Samples for this research were extracted from the International Corpus of English-Great Britain (ICE-GB; Greenbaum, 1996). Sub-component; a one-million-token corpus comprised of written and transcribed spoken discourse from native British-English speakers. The corpus contains 300 spoken and 200 written text units. Among the 300 transcribed spoken texts, 50 scripted speech texts are more characteristic of a written discourse and, thus, can be classified into written discourse (Kaltenböck, 2005). In this sense, the ICE-GB consists of an equal share of spoken and written texts.

The ICE-GB is intensively marked with grammatical features. The anticipatory it of the INTRODUCTORY IT PATTERN is marked with “PRSU” and “ANTIT” (“provisional subject” and “anticipatory it,” respectively), and the extraposed subject clause is marked with “NOSU, CL” (for “notional subject clause”). Meanwhile, the main clause predicates are also syntactically parsed. These markups are helpful for the syntactic parsing of the INTRODUCTORY IT PATTERN. By locating the “NOSU, CL” markup in each text file in ICE-GB, we identified all the INTRODUCTORY IT PATTERNs with Python 3.6.5. In addition, guided by the text unit or sentence boundary markup in the ICE-GB, we extracted the preceding seven sentences totaling to over 100 tokens as the pre-context, which is the basis for a proper interpretation of stance expression (Du Bois, 2007, p. 149).

After manual selection, we ended up with 1,746 cases, among which, 12 cases had coordinated predicate groups. Each coordinated predicate group (e.g., it may be too late or not desirable to make any change) was treated as two predicates (e.g., may be too late and may be not desirable). Therefore, the total number of predicates was 1,758. Among them, 1,042 (59.27%) were adjectival predicates and 472 were verbal predicates (26.85%), thus, constituting 86.12% of all main clause predicates. The others were nominal predicates (10.07%) and adverbial predicates (3.8%). As this study focuses on adjectival and verbal predicates to overcome data sparsity, we, therefore, concentrate on the combined total of 1,514 adjectival and verbal predicates.

Collexeme analysis analyzes the attractions between lexemes and various abstract (e.g., passivation, dative) or specific constructions by calculating collostruction strength, which is “the degree to which particular slots in a grammatical structure prefer, or are restricted to, a particular set or semantic class of lexical items” (Stefanowitsch and Gries, 2003, p. 211). The analysis has been performed with the R script from Gries (2007). Methods generally used to calculate collostructional strength (CLS) in the R script include the Fisher-Yates Exact test, log-likelihood, chi-square, and odds ratio (Gries, 2007). Following Larsson and Kaatari (2019), this study applied the Fisher-Yates Exact test to statistically measure the CLSs between lexemes and grammatical patterns. Regardless of the choice of the statistical method, lexemes whose CLSs are above the threshold value are considered significantly attracted to the target slot and are called “collexemes” (Stefanowitsch and Gries, 2003).

First, two native English speakers were recruited as the annotators for the manual classification. The annotation tool was the UAM CorpusTool 3.3s. The classification was based on the interpretative result by annotators of what the main clause predicate commented about the content of the extraposed subject clause. The scope of the main clause predicate is exemplified by the underlined segment in (2). For all example sentences starting from (2), the source of the example sentence is provided at its end in the brackets (e.g., “s2b-032#039”) to indicate the name of the corpus file (e.g., “s2b-032#039”) and the exact index of the sentence (e.g., “039”).

To obtain reliable annotation results, the researchers trained the two annotators for the manual classification. The two annotators were first imparted necessary grammatical knowledge about the INTRODUCTORY IT PATTERN, and were then taught about the framework of Herriman (2000a) and the operational procedures of the annotation tool. For the few rounds of training, the annotators had several hands-on annotations of the random samples and their inter-annotator agreements (IAA) were calculated with Krippendorff’s alpha. After the IAAs were steadily above 0.8, which is the cut-off point for good agreements (Krippendorff, 2019), the training was considered successful. The two annotators were then given all the samples of the INTRODUCTORY IT PATTERN for annotation.

After the annotation was completed, the results were statistically analyzed to ensure their validity. The raw agreement for the annotation was 0.904 and the IAA was 0.884 (>0.8), suggesting good and reliable results. The disputed cases (n = 151) were then discussed by the two annotators to reach agreements. Six cases could not be agreed upon due to the different interpretations of the two annotators of the same cases. These cases were then arbitrated by the researchers.

