1. Introduction
Previous research on English modality has focused extensively on traditional modal and semi- modal verbs (e.g. can, could, may, might, must, will, would, etc.). Notable among the heterogeneous studies on the topic are works such as Palmer (Reference Palmer1990), Lyons (Reference Lyons1995), Collins (Reference Collins2009a) and Narrog (Reference Narrog2012), illustrating research often rooted in grammaticalization theory. More recently, attention has shifted towards analysing these modals through the lens of Construction Grammar, as seen in Hilpert et al. (Reference Hilpert, Depraetere and Cappelle2021). However, to our knowledge, no study has so far been conducted on the emerging be having to (BHT) construction comparing it with the semi-modal have to, shown in (1) and (2), respectively.
Building on the assumption that no two constructions can be exactly ‘equivalent’ semantically or pragmatically in a constructicon (Leclercq & Morin Reference Leclercq and Morin2023), we intend to investigate how the functions of have to and BHT vary and the need for the emergence of the latter in a modal system where the progressive aspect usually has no reason to exist. Traditional accounts of modality, such as van der Auwera et al. (Reference van der Auwera, Kehayov, Vittrant, Hogeweg, de Hoop and Malchukow2009: 287), claim in fact that since core modal verbs originate historically from expressions of might, knowledge – or possession, in the case of have to – they are stative and therefore incompatible with the progressive.
In the article, the following three Research Questions (RQs) will be addressed:
RQ1. How does the semantic pole of BHT vary from that of have to?
RQ2. What triggered the emergence of BHT?
RQ3. Why is BHT more productive in British English than in postcolonial varieties of English?
Section 2 will provide a brief account on past studies on have to and its diachronic evolution. Section 3 will describe the corpora and the methodology used for the analysis of the data. Section 4 will present the results of the corpus investigation through statistical analyses comparing the distribution of the two constructions, and section 5 will open a discussion aimed to answer our RQs. First, we will focus on what distinguishes BHT semantically from have to, while also considering ongoing changes in the aspectual meaning of the progressive itself. Second, we propose the hypothesis that the emergence of BHT may be motivated by a functional gap in a modal system wherein highly grammaticalized modal constructions are overtaking less grammaticalized dynamic meanings. Third, the hypothesis of negative retentionism (Ziegeler Reference Ziegeler2024) is presented to elucidate the restricted use of BHT in data from postcolonial varieties of English. Section 6 concludes the article.
2. Have to: a historical account
It is generally claimed that in present-day English have to conveys meanings related to obligation and necessity. Numerous studies have aimed to investigate how lexical have came to acquire modal meanings historically (e.g. Krug Reference Krug2000; Łęcki Reference Łęcki2010). One of the earliest accounts on the topic can be found in van der Gaaf (Reference van der Gaaf1931: 180–4), who identifies three diachronic stages corresponding to a gradual reanalysis of have from possession to obligation. In stage A, the verb still retains a meaning of possession, with a noun phrase (NP) functioning as its direct object, as shown in (3). At this point, the infinitive following the construction is dependent on the NP and serves as an adjunct. In stage B (4), have co-occurs with the infinitive, leading to both a possession and obligation interpretation. In stage C, have has lost its original lexical meaning and now expresses only obligation and duty, as in (5), where the fact of buying something cannot entail its possession. Have to + V operates as a sole construction conveying a sense of obligation, with the original NP functioning as the argument of the infinitive. The semantic reanalysis ultimately leads to a word order change with the infinitive verb preceding the object NP (cf. also Basile Reference Basile2024: 72).
