Introduction

According to Halliday and Matthiessen (2014), language is a complex semiotic system composed of three strata: phonology/graphology, lexico-grammar, and semantics. In Systemic Functional Linguistics or SFL, the modal meaning in the semantic stratum is realized by expressions of modality in the lexicogrammatical stratum; this realization happens either within the clause (e.g., modal verbal operators)Footnote 1 or outside the clause (e.g., ‘explicit objective’ orientation) (Taverniers, 2003). Consider the following examples,Footnote 2 wherein expressions of modality are italicized.

  1. (1)

    ▓

    1. a.

      One consideration that might have played a part is a gender imbalance on the US governing board. (COCA_newspaper_2019)

    2. b.

      The sex differences in auditory acuity are possibly linked to the sex differences in vocal production in rats. (COCA_academic_2019)

    3. c.

      And a lot of folks are getting very excited about the possibility of phase one in the trade deal happening. (COCA_spoken_2019)

    4. d.

      I suppose I have to backtrack on that a little bit. (COCA_magazine_2016)

    5. e.

      Although the molecular determinants of dynein force production are not well understood, it is possible that these mutations specifically affect the ability of dynein to remain bound to microtubules under conditions of high load. (COCA_academic_2019)

Modal meanings in (1a–e) are realized by the modal verbal operator might in (1a), the modal adjunct possibly in (1b), the modal nominalization possibility in (1c), the ‘explicit subjective’ orientation I suppose in (1d), and the ‘explicit objective’ orientation it is possible in (1e) respectively. Expressions of modality in the first three examples (i.e., 1a–c) could be regarded as expressions of modality at word rankFootnote 3, realizing the modal meaning within clauses, while those in the last two examples (i.e., 1d-e) could be considered as expressions of modality at clause rank (or interpersonal metaphor of modality in this research), realizing the modal meaning outside clauses.

There are three lines of studies that have examined expressions of modality. The first line pertains to the restricted consideration of expressions of modality based on different theoretical traditions (e.g., Nuyts, 1993; Halliday, 1994; Palmer, 2001; Matthews, 2003; Charnock, 2009). For example, Nuyts (1993) in the traditional sense documented the difference between constructions with modal adverbs, i.e. modal adjuncts at issue, and constructions with predictively used modal adjectives, i.e. ‘explicit objective’ orientation in question, in terms of three different underlying factors: the discourse functionality of the constructions, the interaction of the epistemic modal qualification with evidential marking, and the performative vs. descriptive nature of the modal expression. In addition, specific adverbs in modal adjuncts, i.e. maybe and perhaps, were examined with respect to their use in the context (Suzuki, 2018) and the multifunctionality of the ‘possible’ modal adjuncts by comparing with conceivably with perhaps (Suzuki and Fujiwara, 2017). The former study uncovered that maybe is more prone to subjective use while perhaps is a more strongly grammaticalized item; the latter study demonstrated that conceivably and perhaps display opposite functional characteristics, and the factors influencing the use of these adverbs are strongly associated with the contexts of modality and discourse. Issues in this line are two-fold. First, investigation of subtypes of expressions of modality is restricted. Foci of these studies were either on modal verbal operators (Palmer, 2001), or on predictively used modal adjectives (Nuyts, 1993), or on modal adjuncts (Nuyts, 1993; Suzuki and Fujiwara, 2017; Suzuki, 2018). Second, dynamic nature of these expressions of modality needs to be addressed intricately. Variation or change is the nature of language (de Saussure, 1959; Zhou, 2023a), and thus expressions of modality are of no exception. Therefore, expressions of modality should be explored in the context of different registers.

The second line concerned the distribution of modal verbal operators both diachronically (e.g., Leech, 2011; Leech, et al. 2009) and synchronically (e.g., Biber, et al. 1999; Collins, 2009). In relation to the diachronic studies, Leech, et al. (2009) for example examined the frequency change of modal verbal operators (modal auxiliaries in their research) and documented that some core modal auxiliaries are losing their deontic and epistemic uses and gradually substituted by semi-modal auxiliaries in the process of grammaticalization (must – have (got)to – want to / need to). In relation to the synchronic studies, Collins (2009) explored the distribution of modals and quasi-modals in different varieties of English and uncovered that there is strong tendency for American English to be leading the way in the English language developments. Additionally, it is uncovered that quasi-modals flourish in speech while their modal counterparts are favored more in written registers. Apart from the synchronic investigation of the distribution of modals and quasi-modals, some other modal verbal operators such as should and ought to were also examined in terms of their distribution in present-day British English (Verhulst, et al. 2013). However, a fuller scope of expressions of modality, inclusive of both subtypes of expressions of modality at word rank and those at clause rank, is not systematically concerned, nor is the influencing factor of registers.

The third line of research documented the distribution of expressions of modality in English across registers (e.g., He, 2020; Zhou, 2023a, 2023b, 2023c). For example, He (2020) compared normalized frequencies of specific subtypes of interpersonal metaphor of modality in different registers. It is uncovered that ‘explicit objective’ orientation is register-sensitive and occurs preferentially in formal registers such as academic texts, while ‘explicit subjective’ orientation occurs preferentially in informal registers such as spoken texts and fiction. However, he limited his study within modal verbal operators, modal adjuncts, ‘explicit subjective’ orientation, and ‘explicit objective’ orientation (i.e., congruent and metaphorical expressions of modality in his study) and left modal nominalizations untouched. In addition, he examined the diachronic distribution of these expressions in such registersFootnote 4 as fiction, magazines, newspapers, and non-fiction, leaving the academic register which preferentially accommodates metaphorical expressions (i.e., the ‘explicit objective’ orientation in this study) in Hallidayan sense (cf. Liardét, 2018) unexamined. Zhou (2023a) restricted his topic within interpersonal metaphor of modality (cf. Section 2), particularly the ‘explicit objective’ orientation, and he later extended his topic to the relationship between congruent expressions of modality and metaphorical expressions of modality across registers (2023b) or to the attracted expressions of modality to the construction of interpersonal metaphor of modality (2023c). Based on Zhou’s studies, the synchronic distribution of all subtypes of expressions of modality in terms of the formality of registers is undertheorized.