Second, a collexeme analysis was performed on the INTRODUCTORY IT PATTERN. One vital step was the selection of the target slot in the main clause predicate of the pattern. Guided by our research goal of revealing the stance meanings of the INTRODUCTORY IT PATTERN, the target slot should host the essential stance meaning of the main clause predicate. For our target construction, the two subtypes of INTRODUCTORY IT PATTERN with adjectival and verbal predicates, we selected the predicate head position as the target slot. In this regard, the two subtypes of the pattern can be expressed as it verb-link X_ADJ and it X_VERB, within which the X_ADJ/X_VERB is the predicate head and the verb-link is the linking verb, such as copular verbs. In addition, other linguistic elements in the main clause predicate might include modal auxiliaries, semi-modals, adverbs, or negation devices. Thus, within the example sentences in (2), the predicate heads are interesting, common, possible, and right. In these and other cases, X_ADJ represents the essential semantic content of the main clause (Mair, 1990; Larsson, 2016).

However, the X_VERB is slightly different. It can be in an active or passive voice and can be an uncomplemented verb, a verb with various forms of complementation, or an idiomatic verb phrase (e.g., turn out, make sense). Accordingly, the X_VERB alone might represent the essential semantic content of the verbal predicate like the underlined verbs in it follows that and it was found that. In this case, the X_VERB could be modified by auxiliaries or adverbials, or it could be negated, but the meaning expressed by the whole predicate is closely associated with the meaning of X_VERB (Goldberg, 2006, p. 28). Alternatively, the X_VERB can be verbs with general meanings (Dixon, 2005; Goldberg, 2006), or can bear less semantic weight than their complements, like the come in it has come to my attention. Such cases are rare (n = 4). Meanwhile, the range of complementation to the general verbs is restricted. Thus, when used as a collexeme of the INTRODUCTORY IT PATTERN, come only occurs in the phrasal form of come to my attention/notice (n = 3), meaning being perceived. The restricted use of come is associated with the grammatical role of the come phrase as the predicate of the INTRODUCTORY IT PATTERN because the clausal semantic subject of the phrase is obviously incapable of coming (Dixon, 2005), but of being perceived.

Likewise, for other verbs that do not represent the complete essential meanings of the predicates, their occurrence in the X_VERB slot is strongly associated with one corresponding stance meaning. For example, the verb take in X_VERB signals the expression of Circumstance stance subtype, as it is almost invariantly used to comment on empirical conditions (e.g., it takes us years). Other examples include annoy for Emotive Reaction (e.g., it annoys me to), and dawn for Perception (e.g., it dawned upon her that). Thus, each verb occurring in the X_VERB slot is closely associated with one subtype of stance (Goldberg, 2006, p. 28). Therefore, it is fair to say that in the case of the INTRODUCTORY IT PATTERN, the X_VERB can represent the meaning of the main clause predicate.

Therefore, the predicate head slot (X_ADJ and X_VERB) of adjectival and verbal predicates can represent the essential semantic content of the two predicates and, thus is an ideal target slot for collexeme analysis.

After the target slot was settled, the main clause predicate heads were manually selected and were lemmatized into lexemes. The linguistic data required for a collexeme analysis include the observed frequency count of each lexeme in the target slot, the corpus frequency count of each lexeme, the observed frequency count of the INTRODUCTORY IT PATTERN, as well as the corpus size (Stefanowitsch and Gries, 2003; Gries, 2007; Hilpert, 2014). Based on these data, the collexeme analysis was then performed with an R script (Gries, 2007) to calculate the CLSs of lexemes that occur in the target slot of the INTRODUCTORY IT PATTERN. The significant CLS threshold was set to >3 (p < 0.001) to ensure a high linguistic significance of collexemes, and, therefore, to render lexemes with CLS values that are higher than 3, as the valid collexemes, for our analysis. Consequently, we ended up with 124 collexemes that accounted for 82.17% of the stance expression by the INTRODUCTORY IT PATTERN, with adjectival and verbal predicates.

Apart from CLS values, the dispersion information of collexemes was also considered to rule out misguided findings due to the uneven distribution of the INTRODUCTORY IT PATTERN (Gries, 2008, 2019). The measurement of dispersion was 1-DP (Gries, 2008) and its values for all the 124 collexemes were around 0.8, suggesting an even distribution of the INTRODUCTORY IT PATTERN in ICE-GB.

