000 words). Finally, Vol. 8(2)(2020):  the identified sentences with necessity modals were analysed semantically following the framework described above.

Distribution of necessity modals
Considering that English makes use of an impressive number of devices to convey the modal meaning of necessity, the findings point to a distinct lack of variety in this respect. Of 12 available expressions, as few as six feature in the investigated material, all of which being represented only in the conversational part of the data, whereas neither modal idioms nor a marginal modal occurs in written texts. Also, contrary to initial expectations, the quantitative results presented in Table 1 indicate that necessity modals are not a very common class of verbs in the data. As they average less than two cases per 1,000 words, they may, in fact, be regarded as a minor feature of EfT. Low figures are possibly due to the availability of numerous lexical items and syntactic constructions in English that can perform the same discourse functions. Further, they can be ascribed to the trend observed for general English suggesting that necessity modals are considerably less common overall than the other modal categories (Biber et al., 1999: 487, 493). Interestingly, the findings on the distribution and frequency of the investigated items match those of the previous extensive research on modals conducted on large-scale corpora of general English (the LSWE Corpus [Biber et al., 1999]; the BNC [Kennedy, 2002] the two written registers, they are substantially outnumbered by those in conversation (occurring almost eight times more often). The table shows that the three registers converge in one respect: the predominance of central modals. They are more frequently used in conversation than in written registers, which parallels previous findings (Biber & Conrad, 2019: 96;Kennedy, 2002: 78). The observed slightly greater frequency of should in comparison to that of must is also consistent with previous results (Biber et al., 1999: 486). It may come as a surprise that have to is almost non-represented in the investigated written registers given the overall increase in the use of semi-modals (Smith, 2003) and the fact that have to, despite being a fairly recent addition to the language (Palmer, 2003: 3), is relatively common in written texts (Biber et al., 1999: 488). Conversely, the greatest presence of have to in CON may be expected as semi-modals are especially salient in conversation (Biber et al., 1999: 487). Unlike have to, have got to is highly occasional in CON, which can be accounted for by the fact that modal idioms belong to a more familiar style related to informal language (Palmer, 1990: 114;Smith, 2003: 250), uncharacteristic of professional interactions constituting CON. The non-occurrence of the marginal modal and modal idioms in written texts is quite understandable because ought to is generally rare and modal idioms are restricted primarily to conversation (Biber et al., 1999: 486).

Epistemic necessity
The findings indicate that epistemic necessity is a minor meaning of the investigated items. Only central modals are used as the markers of epistemic modality and that being the case in a highly limited number of instances: three uses of must are clearly epistemic and should appears with an epistemic function in six examples. These are almost evenly distributed between CON and TAR as well as between present and past reference. A complete lack of epistemic modals in VISL contradicts some previous research mentioned in the Introduction (e.g. Suau-Jiménez, 2011, 2012a, 2012b). As an epistemic modality marker, must is typically related to the logical certainty based on what is known and viewed as involving objective modality paraphrasable by 'It is necessarily true/the case that…' (Collins, 2009: 38;Huddleston & Pullum, 2002: 181). This 'objective' must is not exemplified in the data. Rather, epistemic must is used to make judgments which typically rest with the speaker. Hence, it has subjective interpretation in terms of a "confident inference" (Huddleston & Pullum, 2002: 181) paraphraseable by 'The only conclusion that I can draw is that…' as Palmer (1990: 51) suggests. Therefore, examples (1) and (2) should be interpreted as inferred conclusions about the Vol. 8(2)(2020):  present or past situation respectively based on the reasons which are either implied or stated.
(1) There must be another way. (CON) (2) …we are incredibly sorry for any inconvenience this must have caused you during your stay. (TAR) In the examples above, must is used by a speaker/writer to draw the most obvious conclusion. As it expresses his/her absolute conviction, must is placed on the highest point of the epistemic scale. In comparison to this modal, should occupies the lower position of the scale, which is "intuitively closer to the strong end than to the weak", and expresses the modality strength comparable with that of probable (Huddleston & Pullum, 2002: 177, 186). Since should is purely epistemic in few cases (Palmer, 1990: 59-60, 62), the so-called "primacy of the deontic use" (Huddleston & Pullum, 2002: 186-187), none of the examples in my corpus seems to be completely devoid of root meaning. Nonetheless, should in (3) functions as the tentative marker of epistemic necessity expressing "rather extreme likelihood, or a reasonable assumption or conclusion" and "implicitly allowing for the speaker to be mistaken" (Palmer, 1990: 59).