Against the backdrops, this study aims to investigate the synchronic distribution trends of a fuller coverage of expressions of modality across registers, specifically the relationship between expressions of modality and the formality of registers, and the motivation underlying this relationship. In fulfilling our aims, the following research questions are proposed.

  1. 1.

    How are expressions of modality synchronically distributed across registers?

  2. 2.

    What is the relationship between expressions of modality and formality of registers, and what explanations can be offered for the relationship?

This paper is structured as follows. Section 2 delineates subtypes of expressions of modality. Section 3 introduces the information of COCA, confirmation of formality of registers, methods, and the way that data are collected. Sections 4 and 5 consider the distribution of subtypes of expressions of modality and the general distribution trend of expressions of modality, respectively. Results are discussed in Section 6, and a conclusion is made in Section 7.

Theoretical foundations

Expressions of modality reflect the speaker’s judgment of his/her saying (Fontaine, 2013). They incorporate expressions of modality at word rank and expressions of modality at clause rank (cf. Halliday and Matthiessen, 2014). The former is further subcategorized into modal verbal operators, modal adjuncts, and modal nominalizations; the latter is further classified into ‘explicit subjective’ orientation and ‘explicit objective’ orientation (Zhou, 2023c). In this section, key terms in relation to expressions of modality and interpersonal metaphor of modality in SFL, a concept employed to expound the relationship between expressions of modality and formality of registers, are delineated. Subsection 2.1 profiles expressions of modality at word rank, and subsection 2.2 expounds the concept of interpersonal metaphor of modality, which corresponds to expressions of modality at clause rank, i.e., ‘explicit subjective’ orientation and ‘explicit objective’ orientation within SFL (Halliday, 1994).

Expressions of modality at word rank

In English, modal verbal operators (Modal auxiliaries in Nuyts’ (1993) term, or modal auxiliary verbs in Fontaine’s (2013) term, or core modal auxiliaries in Leech’s (2011) term), generally include can, could, may, might, shall, should, will, would, and must. They share a number of typical grammatical properties (e.g., negation and inversion) that make them distinguishable from other modal operators such as peripheral/marginal modal operators or semi-modal operators (e.g., need, have to, seem to).Footnote 5 These grammatical properties include the main verb in the bare infinitive form and the lack of non-tensed form or person-number agreement. These modal operators, except for must, are usually paired as present and past tense forms of single lexemes (i.e., can/could, may/might, shall/should, will/would). In this paper it is useful to regard these pairings as individual lexemes, although their relationships are complex, because we only consider their distribution across registers, precluding the semantic differences that lie in each pair. Consider the examples in (2).

  1. (2)

    ▓

    1. a.

      The storm system could affect tens of millions of people in Virginia. (COCA_spoken_2016)

    2. b.

      The Golden Eagle Inn might have a room for you. (COCA_magazine_1997)

    3. c.

      The warm January should bring a big drop in oil and gas prices, right? (COCA_newspaper_1990)

    4. d.

      The train would round the bend, and the rails would start to quake a little underfoot. (COCA_fiction_2019)

    Could, might, should, and would in (2a-d) exemplify modal verbal operators in each pair. All of them demonstrate the grammatical properties of being followed by a bare infinitive verb form and the lack of person-number agreement between the subject and the modal operator.

    Modal adjuncts are adverbials that demonstrate speakers’ subjective comment and attitude towards a proposition. Halliday and Matthiessen (2014: 177–178) argued that modality is subcategorized into modalization (referring to the scales of probability and usuality) and modulation (referring to the scales of obligation and inclination). They (2004: 618) expounded that “[i]n philosophical semantics probability is referred to as ‘epistemic’ modality and obligation as ‘deontic’ modality”. Considering modal adjuncts at issue, those expressing modalization include certainly, probably, possibly, and surely, and those expressing modulation include essentially, necessarily, and obligatorily.

    Modal nominalization, which is named analogously from Halliday and Matthiessen’s (2014: 710) terms such as verbal nominalization and adjectival nominalization, is generally derived from modal adjectives (e.g., from necessary to necessity) or modal verbs (e.g., from can to ability or from may to possibility). Modal nominalizations that express modalization generally include certainty, probability, possibility, and likelihood, and those that express modulation include essentiality, necessity, and obligation.

    Modal adjuncts and modal nominalizations in relation to expressions of modality at word rank are exemplified in (3) and (4), respectively.

  2. (3)

    ▓

    1. a.

      She doesn’t mean anything by it. She’s half-drunk and probably has no idea what she’s saying. (COCA_academic_2009)

    2. b.

      So, consumers, when they walk down the grocery aisle, are necessarily even going to know what they’re buying. (COCA_spoken_1992)

  3. (4)

    ▓

    1. a.

      The figure shows the probability of stroke-free survival in the two treatment groups. (COCA_academic_1991)

    2. b.

      As a matter of practical necessity the state may have to approve appropriations despite foreseeable harm to public trust uses. (COCA_academic_1996)

Value is an important parameter in SFL while denoting the meaning of modality. An expression of modality could be of a high, median, or low value. “Value” used in this sense is synonymous to the more common term in the mainstream literature, i.e., “strength”. It should be noted that some minor differences also exist between the two terms. For instance, modal verbal operators shall and will are regarded as expressions with median modal value in SFL, while they are with high modal value outside SFL. This research follows the SFL framework (underlying the fact that recent studies on the distribution of expressions of modality are generally SFL-based), so they are categorized into modal verbal operators denoting modality with median value. The aforementioned items that are included in the three subtypes of expressions of modality are tabulated in Table 1. It should be also noted that Table 1 only highlights the modal value of subtypes of expressions of modality at word rank, so the semantic difference between certainly and obligatorily is not addressed. The former expression of modality denotes the epistemic meaning and the latter the deontic meaning, which corresponds to Halliday and Matthiessen’s (2004) terms of Probability and Obligation, respectively.