Van der Gaaf suggests that, although these three uses were already productive in Old English times, the shift in word order leading to a post-infinitival object did not emerge until Middle English (ME) times (Reference van der Gaaf1931: 184). This assumption was later challenged by Mitchell (Reference Mitchell1985), Brinton (Reference Brinton1991) and Fischer (Reference Fischer1994), who argue that the examples van der Gaaf used to demonstrate the three stages are ‘indistinguishable on the surface’ (Fischer Reference Fischer1994: 138). Later accounts sought indeed to offer more fine-grained analyses, addressing both the motivations behind the word order change and its semantic consequences. Fischer (Reference Fischer1994, Reference Fischer1997), for instance, supports a syntactically led grammaticalization hypothesis. She argues that the word order shift from SOV to SVO in ME times resulted in the placing of have next to the infinitive, thus triggering further the semantic reinterpretation from possession to necessity (Fischer Reference Fischer1994: 137, as cited in Basile Reference Basile2024: 73). Ziegeler (Reference Ziegeler2010) offers an alternative perspective by suggesting a semantically driven grammaticalization hypothesis. She contends that Fischer’s explanation falls short of clarifying why the older word order, as seen in examples (3) and (4), remains possible in modern usage. If the shift in word order were responsible for the modal interpretation of have to, it is unclear why this pattern did not apply generally to all constructions involving have and to. Drawing on Bybee et al. (Reference Bybee, Perkins and Pagliuca1994), Ziegeler argues (2010: 51) that the sense of obligation in languages typically arises either from sources related to negative permission or from lexical contexts contributing to a specific interpretation (what she calls ‘lexical optimality’). Lexical optimality, for example, contributed to the emergence of modal ought to (cf. Dekeyser Reference Dekeyser, van der Auwera, Durieux and Lejeune1998). Ziegeler dismisses the negative permission hypothesis, which typically involves a shift from negative possibility/permission to necessity in increasingly negative environments – a pattern not reflected in her data. This is further supported in Brinton (Reference Brinton1991), who notes that negative constructions with have to, such as those in (6), have preserved the original SOV word order in modern English rather than undergoing change. Ziegeler (Reference Ziegeler2010: 52) views lexical optimality as a better explanation, since the construction began to express meanings of possession linked to social duty well before the shift in word order occurred. This is demonstrated in examples (7) and (8) (both from van der Gaaf Reference van der Gaaf1931: 183), where the writer conveys ‘a generic or habitual modal reading’ associated with ‘ongoing customs and duties’ and not simply to possession (Ziegeler Reference Ziegeler2010: 52). This interpretation is clear in example (7), where the subject (the vicar) cannot possess the souls of the parishioners. Instead, he has a duty to fulfill, making the concept of future accomplishment central to the meaning.
However different, past accounts of have to all show that the construction originated from a meaning related to possession, which subsequently grammaticalized to express social duties and, eventually, obligation. Examples (7) and (8) show that after the original meaning of possession, have to started to be used in ambiguous contexts, conveying the meaning of customary and social duty/necessity. The source of necessity, in this case, lies external to the subject referent, who has to accomplish something due to general conditions, making the reading associated with ‘dynamic’ modality. The category of dynamic modality, built upon Palmer (Reference Palmer1990), Collins (Reference Collins2009a) and Basile (Reference Basile2024), corresponds indeed to those contexts where the speaker/writer does not impose any obligation. It also ‘tend[s] to have generic subjects […] [and has] to do with circumstances, rules, regulations, physical or natural laws, etc.’; it may also ‘refer to internal need dictated by external conditions as long as there is no deontic source involved’ (Basile Reference Basile2023a: 285). Another example is in (9), where the accomplishment of the state of affairs is not imposed on the subject referent by any deontic source. The described necessity refers instead to a general situation or circumstance, which – accordingly – does not have a future-projecting value.
Similarly to must in the history of English, have to subsequently grammaticalized to ‘deontic’ and later ‘epistemic’ functions (cf. Gregersen Reference Gregersen2020; Basile Reference Basile2024, inter alia). Deontic modality is ‘concerned with influencing actions, states or events and expressing what Searle (Reference Searle1983: 166) calls “directives”’ (Palmer Reference Palmer1990: 6). Differently from dynamic modality, the time reference is future-projecting, as the necessity behind the realization of the state of affairs is to be read with a future perspective from the moment of utterance (cf. Bybee et al. Reference Bybee, Perkins and Pagliuca1994: 185). An example is given in (10) below, where the speaker imposes their will and directs the subject referent to act in order to realize the state of affairs in a more or less close future.
Epistemic modality (i.e. the kind of modality that conveys the degree of probability of the truth of a proposition) associated with have to is more recent. The Oxford English Dictionary (OED; s.v. have, v., VII.ii.45) dates its first occurrence (11) back to approximately 1829. Denison (Reference Denison and Romaine1998: 172) provides some examples, such as (12), from the second half of the twentieth century. Nevertheless, epistemic readings of have to remain infrequent in present-day English, with must supplying the most common epistemic alternative (see Palmer Reference Palmer1990; Krug Reference Krug2000; Collins Reference Collins2009a, Reference Collins2009b).
As will be shown in the following sections, this historical evolution and grammaticalization of have to lays the groundwork for investigating the productivity and functional competition of BHT, i.e. its progressive counterpart.