Finally, the 124 collexemes were checked for their association with different stance subtypes according to our annotation results. As the results suggested, out of the 124 collexemes, 109 featured a strict one-to-one association with the stance subtypes. For example, interesting is solely used to denote an Emotive Reaction [see (2a)], common for Frequency [see (2b)], and right for Appropriateness [see (2d)]. Seven other collexemes are primarily used to express one stance subtype but can be used to express other subtypes as well. The two groups of collexemes, totaling to 116 collexemes, will be reported with their associated stance types/subtypes in the following two sections.

The remaining eight collexemes are not prone to be associated with specific stance subtypes. Instead, they are indiscriminately used to express multiple subtypes of a stance with no statistical inclinations. These collexemes occurred 13 times in the target slot. Their multiple uses resulted from their polysemous nature and the differing linguistic circumstance. For example, extraordinary could mean exceedingly good or strange in different linguistic contexts, thus, denoting General Evaluation and Appropriateness, respectively. However, a small number of such cases will be insufficient to reveal about the stance meaning of the INTRODUCTORY IT PATTERN. Thus, these collexemes will not be discussed in the following sections.

According to Table 1, Epistemic stance is the most frequently expressed stance type (n = 437), and Figure 1 reports the lexical basis for its linguistic realization. According to Figure 1, the collexemes most strongly associated with the expression of the Existence, Perception, and Truth subtypes are turn out (with the meaning of appear to be), the intransitive seem, and the adjectival clear, respectively.

To express the Existence subtype, collexemes (as reported in Figure 1) highlight a direct or indirect perception of the existence or coming into existence of events/propositions. The prototypical Existence collexeme turn out, possessing the highest CLSs among all Existence collexemes, primarily introduces the observed events/propositions. For example, in (5), the speaker was a judge analyzing a case of business’ secret infringement, where testimonies were needed to prove that the secret was indeed treated as such. In this case, the testimonies that would be directly perceived in the court could tell that the plaintiff was right. Hence, the perception verbal phrase turn out presents the observed propositions/events or expresses Direct evidentiality. The same function is also seen with discover. A few other Existence collexemes, emerge and transpire, can express both Direct and Indirect evidentiality. For example, the emerge in (6a) presents a proposition about the meaning of a word that is deducible from other evidence. Alternatively, in (6b), the emerge introduces visually observable events, thus, expressing a Direct evidentiality. The same multiple uses were observed with transpire.

Furthermore, another Existence collexeme, follow, is only used to express Indirect evidentiality. It displays Indirect evidentiality by indicating that propositions/events should happen rather than have happened or are observed. For example, the follow in sentence (7) is used to argue that an event will not necessarily happen because it is not supported with a convincing reasoning. In this case, the judgment is not based on a direct observation but on an indirect deduction. In other cases, follow is complemented by from-prepositional phrase (PP) to suggest the basis of reasoning, as in it follows from that/the very terms of that/those assumptions.

According to the above results, the Existence subtype boils down to Direct and Indirect evidentiality as primarily actualized by verbs of perception. This finding is in line with Francis et al. (1996). Additionally, we found that the prototypical Existence collexeme, turn out, whose CLS is the highest, manifests Direct evidentiality while others feature Indirect, or both Indirect and Direct evidentiality. Hence, the expression of Existence via the INTRODUCTORY IT PATTERN features a mixed expression of evidentiality, which mainly features a non-modalized/Direct evidentiality.

Compared with Existence, the subtype of Perception put more emphasis on Indirect evidentiality, indicating that the perceived events/propositions originate from an indirectly observed evidence. Among Perception collexemes, the five collexemes with lower CLS values (see Figure 1) must be complemented to indicate indirect evidence or the lack of direct evidence. For example, look and sound are somehow complemented by the co-occurring subordinators, including as if, as though, and like, to present propositions/events that can only be felt or inferred to be the case/likely/true (Field, 1997; Dixon, 2005). Likewise, occur has to be complemented by to-PP, strike co-occurs with me, and dawn with on-phrase or upon-phrase to indicate that events/propositions are perceived in the inner world of writer-speakers. In comparison, the collexemes, seem and appear, can be interpreted as seem to be true/the case/likely or appear to be the case. Hence, they can deliver the probability/likelihood judgment without complementation. Nonetheless, seem can be complemented by to-PP (47.57%) to reveal the perceiver as a first-person perceiver [me for 96.55% vs. us for 3.45%; see (8)], while appear is almost invariantly used without any complementation and is associated with events/propositions expected to happen [see (9a) and (9b)].