Table 1 Instances of expressions of modality with various modal values.
Full size table

Interpersonal metaphor of modality

Interpersonal metaphor of modality involves the remapping of modal meanings on the semantic stratum onto expressions of modality at different ranks on the lexicogrammatical stratum (Zhou, 2023c). Metaphors of modality, alongside interpersonal metaphors of mood, represent one of two types of interpersonal metaphor related to the ways in which propositions are negotiable or not (Yang, 2019). Modality, a subsystem of interpersonal meaning, includes four main types: probability, e.g., it is possible, likely, usuality, e.g., it is usual, customarily, obligation, e.g., it is obligatory, necessarily, and inclination, e.g., will (Halliday and Matthiessen, 2014). Interpersonal meaning, which enacts the interpersonal relationships between interlocutors, is one of the three meanings denoted by a clause. The other two meanings are ideational meaning that construes experiences of the world through language and textual meaning that packages ideational and interpersonal meanings to form texts in contexts. Each of these four types of modality can be realized by means of different orientations that represent different degrees of judgment on the validity of a proposition, either subjective or objective, and either explicit or implicit. Subjective orientation frames the meaning of modality as the speaker’s subject judgment on the validity of a proposition, e.g., I think, must, etc.; objective orientation frames modality as an objective evaluation, e.g., it’s likely, possibly, etc.; explicit orientation directly states the source of judgment in a separate clause, e.g., I think, it’s likely, etc.; while implicit orientation leaves the source of judgment implicit, e.g., must, possibly, etc. (Halliday and Matthiessen, 2014, p. 181). SFL scholars argue that only ‘explicit subjective’ and ‘explicit objective’ orientations realize interpersonal metaphors of modality (cf. Taverniers, 2003; Halliday and Matthiessen, 2014; He, 2020).

‘Explicit subjective’ orientation is constructed by the formation of a pronoun in its first personal nominative case (i.e., I and we) and a mental process verb such as think, believe, guess, etc. Specific instances of this type that are employed to explore their register-based distribution are taken from Zhou and Gao’s (2021) research, who used them to investigate their diachronic distributions during the past two centuries in American English, including I/we think (assume, believe, guess, suppose). Consider the following examples in (5). I think in (5a) and we believe in (5b) are instances of ‘explicit subjective’ orientation in interpersonal metaphor of modality in that speakers’ judgment is positioned outside the projected clause or the proposition i.e., that culture is changing, and it’s changing for the better in (5a) and women are going to make the winning difference in this campaign and in this election in (5b). They are subjective because of the first-person pronoun I in (5a) and we in (5b) realizing the subject of the projecting clauses respectively.

  1. (5)

    ▓

    1. a.

      I think that culture is changing, and it’s changing for the better. (COCA_spoken_2019)

    2. b.

      We believe that women are going to make the winning difference in this campaign and in this election. (COCA_spoken_2000)

    ‘Explicit objective’ orientation is structured by the sequence of it, any form of the copula be, a modal adjective, and a to/that clause. This grammatical pattern is also termed by other scholars as “it-extraposition” (Kaltenböck, 2005). Consider the examples in (6). It was likely in (6a) and it is obligatory in (6b) are instances of ‘explicit objective’ orientation in terms of interpersonal metaphor of modality in that the modal meanings of probability, i.e., likely, and obligation, i.e., obligatory, are realized outside the propositions. Variations of this grammatical pattern could also be attested. That is, the projecting clause could either have the copula be been modified by a modal verbal operator or have the modal adjective been qualified by an adverb. Variations of this kind are instantiated by (7a-b). In (7a), the modal verbal operator might is used to modify the copula be in the projecting clause it might be possible, and in (7b), the adverb entirely is interpolated in the projecting clause it’s entirely possible to qualify the modal adjective possible.

  2. (6)

    ▓

    1. a.

      This told Bosch that it was likely that the woman had been a photographer. (COCA_fiction_2012)

    2. b.

      Because of all of what I have mentioned it seems that it is obligatory to pay respect to the tablets of the loudspeaker, that is, to the cylinders. (COCA_academic_2018)

  3. (7)

    ▓

    1. a.

      It might be possible that a respectable case could be made in favor of them. (COCA_magazine_1990)

    2. b.

      It’s entirely possible that this year’s draft could unfold like last year’s. (COCA_newspaper_2019)

Interpersonal metaphors of modality could be also of various values. ‘explicit subjective’ orientation and ‘explicit objective’ orientation, with respect to high, median, or low value, are presented in Table 2.Footnote 6

Table 2 Instances of interpersonal metaphor of modality with various values.
Full size table

Methodology

Corpus

The on-line corpus COCA (cf. Davies, 2010) is employed in this research.Footnote 7 COCA is the only large, genre-balanced corpus of American English,Footnote 8 ranging from 1990 to 2019 and containing more than one billion running words of text (words considered in the research are exactly 618,200,644, because we do not consider such registers as the general web, blog, and TV/movies).Footnote 9 Registers considered in this study include spoken texts, fiction, magazines, newspapers, and academic texts. The general information on registers at issue in COCA is tabulated in Table 3.