3. Data and methods
The first stage of our investigation aimed to assess the productivity of the overlooked BHT construction and shed light on its origins. The main analysis was conducted using the Corpus of Global Web-based English (henceforth GloWbE; Davies Reference Davies2013; Davies & Fuchs Reference Davies and Fuchs2015), a collection of about 1.9 billion words of texts retrieved from web pages and blogs.Footnote 2 The corpus is made of 20 subcorpora each for a different variety of English. Since our study also aimed to statistically quantify the frequency of BHT geographically (RQ3), analyses were carried out on two of Kachru’s (Reference Kachru1986) ‘inner-circle’ varieties (i.e. British English and US English) and on three ‘outer-circle’ or postcolonial varieties (i.e. Singapore English, Hong Kong English and Indian English). These subcorpora differ in size, so the results on the general frequency of BHT in each variety were normalized per 10,000 words to ensure comparability.
BHT was found to be particularly productive in British English (see section 5), making this variety the central focus of our investigation. However, the sheer number of occurrences in GloWbE made it impractical to perform a qualitative analysis within such a large corpus. To overcome this challenge, we turned to the British National Corpus (BNC 1995), a more manageable dataset compiled between the 1980s and 1990s, consisting of 100 million words across various genres. For the analysis of have to in the same variety, we used the first 180 texts from the spoken subsection of the International Corpus of English – Great Britain (ICE-GB), which was compiled around the same time as the BNC and comprises approximately 360,000 words. Additionally, to explore the historical productivity of BHT in British English during the period 1710–1920, corresponding to the colonization period of regions such as India, Singapore and Hong Kong, we analysed the Corpus of Late Modern English Texts (CLMET, version 3.1; De Smet, Flach, Diller & Tyrkkö Reference De Smet, Flach, Diller and Tyrkkö2011; cf. Diller et al. Reference Diller, De Smet and Tyrkkö2011), which includes roughly 34 million words.
Finite occurrences of have to and of BHT (in the forms am/is/are having to) were retrieved and categorized semantically, using the text analysis software available at English-corpora.org and Sketch Engine (version 2.36.5; Kilgarriff et al. Reference Kilgarriff, Baisa, Bušta, Jakubíček, Kovář, Michelfeit, Rychlý and Suchomel2014). The purpose was to examine the functional competition between BHT and have to as modal candidates. Non-finite forms such as (13) and instances in past or future tenses (e.g. had to, was/were having to or will be having to/will have to) were thus also excluded.
This exclusion was driven by the observation of the comparatively smaller rates of productivity of BHT in such contexts (e.g. less than 0.003 occurrences of were having to per 100,000 words in GloWbE-GB).
Each modal occurrence was annotated according to three groups of variables, aiming to identify statistically significant trends between the two constructions. First, we annotated the type of functional category expressed by the construction (i.e. dynamic, epistemic, deontic), each being a potential indicator of a more or less advanced level of grammaticalization, as shown in section 3. Second, we coded the grammatical subject (i.e. I, you, he, she, it, we, they). Modal constructions with first- or second-person subjects are in fact often said to reflect greater personal involvement and imposition (i.e. deontic readings) compared to those with other types of subjects (cf. Leech et al. Reference Leech, Hundt, Mair and Smith2009: 114). Third, we annotated whether the semantic reference to the expressed subject was generic (G) or non-generic (NG). Non-generic reference invokes a specific, identifiable entity or set of entities, rather than an abstract category, as in (10) above, where the noun NP serving the function of subject (i.e. you) refers to a specific and concrete entity. On the contrary, generic reference invokes the whole class of entities in general, without indexing one entity in particular. In (9), the subject has generic reference as it denotes the generic group of people who are confined to wheelchairs, from which the speaker does not distinguish any submembers. The semantic category of dynamic, as we will see, tends to prefer subjects with generic reference. It is, in fact, hard to generate deontic modality with a subject that conveys some generic reference (cf. Basile Reference Basile2024: 133).
4. Results
The analysis of the geographical components of GloWbE reveals a significantly different distribution of BHT across regions, as illustrated in figure 1. The figure suggests that the use of BHT is more productive in the data from inner- than outer-circle varieties of English. The higher frequency of production of the construction in British English is easily recognizable, being almost 2.5 times as frequent as in the US English data, where the construction has a frequency of 0.02 per 10,000 words. British English seems therefore to have a leading role. Nonetheless, the frequency of BHT in British English remains much lower than that of have to (0.02 instances per 10,000 words for BHT in the BNC, compared to 7.2 instances per 10,000 words for have to in ICE-GB).