The use of Perception collexemes and their complementation statistics entail different degrees of (inter)subjectivity in the expression of Indirect evidentiality or epistemic judgment. The expression can lean toward the intersubjective end by revealing the perceiver as a first-person or third-person perceiver (Whitt, 2011) via various forms of complementation, thus, highlighting a self-sourced or other-sourced perception. As the self-sourced perception is less reliable than the other-sourced perception, an expression of evidentiality that features the first-person perceiver, thus, entails higher subjectivity (Whitt, 2011). Most importantly, the Perception collexeme possessing the highest CLS value is seem (see Figure 1), which features a less solid observation of the inner world of reasoning/inference of the first-person perceiver. The high CLS of seem suggests that the INTRODUCTORY IT PATTERN favors more subjectivity in the expression of Perception.

As opposed to Existence and Perception, the subtype of Truth displayed mixed modes of information presentation that either explicitly evaluates likelihood or exhibits a modalized Indirect evidentiality. The two modes are realized by adjectival and verbal Truth collexemes, respectively. Among adjectival Truth collexemes, clear has the highest CLS value, followed by likely/unlikely, and then by obvious/true. Among them, clear is frequently modified (25.64%) by intensifying degree modifiers, such as quite and very [see (10)], but is hardly modified (n = 1) by the totality reinforcing modifiers (e.g., absolutely or definitely). Moreover, it can be used in comparative form, as in it could not be clearer. Comparatively, likely and unlikely are more frequently pre-modified (31.82%) by a wider range of degree intensifiers, especially by comparative adverbials, more and less, to be used in a comparative form. According to Paradis (2001), modification by intensifying modifiers and by the comparative/superlative use of adjectives indicate their scalarized meanings. With reference to these standards by Paradis (2001), clear and likely/unlikely feature scalarized meaning. Moreover, likely/unlike manifests typical characteristics of scalarized meaning as they are more frequently used in the comparative form than clear. In contrast, adjectives with extreme meanings are modified by the totality reinforcing modifiers like definitely and almost (Paradis, 2001). Thus, clear possesses a scalarized meaning that leans toward the extreme end as it can be occasionally modified by the totality adverbial perfectly.

In comparison, obvious and true are the extreme Truth collexemes denoting a higher certainty because they are rarely modified by degree intensifiers and can be pre-modified by totality adverbials [see (10b)]. Thus, obvious and true are less typical of scalarized adjectives but more characteristic of extreme adjectives. Adding up to the scalar, there is also a “dubitative” likelihood judgment, as in it is debatable or it is not clear, functioning to discharge the epistemic obligation of writer-speakers (Marín-Arrese, 2021).

Compared to adjectival Truth collexemes, the verbal Truth collexemes indicate the source of evidence for events/propositions. Among these collexemes, verbal collexemes of reporting feature a moderately modalized Reported evidentiality (see the previous section for detail), while the remaining verbal Truth collexemes manifested a modalized Indirect evidentiality by suggesting the existence of some supporting evidence for the presented events/propositions. For estimate and calculate, the evidence is the certain measurement (Dixon, 2005). On the other hand, for prove, accept and expect, it is some convincing evidence, consensus, and generally known routines, respectively. Meanwhile, for assume, allege, and recognize, the reason is the certain personally held convictions over the likelihood of the presented events/propositions [see (11)]. Interestingly, these remaining verbal collexemes are rarely pre-modified by modal verbs to alter the expression of stance, like the case of reporting collexeme say. The occasional pre-modification simply functions to intensify or tone down the degree of certainty in the likelihood judgment, as in it must be presumed that or it might be expected that.

Therefore, Existence, Perception, and Truth expressed by verbal collexemes, demonstrate evidentiality that is modalized to different degrees. Perception expresses fully modalized/Indirect evidentiality that features higher subjectivity, while Existence and Truth display a mixture of modalized and non-modalized evidentiality. The modalization is observed with verbal collexemes of existence and perception, reporting as well as verbs of cognitive activity despite the Direct/Reported evidentiality entailed in their lexical semantics. In some rare cases, reporting verbs are used to deliver deontic meanings (e.g., it must be said/admitted that), disconnected from the Indirect evidentiality or Reported evidentiality (Auwera and Plungian, 1998), resulting in a mere deontic assessment of events/propositions. A plausible conclusion is that non-evaluative lexis is associated with evaluative meanings when used as collexemes of the INTRODUCTORY IT PATTERN via the modalized evidentiality and the linguistic circumstance, especially one with modal verbs.