Table 3 Information of registers in COCA.
Full size table

Formality of registers

With respect to the formality of registers under consideration, they could be either very informal (e.g., spoken texts), or very formal (e.g., academic texts), or somewhere in between (e.g., magazines, newspapers) (cf. overview information on registers in COCA by Mark Davies (2010)). In this research, we employed the formula proposed by Heylighen and Dewaele (1999) and also testified by Zhou (2023b, 2023c) to establish a continuum of formality across registersFootnote 10 based on two considerations. One is that the formula counts nouns positively but verbs negatively in establishing the formality of a register, which is in accordance with the argument that nouns are preferred in a formal style while verbs are preferred in an informal style (cf. Heylighen and Dewaele, 1999; Biber and Gray, 2016). Additionally, adjectives, prepositions, and articles are used to specify additional details that nouns contain, while adverbs are used to supplement information to verbs. Pronouns and interjections are closely associated with the context in that their meanings are recovered from the linguistic and situational contexts; context-dependent word classes are generally used in an informal style (cf. Heylighen and Dewaele, 1999). The other is that the formula is practicably operationalizable in that frequencies of each word class could be directly obtained by constructing relevant search queries (e.g., [v*] and [nn*] could be implemented to retrieve all occurrences of verbs and nouns, respectively, in COCA). By implementing the formula, the working continuum of formality of registers is confirmed. Precisely, from the most informal register to the most formal one, these registers in question follow the sequence of spoken texts (normalized frequency or NF = 47.21), fiction (NF = 49.26), magazines (NF = 58.94), newspapers (NF = 59.07), and academic texts (NF = 64.97). The formality of registers confirmed by Heylighen and Dewaele’s (1999) formula is also in accordance with that proposed by Biber (1988) and Biber and Conrad (2009) who captured register differences based on a multi-dimensional analysis.

Data collection

For obtaining occurrences of subtypes of expressions of modality in COCA, relevant search queries or SQs were constructed. We first retrieved occurrences of subtypes of expressions of modality at word rank according to SQs 1, 2, and 3. SQ1 reads as a sequence of modal verbal operators at issue and a verb in its bare infinitival form; this query will successfully and exhaustively retrieve such clauses as exemplified in (2). SQ2 was used to retrieve modal adjuncts at issue, and hence, such clauses as exemplified in (3) will be obtained. In a similar vein, modal nominalizations that occur in such clauses as shown in (4) are obtained by SQ3.

SQ1: [vm*] [v?i*]

SQ2: certainly|probably|possibly|surely|essentially|necessarily|obligatorily

SQ3: certainty|probability|possibility|essentiality|necessity|obligation|likelihood

For retrieving occurrences of “explicit subjective” orientation and ‘explicit objective’ orientation in interpersonal metaphor of modality, SQs 4 and 5 were constructed respectively. Occurrences of the former type were retrieved by SQ4, which reads as a grammatical pattern in the sequence of a punctuation mark, a personal pronoun, a mental process verb, a pronoun or noun, and a verb. This query will successfully retrieve such clauses as exemplified in (5). Those of the latter type as shown in (6) will be retrieved by SQ5, which means a grammatical pattern in the sequence of it, any form of copula be, any modal adjective, and a to or that clause. Owing to the fact that SQ5 could not efficiently retrieve such clauses as shown in (7), SQ6, which means the copula be is modified by a modal auxiliary, is supplemented to SQ5 to retrieve such clauses as shown in (7a), and SQ7, which reads as the modal adjective is qualified by an adverb, is supplemented to SQ5 for the purpose of retrieving such clauses as shown in (7b).

SQ4: [y*] [p*] think|assume|believe|guess|suppose [pp*]|[nn*] [v*]

SQ5: it [vb*] [j*] to|that

SQ6: it [vm*] [vb*] [j*] to|that

SQ7: it [vb*] * [j*] to|that

Raw frequencies by applying the seven search queries were first normalized to per million words (PMW) for the ease of comparison, and then converted according to the principle of equal totality if necessary (cf. He, 2019; Zhou and Gao, 2022).

Statistical analysis

For the purpose of quantitatively examining the distribution of expressions of modality across registers, this research employed some statistical measures in R language, a project for statistical computing (see R project’s webpage: https://www.r-project.org/), to examine the variation or correlation between variables. The variation analysis is employed to answer the first research question and the correlation analysis is used to address the second research question. Specifically, the Shapiro-Wilk normality test (the function in R language is Shapiro.test(x)) was used for testifying the normal distribution of variables. The Student T test (t.test(x,y)) was used to testify the difference when variables are normally distributed and the Mann-Whitney test (Wilcox.test(x,y)) for variables that are not normally distributed (cf. Stefanowitsch, 2020; Zhou and Gao, 2022). The Pearson’s correlation coefficient test (cor.test(x,y)) was used to testify the correlation when variables are normally distributed, and the Spearman’s correlation coefficient test (cor.test(x,y,method = “spearman”) was adopted if one variable is not normally distributed. In addition, figures in this research were visually presented by using the Excel software.

Results

The distribution of subtypes of expressions of modality is shown in Fig. 1, and results of variation analysis between subtypes, inclusive of instances of each subtype, are tabulated in Table 4.Footnote 11

Fig. 1: Normalized frequencies of subtypes of expressions of modality across registers.
figure 1

a Modal verbal operators, (b) modal adjuncts, (c) modal nominalizations (possibility on the right-hand axis), (d) explicit subjective orientation (think on the right-hand axis) (e) explicit objective orientation (possible on the right-hand axis), (f) expressions of modality (modal verbal operators on the right-hand axis).