Frequency (normalized per 10,000 words) of BHT in GloWbE
The annotation of the examples of BHT in the BNC suggests that all the occurrences of this construction express dynamic modality, and 54 per cent of the total have generic subject reference. With the aim of comparing BHT with have to as modal contestants, a binary regression model was computed, following the formula in (14).
The formula calculates the impact of subject type (i.e. I, you, he, she, it, we, they) and subject reference (i.e. ‘gen’ in glm1) on the choice of BHT and have to. The variable of functional category was not included, considering the exclusivity of dynamic modality associated with BHT, as observed in the manual annotation. The model’s explanatory power is substantial (Tjur’s R2 = 0.36). The model’s intercept, corresponding to subj = he/she and gen = G (i.e. generic), is at -0.42 (95 percent CI [-1.17, 0.32], p = 0.268). The results of the logistic regression model are shown through a forest plot in figure 2. The plot illustrates significant patterns as far as the choice of different subject types is concerned. The logits suggest indeed that first and second person subjects (i.e. you, I, and we) are significantly more often associated with have to. BHT, on the contrary, seems to be associated more frequently with third person plural subjects, including full NPs (with generic reference in 75 per cent of cases according to the annotation).
Log odds for using have to in opposition to BHT according to the independent variables of subject type and subject reference (data for BHT: BNF, data for have to: ICE-GB)
Examples of generic they subjects are given in (15) where the category of librarians serves the function of subject, and (16), where the writer refers to a generic sub-group of people, namely ‘men and women discharged from hospital’.
The results of the significant observations of glm1 are summarized in table 1.
Statistical results of the statistically significant predictors in glm1
A diachronic analysis was also conducted to determine when BHT first emerged in the history of British English. According to the OED (s.v. have, v., VII.i.42.a), (17) constitutes the first registered example of BHT, dating back to 1871.Footnote 3 In the example, the writer – J. Darby – describing life in the countryside, reports that rafters, intended as a third-person-plural subject with generic reference tied to the phrase ‘a flat roof’, have to carry a heavy weight when it snows. The functional category associated with the modal occurrence is dynamic as the necessity is presented as a fact that arises in general circumstances. The use of an inanimate subject with the construction appears as an exception in the data for BHT. As a result, it is likely that the author attempts to personify the subject for stylistic effects (cf. ‘the rafters are suffering’).
The CLMET dataset registers only one occurrence of BHT from 1920, shown in (18). The limited productivity of BHT during that time frame further suggests that the appearance of BHT should be regarded as a recent phenomenon.
5. Discussion
Based on the results of the corpus investigation, this section aims to address the central research questions of the present study.
RQ1. How does the semantic pole of BHT vary from that of have to?
The BHT construction stands out for its formal combination of a modal root and the marking of the progressive aspect (i.e. be V-ing), a pairing that is atypical for modal constructions in English. Aside from have to, our data exploration suggests there are few – if any – other modal or semi-modal constructions within the English domain of necessity that are typically used in the progressive. A quick search on GloWbE for progressive forms like are needing to + V reveals an extremely low frequency – fewer than 0.05 occurrences per million words.Footnote 4
Cross-linguistically, van der Auwera et al. (Reference van der Auwera, Kehayov, Vittrant, Hogeweg, de Hoop and Malchukow2009) claim that when an acquisitive verb grammaticalizes into a (necessity) modal, it may carry with it the meaning of temporariness that an original modal verb is lacking, since modal verbs are permanent statives in the first place. It may be fairly assumed then that the use of the progressive with an acquisitive verb such as have makes the BHT construction doubly loaded with a sense of temporariness.Footnote 5
The incorporation of the progressive in the BHT construction seems to reinforce the aspectual feature that is under the scope of the modal verb, while simultaneously enhancing the modal meaning of necessity with shades of intensity (cf. Brisard & De Wit Reference Brisard and De Wit2013). It is in fact not uncommon to encounter in the data examples like (19) which incorporate adverbs of intensity such as really. These uses seem to emphasize the lack of control that the subject has over the state of affairs and their potential unwillingness. The agents, often represented by third-person-plural subjects (as indicated by the logistic regression analysis), feel obliged to comply with external necessities, despite their reluctance.
This is also manifest in examples including lexical verbs that would usually entail that the subject has control over a situation, as with the verb decide. Example (20) shows that the choice the subject referent has to make is still in contrast with their desires. The very fact of being forced to choose between the two scenarios (turning the heat on or buying shoes for their children) is a result of external circumstances that are outside of their control. It is therefore not surprising that, as discussed in section 4, all the retrieved occurrences of BHT in the data convey a sense of dynamic and not deontic necessity.