(3) …wherever you arrive, there should be a welcome service for you. (CON) Under the epistemic interpretation, (3) is an inference regarding the probability of the situation being the case ('I am fairly confident that there will be…'). Being a matter of what the speaker expects will happen (Huddleston & Pullum, 2002: 186), the deductions made with should are quite distinct from those with must, with the major difference arising from the type of deduction. Whereas must involves the deductions based on knowledge or perceptions, the deduction in (3) is largely based on what the normal course of events is (Depraetere & Langford, 2020: 264). Like must, should is used with reference to past situations. Naturally, all such examples are found in TAR clearly reflecting the emphasis on past activities.
Although the examples like (3) may get root interpretation, particularly that of internal-source necessity, they are considered instances of epistemic necessity since an epistemic reading is much more likely with favourable situations than with unfavourable ones (Huddleston & Pullum, 2002: 187), which accounts for a seemingly high frequency of epistemic should in the corpus.

Root necessity
Root necessity meaning is identified when modal verbs convey the idea that it is necessary for an event to occur or that it is necessary for someone to do something (Depraetere & Langford, 2020: 273). Accounting for 97 examples in the material, root necessity is prevalent meaning in the data ascribed to all of the identified modals with more or less different shades of modal meaning.
Vol. 8(2)(2020):  As a root modal, must has been traditionally associated with the semantic notion of obligation and qualified as discourse-oriented, deontic (Huddleston & Pullum;2002;Palmer, 1990). When collocating with the 2nd person subject, it is used performatively to impose a directive (Collins, 2009: 35). The fact that neither this subjective deontic must nor mustn't expressing prohibition is found in the data can be plausibly explained by the "general tendency to avoid the face-threatening force of expressions with an obligation meaning" (Biber et al., 1999: 489), which is particularly strong in the discourse of service-providing industries. Rather, you must is used in CON, as exemplified by (4), to give a word of specialist advice or suggestion. As (6) illustrates, have to is used in quite similar discourse settings.
(4) You must also establish the client's priorities. (CON) (5) First, you must remember that last year the City considered the early launch of brochures a mistake. (CON) (6) You have to ring up to order something, that's the only way you can do it. (CON) The difference between the examples above lies in the source behind necessity: it is discourse-external in (4) and (6), while (5) involves a discourse-internal source. The source of the latter type is also identified in: (7) Wow, I must take a photograph... (CON) Here, must expresses an internal need/compulsion in the subject referent paraphrasable by 'it is necessary for me to...', i.e. dynamic modality in Palmer's terms (1990: 113). The same might apply to the examples involving the quasiformulaic use of root must with communication verbs (say, tell), or, in TAR, with apologise. In this use, the speaker/writer imposes the obligation on himself/herself and by so doing actually performs speech acts (Palmer, 1990: 74). Have to readily lends itself to expressing this personal obligation (collocating with say, inform) in CON. Other than these instances of weak necessity, the examples of strong modals with a discourse-internal source are quite difficult to pin down. Example (8) can be viewed as such since the writer states what is necessary for him/her to include based on personal disposition and preferences. Yet, it may be argued that what gives rise to the necessity are the characteristics of the mentioned sites, a discourse-external source, so that (8) could possibly be seen as a mixed-source. The impression is that prevalent are the sentences with a discourse-external source, with rules being more frequently associated with must and circumstances with have to. All three registers offer examples with the sources classifiable as rules and regulations. Following Depraetere and Verhulst (2008: 7), spoken Vol. 8(2)(2020):  instructions given in interactions, such as (4) above, are classified in this subgroup as they have a broad field of application and are not restricted in time. 10 Rules can thus be discerned in: (9) There's also a 20-minute spa trial available, but you have to book it beforehand… (CON) The data suggest that, in CON, the difference between the uses of must and have to for conveying this sense of the requirement which stems from rules is related to the context and participants in the interaction. With the 2nd person subjects, must appears in conversations between tourism professionals concerning the standards of performance and codes of conduct, as (4) shows, whereas have to occurs in professional-customer dialogues in statements about regulations relating to the provision of services, as in (9). Besides, the preference given to have to in (9) possibly arises from the view that must implies some authority on which the speaker relies. While a degree of authority can be felt in some face-to-face interactions with co-workers, the authorial-sounding modal is quite likely to be avoided in service encounters due to potential detrimental effects.