Full size image
Table 4 Results of variation analysis of expressions of modality.Footnote

In tables of this paper, * stands for significance at the 0.05 level and ** stands for significance at the 0.01 level

.
Full size table

According to Fig. 1a, such modal verbal operators with high values as must and shall/should are not, comparatively, preferentially used by language users while constructing texts at issue. In addition, they are plateauing irrespective of the formality of registers. That is to say, no obvious preferences for the five types of registers could be identified although registers as such are of various degrees of formality. Those with low modal value can/could preferentially occur in more informal registers (e.g., normalized frequency or NF = 1982 in spoken texts and 1502 in academic texts). Those with low value may/might and median value will/would seem to be competing with each other in more formal registers, i.e., the latter occurs less frequently while the former occurs more frequently. Figure 1b shows that such modal adjuncts with high and median values as certainly and probably occur more frequently in more informal registers (e.g., in spoken texts, 308 for certainly and 388 for probably) but less frequently in more formal registers (e.g., in academic texts, 93 for certainly and 123 for probably). Modal adjuncts with median value, such as essentially and necessarily are particularly disfavored by fictional texts while preferred by academic ones. Figure 1c shows that such modal nominalizations as possibility and obligation are particularly infrequently used in fictional texts (NF = 33 PMW and 7 respectively), but more frequently in formal registers, peaking in academic texts (95 and 25 respectively). Similarly, such modal nominalizations with high value as certainty (7) and with median value as probability (4) and necessity (5) are most infrequently used in spoken texts, and fluctuate rather minorly in registers like fiction, magazines, and newspapers. However, the fluctuation of these modal nominalizations is greatly broadened in academic texts (14 for certainty; 46 for probability; 34 for necessity).

Figure 1d shows that such ‘explicit subjective’ orientation as I/we think is the most frequently used expression in registers of COCA, particularly in the most informal spoken register (410.34). The second frequent instance of this subtype is I/we guess, with I/we believe, I/we suppose, and I/we assume following. It also shows that all instances of ‘explicit subjective’ orientation are rather preferred in such informal registers as spoken and fictional texts, and generally dispreferred in such formal registers as magazines, newspapers, and academic texts. Figure 1e demonstrates that expressions of ‘explicit objective’ orientation are less frequently used in fictional (0.19 PMW for probable, 9.78 for possible, 0.43 for essential, 2.62 for necessary, and 1.9 for likely) and spoken (0.19 for probable, 13.34 for possible, 1.24 for essential, 3.16 for necessary, and 6 for likely) texts than the other types of registers. That is, they prefer formal texts to informal ones.

Figure 1f demonstrates that modal verbal operators are diminishing in usage drastically from spoken texts (11359 PMW) to academic texts (9706). This diminishing is by and large contributed by the less frequent use of such modal operators with low or median value as can/could (1982 in spoken texts and 1502 in academic texts) and will/would (2896 in spoken texts and 1586 in academic texts) (cf. Figure 1a) in more formal registers. Modal adjuncts occur frequently in the most informal register like spoken texts (917), but their occurrences do not make any difference in more formal registers such as newspapers (411) and academic texts (417). The register-sensitive nature (He and Yang, 2018) is mainly characterized by the frequent use of modal adjuncts with high or median value like certainly and probably in more informal registers and their infrequent uses in more formal registers (cf. Figure 1b). Modal nominalizations are nearly equally preferred in both informal and formal registers (e.g., 585 in spoken texts and 490 in newspapers) except for the most former register – academic texts (1370). This particular preference in academic texts is mainly caused by such modal nominalizations with median or low value as probability and possibility (cf. Figure 1c). Pertaining to the two subtypes of interpersonal metaphor of modality, instances of ‘explicit subjective’ orientation are frequently used in more informal registers but infrequently used in more formal registers, which is by and large contributed by the use of I/we think with low value from spoken texts (410.34) to academic texts (13.37) (cf. Figure 1d). Expressions of ‘explicit objective’ orientation generally keep a minor fluctuation from spoken texts (25) to newspapers (23), but this minor fluctuation is broken up and these expressions are extremely preferred in academic texts (98). This fluctuation is primarily caused by the use of the expression it is/was possible with low value from spoken texts (13.34) to academic texts (48.83) (cf. Figure 1e).

Table 4a shows that occurrences of modal verbal operators are either significantly or extremely significantly different from each other except for the one between can/could and will/would (t = −2.0465, p = 0.0901 > 0.05), indicating the dominance of can/could and will/would over the others, with may/might, shall/should, and must following. Table 4b shows that certainly (W = 25, p = 0.0079 < 0.01 with possibly, essentially, and necessarily respectively) and probably (t = 4.1739, p = 0.0135 < 0.01 with possibly; t = 4.3633, p = 0.0096 < 0.01 with essentially; t = 4.2006, p = 0.0106 < 0.01 with necessarily) are significantly different from the other modal adjuncts, and this difference is generally contributed by their preferential occurrences in spoken register (cf. Figure 1b). Table 4c shows that occurrences of modal nominalizations certainty, probability, necessity, and obligation demonstrate no significant difference between each other, indicating no perspicuous preference in these registers of COCA. Concerning the low value modal nominalization possibility, it is statistically significantly different from the other four modal nominalizations (t = −4.2787, p = 0.0119 < 0.05 with certainty; W = 2.5, p = 0.0459 < 0.05 with probability; W = 24, p = 0.0159 < 0.05 with necessity; t = 3.6921, p = 0.0162 with obligation), which suggests that language users frequently employ the low modal item in its nominalized form, i.e., possibility, to objectify the modal meaning within the proposition.

In relation to the variation between ‘explicit subjective’ orientation in interpersonal metaphor of modality, Table 4d shows that I/we think is significantly different from I/we assume (W = 25, p = 0.0079 < 0.01), I/we believe (W = 24, p = 0.0159 < 0.05), I/we guess (W = 23, p = 0.0318 < 0.05), and I/we suppose (W = 24, p = 0.0159 < 0.05); in addition, I/we believe and I/we assume (t = −2.8489, p = 0.0437 < 0.05) are also significantly different from each other. In relation to this variation between ‘explicit objective’ orientation, it is shown in Table 4e that probable and possible, possible and essential, probable and likely, and probable and necessary are significantly different from each other, respectively. These differences indicate that when “possibility” in modalization needs to be expressed, ‘explicit objective’ orientation that denotes low value such as possible and likely is extremely likely to be employed; when “necessity” in modulation needs to be expressed, the one with median value like necessary is likely to be used.