This is not necessarily the case for have to. As Sweetser claims (Reference Sweetser1990: 54), have to typically conveys a type of obligation that is ‘resistible’, enabling the production of utterances such as (21), where the speaker decides to prioritize their desires over the imposed necessity.
The sense of control over the realization of the state of affairs conveyed by have to is absent in BHT. In fact, using BHT in (21) would most likely sound odd. A more illustrative contrast between have to and BHT in this sense is given in (22).
In this example, the two modal constructions seem to express different levels of necessity. The use of BHT suggests that selling their homes is a requirement that is being imposed against the pensioners’ will. The use of have to appears instead to be more general and neutral. The modality expressed is also dynamic, but its use seems to entail that relying on assistance provided by the state is not as burdensome as the act of selling their properties.
BHT appears to be associated with a combination of three potential semantic features, that is, (a) inchoativity, (b) contingency (in Comrie’s Reference Comrie1976 terms) and (c) adversity.
(a) Inchoativity . This term corresponds to the morphological expression of elements indicating the beginning of a situation or entrance into a state (Bybee Reference Bybee1985: 147–50). In (22), for example, the subject referents (‘the elderly’) are faced with a situation that they wish they could escape, especially since they find themselves having to do something that they did not have to do before. Such a change of state is overtly expressed in the previous context by the adverbial now, which is itself reinforced by the intensifier especially. The internal point of view of the imperfective in BHT can be seen as reorienting the typical future-projecting obligation of have to to the present. Hence, the functional need expressed by BHT in (22) to suggest that the present pension system, which used to be efficient, is not anymore.
(b) Contingency . Relying on Comrie (Reference Comrie1976), who posits that imperfectivity subsumes habituality, continuousness and progressivity, the progressive marker in BHT is, among other features, indicative of a potential duration (1976: 41–4). Comrie (Reference Comrie1976: 38) notes that the basic aspectual meaning of the progressive would be better defined as the denotation of a contingent situation, which itself comprises progressiveness, ‘temporary (contingent) states’ and contingent habitual situations. Comrie clearly seems to equate the meaning of the adjective contingent with that of temporary (see also Comrie Reference Comrie1976: 50).Footnote 6 Referring to a personal communication by Lyons, the same author tentatively speculates that the dynamicity of the progressive might as well convey a meaning of unexpectedness, surprise or even unreality (Comrie Reference Comrie1976: 37, 49).
(c) Adversity . As shown in (19) above, the BHT construction signals a lack of agentive control, in contrast to have to in (21). This absence of control may be linked to meanings of adversity often associated with the use of stative verbs in other progressives (cf. Ziegeler Reference Ziegeler, Van Olmen, Cuyckens and Ghesquiere2017: 331 and Ziegeler & Lenoble Reference Ziegeler, Lenoble, Núñez-Pertejo, López-Couso, Méndez-Naya and Pérez-Guerra2020: 240, 256). Given that BHT is associated with the start of new situations (inchoativity) and that the unfolding of these states of affairs is beyond the control of the subject referent, it is not surprising that occurrences of BHT are often accompanied by a sense of inevitability or adversity.Footnote 7 Example (23) further shows the strong link between inchoativity/adversity and BHT, with the contrast between the simple past (‘created’) and the progressive, along with the adverb now reinforcing the inchoative meanings and demonstrating the change of state that the referent is made to comply with.
Interestingly, the emergence of BHT may also provide evidence of the diachronic shift from durativity meanings (in Old English) to meanings of temporariness in present-day progressives (Ziegeler Reference Ziegeler1999: 84–5). The extension in question has been recurrently addressed in the literature with a special focus on semantic domains such as temporariness (Comrie Reference Comrie1976; Smith Reference Smith1983; Huddleston & Pullum et al. Reference Huddleston and Pullum2002), markedness (Smith Reference Smith1991), vividness (Smith Reference Smith1983), semantic shifts (Hirtle & Bégin Reference Hirtle and Bégin1991), waxing and waning cases (Huddleston & Pullum et al. Reference Huddleston and Pullum2002), tentativeness (Huddleston & Pullum et al. Reference Huddleston and Pullum2002), colloquialization, conventionalization and hedging (Mair Reference Mair and Binnick2012), cognitive construals of situations (Langacker Reference Langacker2002), and the persistent ongoing grammaticalization of the progressive is scrutinized separately within the class of statives (cf. Martínez-Vásquez Reference Martínez-Vásquez2018; Rautionaho & Fuchs Reference Rautionaho and Fuchs2021, among others). A notable example is the McDonald’s slogan, I’m loving it, which is the focus of the latter scholar’s paper. This example appears to reflect the contingency of the situation in Comrie’s terms, rather than encoding non-obviousness or subjectivity, as suggested by De Wit et al. (Reference De Wit, Petré and Brisard2020: 502).