On similar grounds, we can account for the instances of this "objective deontic must" (Collins, 2009: 35) which have been, contrary to some previous studies (e.g. Manca, 2016), identified in the promotional register. In example (10), and five others, must issues requirements, typically with passive verbs (purchase, present, pay), the source for must being an official position to objectively state the institutional rules of service providers. When compared to an alternative direct imperative structure, must mitigates the face-threatening and somewhat aggressive force of a directive. The use of the passive form adds to this effect by further distancing the writer from the required action.
(10) Admission ticket to Kew Gardens must be purchased for access to Kew Palace. (VISL) The corpus offers examples of both strong modals conveying the necessity brought about by circumstances, which challenges Palmer's (1990: 116) claim that "must would not be used where it is clear that there is external necessity." The meaning of 'circumstances compel' may be implied or explicitly referred to, as in examples below. Another difference between the two strong modals has been seen in terms of implied actuality, with that not being the case with must (Collins, 2009;Palmer, 1990). Yet, judging from (11) and (12), there seem to be no categorical distinctions in this respect.
(11) Windsor Castle sometimes has to limit access to certain areas ... (TAR) (12) The icy Serpentine waters are usually below 4C (40F) degrees in the winter, so swimmers must become acclimatised over a period of time. (VISL) It may be claimed, however, that the differences are register-induced. Whereas the examples of have to with circumstances mainly come from CON, VISL offers only one example, albeit in the negative context: In other semantically similar examples in this register, have to occurs only in the negative form. Although these sentences are beyond the scope of this study, it could be useful to draw attention to this pattern of use as it seems a quite convenient means to increase the text's persuasive force and help catch the reader's attention. As in (13), have to is used to point to the lack of necessity of the adverse situation characterised through words carrying negative collocations, which is brought about by favourable circumstances. By explicitly stating the absence of frequent potential obstacles for tourists, mainly high costs, the text is more likely to persuade readers towards the desired point of view and the assumed action (of becoming tourists). As regards the instances of condition constituting the source of necessity, only one example with must is found wherein the driving force behind the necessity is the wanted situation.
The frequencies of have got to are too low to give any indications of whether it is closer to must than to have to (see Depraetere & Verhulst, 2008: 14). In both sentences, the source of necessity is discourse-external, either rules or circumstances. Whereas, in one case, the modal idiom is readily substitutable by must or have to, with a change only related to a shift towards greater formality, have got to in (14) seems to be replaceable not by strong modals, but by should, which is indicative of the potential advising function of the modal idiom.
(14) Because really you've got to look at attractiveness and user-friendliness… (CON) Compared to strong modals, weak necessity modals are almost equally present in the data. Being less forceful, should and ought to are used if the speaker admits the possibility that the event may not actualise (Palmer, 2001: 73), i.e. in the context of "escapable obligation" (Depraetere & Langford, 2020: 289). Judging from the examples below, in EfT, all three weak necessity modal expressions are used by speakers to point to the best behaviour to follow and simply convey advisability meaning. Although the advice containing had better may carry the implication that "unpleasant consequences may follow if it is not taken" (Palmer, 1990: 82), it can be claimed that the three are almost freely interchangeable in CON. Some restrictions in use may be imposed by a register as, in VISL, should is the only modal used as a device to give advice (Manca, 2016: 127) and issue "a strong recommendation" to the potential customers (Đurović & Silaški, 2015: 189).
(15) …you really should book by calling the education department here. (CON) (16) …maybe you ought to have a look at Museum Island... (CON) (17) OK, well you'd better stick to the most obvious differences… (CON) Considering the number of available examples, it is hard to associate the uses of these modals with specific sources of necessity. It seems, however, that any can be used when a discourse-internal source is involved. Only should is found as a weaker version of must associated with rules and regulations, as in (15) wherein really increases the pragmatic strength of the utterance, or with circumstances behind the necessity. Also, should occurs as the only weak modal used for reference to the necessary past situations that did not actualise.