Regarding the overall occurrences of subtypes of expressions of modality, Table 4f shows that modal verbal operators are extremely significantly different from modal adjuncts (W = 25, p = 0.0079 < 0.05), modal nominalizations (t = 27.368, p = 8.539e-08 < 0.01), ‘explicit subjective’ orientation (t = 32.055, p = 1.643e-06 < 0.01), and ‘explicit objective’ orientation (t = 33.452, p = 4.553e-06 < 0.01) respectively, indicating that modal verbal operators are still the most frequently used expressions of modality in registers of COCA, particularly in more informal registers. In addition, modal adjuncts are significantly different from ‘explicit subjective’ orientation (W = 2, p = 0.0318 < 0.05) and ‘explicit objective’ orientation (t = 5.3053, p = 0.0053 < 0.01), and the significant difference also exists between modal nominalizations and ‘explicit subjective’ orientation (W = 1, p = 0.0159 < 0.05) or ‘explicit objective’ orientation (t = 3.6696, p = 0.0209 < 0.05)

Results of correlation analysis between subtypes of expressions of modality, inclusive of instances of each subtype, are tabulated in Table 5.

Table 5 Results of correlation analysis of expressions of modality.
Full size table

In relation to the relationship between subtypes of expressions of modality at word rank and the formality of registers, Table 5a shows that only the distribution of will/would is significantly negatively correlated with that of must (r = −0.9211, p = 0.0262 < 0.05), which suggests that will/would occur less frequently in more formal registers while must in more informal registers. More importantly, distribution of will/would is negatively but not significantly correlated with that of may/might (r = −0.7182, p = 0.1717), which indicates that the compensation of may/might for will/would does not reach the significant level in more formal registers (cf. Figure 1a). In Table 5b, such modal adjuncts as probably and possibly are positively significantly correlated (r = 0.9065, p = 0.0338 < 0.05), indicating that the two modal adjuncts occur more frequently in more informal registers. The positive significant correlation between essentially and necessarily (r = 0.9268, p = 0.0235 < 0.05) suggests that both modal adjuncts are preferred in more formal registers (cf. Figure 1b). In Table 5c, these modal nominalizations do not reach a significant level (precluding the correlation between possibility and obligation) although they are positively correlated. It is also shown that possibility and obligation are positively extremely significantly correlated with each other (r = 0.9937, p = 0.0006 < 0.01), indicating that they preferentially occur in more formal registers (cf. Figure 1c).

With respect to the relationship between interpersonal metaphor of modality and the formality of registers, Table 5d shows that I/we think and I/we believe (ρ = 1, p = 0.0167 < 0.05), I/we assume and I/we guess (t = 0.9855, p = 0.0021 < 0.01), and I/we assume and I/we suppose (ρ = 1, p = 0.0167 < 0.05) are positively significantly correlated with each other respectively, indicating their similar distribution trend. That is, the more formal a certain register is, the less frequently ‘explicit subjective’ orientation will occur (cf. Figure 1d). In Table 5e, it is shown that probable and essential (ρ = 9747, p = 0.0048 < 0.01), possible and necessary (ρ = 1, p = 0.0167 < 0.05), possible and likely (ρ = 1, p = 0.0167 < 0.05), essential and likely (t = 0.9860, p = 0.0020 < 0.01), and necessary and likely (ρ = 1, p = 0.0167 < 0.05) are significantly correlated with each other respectively, indicating that they share a similar distribution trend that expressions of ‘explicit objective’ orientation as such are preferentially used in more formal registers (cf. Figure 1e).

It can be seen in Table 5f that three types of correlation between expressions of modality could be considered. First, modal verbal operators and modal adjuncts are positively significantly correlated with one another (r = 0.8761, p = 0.05), and this positive significant correlation also exists between modal adjuncts and ‘explicit subjective’ orientation (r = 0.9677, p = 0.0069 < 0.01); additionally, modal verbal operators and ‘explicit subjective’ orientation, to some extent, are correlated with each other (r = 0.8608, p = 0.061) in that this significance is rather close to the threshold of ‘a = 0.05’. The positive significance of correlation between them suggests that the three types of expressions of modality occur frequently in more informal registers but infrequently in more formal registers (cf. Figure 1f). Second, modal nominalizations are positively significantly correlated with ‘explicit objective’ orientation (r = 0.9816, p = 0.003 < 0.01), indicating that expressions of modality as such are preferentially used in more formal registers, particularly in academic texts. Third, modal nominalizations are to some extent negatively correlated with modal verbal operators (ρ = −0.3, p = 0.6833) and modal adjuncts (r = −0.2461, p = 0.6898), although they do not reach the significant level, which indicates that the loss of modal verbal operators and modal adjuncts is compensated for by the gaining of modal nominalizations and this compensation happens obviously in academic texts (1370 PMW). In addition, there is a negative insignificant correlation between ‘explicit objective’ orientation and ‘explicit subjective’ orientation in terms of interpersonal metaphor of modality (r = −0.5124, p = 0.3744), suggesting that the gaining of ‘explicit objective’ orientation is also compensating for the loss of ‘explicit subjective’ orientation.

Discussion

This section discusses the distribution of expressions of modality across registers in relation to the first research question and the relationship between distribution of expressions of modality and formality of registers in relation to the second research question.