The present article thus argues that the temporariness inherent in the meaning of the progressive underlies its extended, but still aspectual, uses with stative verbs, habitual progressives, but also adversative meanings: a grammatical marker expressing temporariness focuses on the limited duration of a new (inchoative) situation of adversity (cf. also Ziegeler & Lenoble Reference Ziegeler, Lenoble, Núñez-Pertejo, López-Couso, Méndez-Naya and Pérez-Guerra2020: 256; Lenoble Reference Lenoble2021: 11). The extension of the progressive to (semi-)modal constructions, such as BHT in British English, illustrates just how far the use of the progressive with statives has grammaticalized. The contrast with the normal state of affairs that the progressive generally indicates is undeniably a defining feature of BHT as well.
Table 2 summarizes some of the traits that seem to set BHT and have to apart. On the one hand, BHT appears to enhance the modal meaning of necessity with nuances of temporariness and adversity. On the other hand, it incorporates the notion of inchoativity, transforming the future-oriented meanings associated with deontic have to, to focus instead on the modal necessity of the current situation. This reoriented obligation sense persists regardless of the subject’s intention, indicating a state beyond their direct control.
Summary of the main properties of have to and BHT
RQ2. What triggered the emergence of BHT?
Building on the arguments made in the previous section, it can be concluded that BHT and have to differ not only in form but also in their semantic/pragmatic functions. This distinction follows the principle that there is no true equivalence within constructions, namely that ‘each construction [in a constructicon] occupies a unique functional niche (be it at the semantic, pragmatic and/or social level)’ (Leclercq & Morin Reference Leclercq and Morin2023: 11), as if two constructions were semantically identical, their co-existence would lead to a ‘level of redundancy that would be intolerable in any natural language’ (Collins Reference Collins2023: 8). The origins of BHT remain however unexplained, and more precisely what motivated the introduction of BHT in the English modal system.
Past studies on modality have shown that have to has been renewing the functions of the declining core modal must (cf. Krug Reference Krug2000; Mair Reference Mair2006, Reference Mair and Collins2015; Collins Reference Collins2009b; Leech et al. Reference Leech, Hundt, Mair and Smith2009; Leech Reference Leech2013; Mair & Leech Reference Mair, Leech, Aarts, McMahon and Hinrichs2020, inter alia). The semi-modal has therefore gradually taken the functions that were previously typically associated with must, and specifically the expression of deontic modality. The gradual replacement of must with have to is likely to have triggered a reconfiguration of the modal have to in inner-circle varieties of English, leading to an acceleration of its grammaticalization. The historical reconstruction of the semi-modal in section 2 has shown that it has in fact grammaticalized from an original meaning of possession to dynamic and later deontic meanings, which, by definition, have future-projecting connotations. In this sense, it is possible to draw a parallel between the recruitment of have to for the expression of deontic meanings and the future tense in Latin. As Hopper & Traugott note (Reference Hopper and Traugott2003: 53), Latin used to have synthetic constructions that conveyed information about the person, the number, and the tense, including one construction for the future. This construction was, however, gradually replaced by a periphrastic future construction made of an infinitive alongside a conjugated form of the verb habeo ‘to have’ in the present tense, as in (24).
In a similar fashion to have to, before undergoing further grammaticalization, habeo served as a transitive verb, denoting possession or mere belonging. It would also fulfil future- projecting functions of obligation in certain contexts, possibly influenced by the Latin gerundive, which diachronically also used to convey a sense of obligation (Hopper & Traugott Reference Hopper and Traugott2003: 53).Footnote 8 Example (25) shows this possible double interpretation. Indeed, the fact of owning a house may actually involve the need for its maintenance in the relatively near future.
The future projection meanings became more prominent with time and contributed to the reinterpretation and reanalysis of the construction. This lies at the root of the future tense in Romance languages, as in the French form je chant-e-r-ai (‘I will sing’) that lost the original periphrastic properties of Latin (cf. also Gisborne Reference Gisborne, Gisborne and Hippisley2017).