PEDAGOGICAL IMPLICATIONS
The desirable general outcome of EfT instruction is to equip learners to function efficiently in the prospective discourse community as participants in successful interactions with customers, co-workers and other stakeholders in the industry. This subsumes the awareness of the language items, including modality markers, appropriate for the given context. Maci and colleagues observe that teaching EfT "requires a shift to a range of specific language skills mostly based on understanding diverse types of discourse in varied language situations and contexts" (Maci et al., 2018: 1-2). Instruction should reflect the specificity and complexity of tourism discourse, which implies addressing the differences in lexical and grammatical items in terms of the probability of their uses in different registers/genres. The findings presented above may thus have practical implications. EfT teachers could find them indicative of the amount and quality of the input provided either in teacher talk or in teaching materials. The findings may, on the one hand, inform teachers' decisions as to which modal expressions of necessity to teach to their students for reception rather than production, and, on the other, influence their choices about the approaches which would help students acquire productive knowledge of pertinent modals.
Much in line with an already observed practice (Larsen-Freeman et al., 2016: 138), EfT textbooks tend to present necessity modals from a speech act perspective (e.g. advice and suggestions [Walker & Harding, 2009]). As this approach may result in a fragmented view of the items in question (Larsen-Freeman et al., 2016: 138), the engagement of ESP teachers in enriching teaching materials is strongly recommended. Rather than following a traditional sentencebased approach, materials should favour discourse-driven treatment of modal verbs based on authentic texts. The texts investigated here seem well suited for this purpose. As online responses to reviews provide a good source for epistemic modals, they can be used to draw students' attention to the basic distinction between the epistemic and the root uses of must and should. The excerpts from Vol. 8(2)(2020): 275-296 destination websites may serve as a viable means to raise students' awareness of the subtle differences between root meanings of should, must and have to. Students should be made aware, directly or indirectly, of the underlying reasons for the choices of specific modals and the ways they shape the genre/register they occur in. The textbook transcripts may be used to enhance students' understandings of the differences in the use between spoken and written registers. Students, for instance, can be prompted to observe the uses of must and have to in conversation and discuss the differences and similarities in speech and writing. Special attention should be given to the factors surrounding the appropriateness of necessity modals and their (im)possible overlap in different registers. Fruitful discussions could help students reach conclusions about the constraints imposed on modals by genres/registers. Although EfT instruction revolves around communicative activities, widely used types of practice exercises (e.g. filling in the blanks, correcting the mistakes in the sentences, rewriting sentences using modals) need not be entirely neglected. As they are typically related to decontextualised usage, they could be modified to fit the discourse-driven approach. For instance, a teacher may design a gap-fill activity with necessity modals which will, instead of isolated sentences, contain short texts illustrative of different tourism registers.

CONCLUDING REMARKS
This study has attempted to offer an account of necessity modals across three registers in EfT. The findings suggest that the differences in the distribution are less related to the semantic nuances than to the appropriateness of modals in particular registers. The analysis has shown that the use of investigated items is contingent on the medium of communication, its purpose and participants. It might be claimed that systematic patterns of choice in different registers of EfT can be observed and some are summarised below.
Necessity modals have a very low incidence of epistemic uses. Whereas should and must encode both epistemic and root necessity in conversation and online responses to guests' reviews, in the promotional register they occur only as root modality markers. The use of both central modals is related to directives, but with different functions. Must is employed to issue instructions (rules and regulations) the potential customers should follow, whereas the instances of should are closely associated with the persuasive character of this register. Also, persuasion underlies the uses of the alternative strong necessity modal, have to, yet strictly restricted to the negative contexts.
The probability of occurrences of the investigated items varies considerably according to the medium of communication. Whereas in written registers the strong preference is given to central modals, in conversation, the modals with overlapping root meanings are almost freely interchangeable. As regards must and have to, the findings indicate that both can equally be used in sentences with Vol. 8(2)(2020): 275-296 different sources of root necessity. However, have to has the potential to overlap with must in conversation only, whilst there are salient differences in written registers. Also, in conversation, the choice between the two is largely determined by the participants and the setting, whether tourism professionals address customers or co-workers.
These generalisations, however, need to be taken with caution due to the major limitation of study related to the corpus size and content. It is the relevance of CON that seems rather dubious because it contains the texts which, although being speech-based, are primarily intended for educational purposes. Further, except for occasional mentions, no attention has been given to the differences related to various spoken registers. As this is a preliminary study, more extensive research is needed to substantiate my findings. Nonetheless, the analysis has suggested some pedagogical implications, presented above. That being said, the study might be informative to EfT instructors. Overall, it has added to the research on the English modals in professional discourses, on the one hand, and to the linguistic studies in tourism genres/registers, on the other. Hopefully, the study has made some contribution in promoting register-based analyses in the field of EfT and paved the way for future research which could explore the registerinduced variation of other modal and lexical categories in tourism discourse.