Distribution of expressions of modality across registers

Concerning expressions of modality in each subtype, those with low or median values are preferentially used across registers of COCA by language users to express their subjective judgment of propositions at issue. More specifically, this inclination is instantiated by can/could, will/would, and may/might in modal verbal operators, probably and possibly in modal adjuncts, possibility in modal nominalizations, I/we think in ‘explicit subjective’ orientation, and possible in ‘explicit objective’ orientation (cf. Figure 1a-e and Table 4). Findings of this research are in accordance with earlier studies on the distribution of modal verbal operators across registers (Collins, 2009; Verhulst, et al. 2013; Zhou, 2023b). They also echo the findings documented by Zhou (2023a, 2023b), who examined the distribution of interpersonal metaphor of modality across registers in corpora of COHA, COCA, and BNC. What distinguishes this study from earlier ones is that the current one expanded the investigation of expressions of modality into other domains, such as modal nominalizations and individual expression of modality in each subtype. Reasons for this preferential employment of expressions of modality with low or median value, particularly the former, are basically two-fold. First, language users actively engage with a multiplicity of voices, embracing the fluid and evolving nature of propositions. This approach not only acknowledges the legitimacy of diverse perspectives but also fosters a collaborative spirit of inquiry and negotiation, leading to a richer and more nuanced comprehension of the issues at hand (for the entertainment of varying voices and the arguability and negotiability of propositions, cf. Martin and White, 2005; Yang, 2019). Consider the example in (8a), in which the modal adjunct probably denotes a median modal value (Halliday and Matthiessen, 2014). The modification of the proposition she has no idea what’s she’s saying by probably with median value of modality is licensed by the preceding proposition she’s half-drunk, because a drunk person’s words are usually uttered unintentionally. In doing so, the use of probably leaves sufficient room for hearers to argue the validity of the proposition she has no idea what’s she’s saying. Second, the judicious use of modal expressions serves to mitigate the impact of one’s words, ensuring that communication remains respectful and considerate. By opting for language that carries a lower or median degree of assertiveness, speakers navigate the delicate balance between conveying their messages and preserving the dignity and self-esteem of their audience, thereby fostering an environment conducive to open and harmonious exchange (for the warranty of the hearers’ or readers’ faces by expressions of modality, cf. Zhou, 2023a, 2023c). Consider the example in (8b), wherein I think is employed to attenuate the speaker’s judgment of the proposition the Republicans need to start talking about the real issues before the country so as to accept other varying voices (e.g., the Republicans do not necessarily need to start talking about the real issues before the country) and consequently others’ faces are protected.

  1. (8)

    ▓

    1. a.

      She doesn’t mean anything by it. She’s half-drunk and probably has no idea what she’s saying. (COCA_academic_2009)

    2. b.

      I think the Republicans need to start talking about the real issues before the country. (COCA_spoken_2004)

Expressions of modality at word rank, i.e., modal verbal operators, modal adjuncts, and modal nominalizations, are used more frequently than expressions of interpersonal metaphor of modality, i.e., ‘explicit subjective’ orientation and ‘explicit objective’ orientation, or that modality at clause rank (cf. Figure 1f). This finding is consistent with what was documented in He’s (2020) or Zhou’s (2023b) research. Both studies uncovered that congruent expressions of modality (modal verbal operators and modal adjuncts in this research) occur more frequently than their metaphorical counterparts across registers. Possible reasons are also two-fold. For one thing, the employed corpus, i.e., COCA, consists of various registers that are constructed for general purposes (e.g., fiction, magazines); that is, their potential readers are the general public. The only register that requires the readers to be trained to understand is academic texts. Academic register generally features grammatical metaphor which incorporates ‘explicit objective’ orientation (Halliday and Matthiessen, 1999). Metaphorical expressions of modality are more difficult to understand than their congruent counterparts (Halliday, 1994; Halliday and Matthiessen, 1999, 2014). Against this backdrop, registers as such (except for academics) usually employ expressions of modality that are easy for readers to understand, such as modal verbal operators and modal adjuncts. For another, implicitness is typically characteristic of these considered registers of COCA. In SFL, modal verbal operators and modal adjuncts denote their modal meanings implicitly (Zhou, 2023b). Regarding expressions of ‘explicit subjective’ orientation, their modal meanings are directly projected by the grammatical pattern I/we think. Owing to its spoken feature, this grammatical pattern is not particularly favored in other written registers of COCA.

Relationship between distribution of expressions of modality and formality of registers

Regarding certain expressions of modality, those with low or median values are used in more formal registers while those with high value are used in more informal registers. This is in accordance with the finding documented by Zhou (2023a, 2023b). As expounded in the foregoing subsection, those expressions of modality are employed by language users to entertain different voices so as to make the propositions negotiable and arguable (Martin and White, 2005; Yang, 2019; Zhou, 2023a). However, the preferential employment of expressions of modality with high value in more informal registers such as spoken texts is motivated by “the strengthening of propositions when sufficient evidence is provided” (Zhou, 2023c: 175). Consider the example in (9). The expression of modality with high value we believe is used in the sense that the validity of the proposition the majority of doctors will actually get paid either the same or a little bit more is backed up by the preceding proposition because we would be covering everybody and we wouldn’t allow duplicative care, which is how we operate with Medicare right now which functions as the needed evidence.

  1. (9)

    Because we would be covering everybody and we wouldn’t allow duplicative care, which is how we operate with Medicare right now, we believe that the majority of doctors will actually get paid either the same or a little bit more. (COCA_spoken_2019)