The grammaticalization of a verb denoting possession into a marker of future tense is not unique to Latin. As Heine & Kuteva (Reference Heine and Kuteva2002: 242–3) point out, have-type predicative possession is also used to indicate future tense in other languages, including Godié & Nyabo (cf. Marchese Reference Marchese1986). In Bulgarian, for instance, the verb ima functions both lexically and grammatically, as illustrated in examples (26) and (27).
As discussed in section 2, much like Latin, the English have to progressed from a former lexical meaning of possession to a derived deontic meaning of necessity and obligation. Since, by definition, deontic modality has been considered to be strongly related to the realization of an event in the future, the future meaning remains pragmatically implied in the construction. This construction may undergo a further transformation, shifting from primarily expressing obligation to functioning as a proper future marker. The ongoing reanalysis of the construction could be intensified by the formal properties it exhibits. In English, the preposition to is commonly associated with purposive meanings, which can also carry future implications (cf. Haspelmath Reference Haspelmath1989). Notably, when these purposive meanings are bleached the notion of futurity becomes more salient (Hopper & Traugott Reference Hopper and Traugott2003: 3). This pragmatic implicature leading to the ‘presupposition of a future event’ (Hilpert Reference Hilpert2008: 109) has in fact also been identified as one of the factors contributing to the reanalysis of going to as a marker of futurity in English (Hopper & Traugott Reference Hopper and Traugott2003: 3). The fact that the agent is moving on a path towards a goal enables its reinterpretation as a future tense (Bybee et al. Reference Bybee, Perkins and Pagliuca1994: 268) also leading to a pragmatic inference of prediction (cf. Budts & Petré Reference Budts and Petré2016). Accordingly, the historical reinterpretation of have to driven by grammaticalization might follow the trajectory shown in (26) (cf. also Basile Reference Basile2023b).
Faced with the acceleration of the grammaticalization of have to strengthened by the renewal of the functions of must, the emergence of BHT might be explained as a systemic strategy to ‘reinforce’ (cf. Lehmann Reference Lehmann2015) the declining dynamic functions of its counterpart have to, with a construction that is in fact unambiguously dynamic (C), and – at least in our datasets – never deontic (D). Yet it is essential to approach this argument with caution. The emergence of BHT is a relatively recent phenomenon, and fine-grained investigations based on the productivity of have to as a pure future marker bleached of much of its obligation meaning necessitate further attention in the years to come.
RQ3. Why is BHT more productive in British English than in postcolonial varieties of English?
A final question concerns the observed different distribution of the construction in British English compared with other varieties. Examples (17) and (18) above show that the emergence of BHT is recent and most likely subsequent to the colonization period of areas such as India, Singapore and Hong Kong. The lack of input containing BHT from British settlers in these areas at the time may thus account for the minimal production of BHT in Singapore English, Indian English and Hong Kong English today. This hypothesis follows the notion of ‘negative retentionism’. This is a term introduced in Ziegeler (Reference Ziegeler2024) essentially to explain the converse of what has been variously described in the literature as ‘colonial lag’ (see Marckwardt Reference Marckwardt1958; Trudgill Reference Trudgill1999; Hundt Reference Hundt, Rohdenburg and Schluter2009), ‘retention’ (usually of the positive kind) (Siemund Reference Siemund2004; Pietsch Reference Pietsch2009) and ‘feature retention’ (Fuchs Reference Fuchs, Rautionaho, Parviainen, Kaunisto and Nurmi2022). In the (positive) description of retentionism, it is the case that a feature which was once used in the history of the model lexifier language (in our case, colonial English), but has become obsolete in the grammar of the original lexifier, continues to be used in the contact variety until the present day. Examples include the Irish English medial perfect (e.g. Pietsch Reference Pietsch2009), the auxiliary do in habitual aspect in Atlantic creoles (cf. Ihalainen Reference Ihalainen, Trudgill and Chambers1991), Kortmann (Reference Kortmann2004), and the retention of an archaic conjunction in Indian English (Fuchs Reference Fuchs, Rautionaho, Parviainen, Kaunisto and Nurmi2022), all of which have been argued to be also influenced by local substrate languages. In terms of grammaticalization, it may be argued that for one reason or another, the retained feature has grammaticalized further in the colonial lexifier language than in the contact varieties, where it stays frozen in time.