    With respect to the relationship between the general distribution of expressions of modality and formality of registers, two respects are to be discussed. First, modal nominalizations preferentially occur in more formal registers such academic texts, while modal verbal operators and modal adjuncts preferentially occur in more informal registers such as spoken texts if expressions of modality at word rank are concerned. This could be explained by the fact that modal meanings in spoken texts are articulated in a subjective way, while those in academic texts should be realized objectively so as to highlight the credibility and persuasiveness of academic texts (Tian, 2017) and ultimately make the academic texts authoritative (Schleppegrell, 2004). Specifically, modal verbal operators and modal adjuncts realize modal meanings directly in SFL (cf. Halliday and Matthiessen, 2014), whereas those modal meanings are presupposed by the employment of modal nominalizations. In other words, such expressions of modality as modal verbal operators and modal adjuncts are important components of Finite element in Mood system (which consists of Mood and Residue, the former is composed of the subject and the Finite element in a clause and the latter refers to the rest of that clause), realizing interpersonal meaning (i.e., enacting the social relationship between interlocutors). The Finite element is expressed by modal verbal operators which are capable of making propositions arguable or negotiable between interlocutors and, hence, conveying the interpersonal meaning subjectively (Halliday and Matthiessen, 2014). Consider the example in (10a), wherein the subject realized by this and the Finite element realized by the modal verbal operator might form the Mood of the clause. That is to say, the interpersonal meaning is directly realized by the modal verbal operator, and thus, the speaker’s subjective meaning is expressly articulated. On the contrary, a modal nominalization realizing the modal meaning is an element of the proposition. In this sense, the modal meaning is presupposed and the interpersonal meaning is therefore “disguised” as the ideational meaning, which construes the experience in social reality. By doing so, the way to express the modal meaning is made indirectly and thus conveyed objectively. Consider the example in (10b), wherein the writer’s subjective meaning is not expressed by a Finite element, but by the modal nominalization possibility, which has been turned into an element of the proposition. Thus, the objectivity of expressing the modal meaning is foregrounded.

  2. (10)

    ▓

    1. a.

      This might be my favorite story of the day. (COCA_spoken_2019)

    2. b.

      Nobody in Los Alamos raised this possibility. (COCA_academic_1992)

  3. (11)

    Although it is extremely difficult to establish legitimacy for an object, it is often possible to point out what’s wrong with it. (COCA_academic_2003)

Second, expressions of ‘explicit objective’ orientation are preferred in more formal registers such as academic texts while those of ‘explicit subjective’ orientation are frequently used in more informal registers such as fictional texts if interpersonal metaphor of modality is concerned. This is corroborated by the argument that grammatical metaphor (i.e., modal nominalizations and ‘explicit objective’ orientation are categorized into ideational metaphor and interpersonal metaphor of modality, respectively, in Hallidayan sense) is particularly preferred in academic texts (cf. Liardét, 2018; He, 2020; Zhou, 2023a, 2023b). Although the modal meaning, by employing ‘explicit subjective’ orientation in terms of interpersonal metaphor of modality, is expressed outside the proposition in a projecting clause, the explicit use of the first personal pronouns (i.e., I or we) foregrounds the commentator who has made the judgment of the proposition. Therefore, this type of expressions of modality denotes a strong sense of subjectivity. However, by means of ‘explicit objective’ orientation, the modal meaning originally represented by a Finite element is now expressed by another proposition, in which the modal adjective functions as the predicate of the proposition. By doing so, the means of representing the modal meaning is concealed so that the objectivity of the proposition is further highlighted. This is considerably in accordance with the fact that academic texts are constructed in an objective way so as to convey the text information persuasively (Tian, 2017) and authoritatively (Schleppegrell, 2004). Consider the example in (11). The modal meaning is not expressed by a Finite element (e.g., possibly or might), but by the ‘explicit objective’ orientation it is often possible in which possible functions as the predicate of the proposition; hence the means of denoting the modal meaning is concealed and the objectivity of the proposition is highlighted.

Conclusion

This research investigated the register-based distribution of expressions of modality in English. The results show that expressions of modality with low value or occasionally with median value are significantly employed by language users to construct texts in registers of COCA so as to entertain varying voices. Specifically, this significant employment is instantiated by can/could, will/would, and may/might in modal verbal operators, probably and possibly in modal adjuncts, possibility in modal nominalizations, I/we think in ‘explicit subjective’ orientation, and possible in ‘explicit objective’ orientation. These results also show that the gaining of modal nominalizations compensates for the loss of modal verbal operators and modal adjuncts in terms of expressions of modality at word rank, and the gaining of ‘explicit objective’ orientation compensates for the loss of ‘explicit subjective’ orientation in terms of interpersonal metaphor of modality. The purpose is to conceal the speakers’ or writers’ subjective meaning, and the objectivity of the proposition at issue is therefore foregrounded.

This study contributes to existing literature on modality in at least two ways. For one thing, a fuller scope of expressions of modality in English is documented. Instead of solely examining modal verbal operators, modal adjuncts, or modal projecting clauses, this research considered, to a large extent, the distribution of modality across registers in relation to expressions of modality both at word ranks such as modal verbal operators and modal nominalizations and at clause rank or interpersonal metaphor of modality such as ‘explicit subjective’ orientation and ‘explicit objective’ orientation. For another, the distribution of modality across registers is theoretically associated with the interpersonal metaphor of modality. This research considered not only the preferential occurrence of expressions of modality across registers, but also the reasons underlying this preference from the perspective of interpersonal metaphor of modality. Specifically, language users employed different kinds of expressions of modality for the purposes of either entertaining other varying voices or concealing the commentators’ subjective meaning to highlight the objectivity of propositions in former registers. This study also contributes to a fuller consideration of the influence of the register variable on expressions of modality in relation to register studies. Inclusive of such registers as spoken texts, fiction, magazines, and newspapers that are investigated in former studies, the current research also considered the academic register regarding the distribution of expressions of modality.

One limitation of the paper is that only the synchronic distribution of expressions of modality in registers of COCA is considered, leaving their diachronic distributions untouched. Another limitation is that COCA is typically characteristic of American English, and thus, it should be to some extent cautious to regard the findings of this research as the common properties of English in general. In addition, as one of the reviewers commented, this research only considered the positive forms of expressions of modality (e.g., may or it is necessary), excluding the negative formulations of modality. This treatment will arguably make the findings of the study partial. Underlying the three limitations, future studies are therefore suggested to consider the diachronic distribution of expressions of modality (inclusive of both positive and negative formulations of modality) in these registers and in other varieties of English, such as British English and Australian English.