In the case of negative retentionism, there is a clear correlation shown between the frequency of a form in a particular variety of English, and the time at which it began to appear in any variety, so that the more recently it is reported as having emerged in an inner-circle variety, the less frequently it is seen to be distributed across other, outer-circle varieties, as we can see with BHT as shown in figure 1 above. Negative retentionism suggests, then, that if a feature was not present in the input English varieties at the time of contact, it will lag behind in its later uptake in outer-circle varieties. The evidence for this has already been shown in Ziegeler et al. (Reference Ziegeler, Mélac and Gast2019) and Ziegeler (Reference Ziegeler2024) in the case of the sentence-final additive adverb either, which appears with much less frequency in Singapore English than in British English. The infrequent use of the adverb was partly attributed to its infrequency in the ancestral, input dialects at the time of historical contact, when, although its use was not absent altogether, it was infrequent in the scope of negation.
While not linking her results with the phenomenon of retentionism at all, Hansen (Reference Hansen2018) provides the most salient evidence for negative retention in the English modal domain. She refers to Mufwene’s (Reference Mufwene1996, Reference Mufwene2001) ‘Founder Principle’, in which the formation of a creole is ultimately associated with the input dialects of the lexifiers at the time of contact. The present approach also accords with this principle as an important factor explaining the differences in frequency of BHT, but with the additional proviso that features appearing in the source dialects later than the time of contact will be adopted more slowly in dialects distant from the varieties in which they evolved. Hansen’s study of Singapore English, Indian English and Hong Kong English reveals that although the older necessity modals she studied, such as must and have to, showed little difference across the dialects, the more recently emerging ones such as have got to and need to were less frequent in postcolonial varieties than in British English, and, significantly, that their epistemic meanings were less grammaticalized than in British English as well, due to their more recent grammaticalization in the source variety. Her evidence points to a possible ‘epicentral’ approach to change across international varieties of a single language (discussed at greater length in Collins Reference Collins2020, for example). But more important is the fact that it was grammaticalization stages that were subject to ‘colonial lag’, and that the pace of gradual diffusion of grammaticalizing items across spatial domains can be clearly measured in the gradualness of diachronic stages. Negative retention regards Hansen’s (Reference Hansen2018) results as demonstrating the preservation of the actual state of the language system at contact time, and not simply a specific feature, so that what was absent at contact time will be also absent or less frequently attested at later times in the dialect’s development.
6. Conclusion
This article has aimed to shed more light on a previously overlooked modal construction and its emergence in the English modal system. Several key findings have emerged in the analysis. BHT stands out due to its incorporation of the progressive aspect, which affects its modal meaning, emphasizing contingency, intensity and a marked lack of control over the situation. It often conveys a sense of external necessity, imposed upon the subject, who must comply despite reluctance or a desire to resist. In contrast, have to expresses a more neutral form of obligation, often leaving room for the subject’s personal agency or choice. We have also shown that the progressive aspect in BHT introduces an element of inchoativity, signalling a transition or change in circumstances – often involving an undesirable or challenging situation. This reorients the temporal focus of obligation: whereas have to typically projects necessity into the future, BHT anchors it in the present, creating an immediate and ongoing sense of urgency. Crucially, these semantic features do not seem to be the result of a mere compositional process involving the progressive and have to. As mentioned in section 5, applying the progressive to other semi-modal constructions (which is itself rare in the data, cf. with need to) does not yield comparable readings of inchoativity, adversity, or contingency – underscoring the non-compositional nature of BHT.
A historical dimension has also been included in the study. We have argued that emergence of BHT may be linked to the accelerated grammaticalization of have to. Diachronically, have to has shifted from expressing possession to encompassing both dynamic and future-projecting deontic modal functions. This trajectory mirrors similar developments in other languages, such as Latin and French, where verbs initially signifying possession have transformed into future tense markers and gives food for thought to the destiny of English have to in the future.
The corpus investigation has also suggested that British English is the birthplace of this emerging construction, while other varieties of English are lagging behind in adopting it. This delay has been explained by the phenomenon of negative retentionism, which posits that because BHT emerged after the period of colonization, it was not present in the input varieties during the initial stages of language spread, resulting in limited use in these regions.
This research only scratches the surface of a complex linguistic phenomenon. BHT, being a relatively recent development, requires further investigation to fully understand its role and evolution in the English modal system. In fact, while this article provides a foundational understanding of BHT and its relationship to have to, there is still much to explore. Our study hopes to inspire future research into this neglected construction and its implications for the evolving English modal and progressive systems.
Open access