It is, therefore, inefficient and inaccurate to maintain the ternary division of Epistemic stance into Existence, Perception, and Truth because they fail to adequately describe the meanings or uses of the verbal collexemes. For example, verbs of existence/occurrence also express the inference/reasoning of writer-speakers, and, thus, should not simply deliver “Existence” but also the epistemic judgment of writer-speakers. Moreover, it should be noted that this epistemic judgment is within the domain of Epistemic stance while the existence/occurrence observation belongs to that of the evidentiality. Based on our research results, it will be more appropriate to have a bipartite division of the Epistemic stance into the Modalized Probability and General Probability associated with verbal collexemes and adjectival collexemes, respectively. The Modalized Probability is a likelihood judgment that also indicates the source of the judgment, including the inner perception [e.g., seem, appear, and look/sound (as if)], general reasoning/inference (e.g., it can be argued; and it could be said), assumption (e.g., assume and allege), certain measurement, and the routine. In comparison, General Probability is a general likelihood judgment featuring different degrees of certainty in the judgment. It favors an intermediate likelihood judgment typically expressed with clear, likely, and unlikely, which, in turn, reflects a lack of evidence and certainty (Haβler, 2010). The INTRODUCTORY IT PATTERN, thus, features a modalization cline of evidentiality ranging from (a) non-modalized observation or reporting, (b) grammatically marked modalization (e.g., it can be said), to (c) semantically oriented modalization actualized by verbal and adjectival predicates that point to a lack of direct access to events/propositions. Along this line, the expression of Epistemic stance by the INTRODUCTORY IT PATTERN is at the modalized end of the cline covering phases (b) and (c).

Besides, Epistemic stance demonstrates the tendency to display positive likelihood judgment. Details reported in Table 5 show that the Epistemic stance features an overall emphasis (86.04%) on a positive likelihood judgment. Moreover, such tendency is highlighted by verbal collexemes as they are almost solely used (above 93% occurrences) to express this positive judgment. In contrast, more adjectival Epistemic collexemes are used to express negative (12.89%; e.g., it is hardly probable) and neutral/dubitative (12.37%; e.g., it is not clear) likelihood judgment.

The prevalent positive likelihood judgment of the INTRODUCTORY IT PATTERN echoes the results of Hyland and Tse (2005), where a general positive Epistemic stance (55.5%) features the expression of Epistemic stance by the evaluative that construction. Interestingly, our findings show that the INTRODUCTORY IT PATTERN demonstrates a more robust tendency to express the positive Epistemic stance. Such tendency is more prevalently associated with Modalized Probability, indicating the overall tendency for modalized evidentiality to be affirmative.

Details reported in Table 3 show that the variety of collexemes expressing the Deontic stance was limited, with only five associated collexemes. Furthermore, observations of their use revealed that Deontic collexemes were associated with different modal verbs.

The expression of the Obligation subtype shows the strongest association with the adjectival collexeme necessary, which can be modified by possibility/permission/ability modal verbs, may [n = 5; see (12)] and could (n = 1), or prediction/volition modal verbs will (n = 2) and would (n = 1). However, no obligation modals or semi-modals, including must, should, and have to, are used with it. As for the other less attracted collexemes, such as essential and imperative, no pre-modifying modal verbs are seen. However, several main verbs of the extraposed subject clauses are pre-modified by the obligation modal, must [n = 1; see (13a)], and semi-modal, have to (n = 1). Besides, among the extraposed subject clauses that follow imperative, two cases of present subjunctive mood are found [see (13b); n = 2].

For the two Volition collexemes, hope and desire, associations with different modal verbs are also found with their use. The verbal collexeme hope are followed by extraposed that-clauses (n = 4), and that-clause with that omitted (short as “zero that-clause”) (n = 1). All extraposed that clauses following hope have modal verbs that pre-modifies the main verb, including the volition/prediction modal will, and possibility/permission modals might and can. In contrast, the verbal collexeme desire uniformly occurs with the extraposed infinitival clauses [see (14)], where the use of modal verbs, subordinators, and grammatical subjects is blocked. Hence, the linguistic elements between desire and the main verb of the extraposed infinitival clause are reduced to a minimum. The eagerness connotated in desire is, thus, maintained.

The above findings on the use of Deontic collexemes point to the different degrees of obligation and volition connotated in the Obligation and Volition collexemes. The concept of the degree of obligatoriness is mentioned in Nuyts (2005, p. 23) and in Van linden and Verstraete (2011). Likewise, the degrees of volition are also noted (Dixon, 2005). In this study, the varying degrees in expressing Obligation and Volition are evident from two formal clues: the pre-modifying modal verbs of the collexemes, and the main verbs in the extraposed subject clauses. The former provides circumstantial meanings that can be read into the connotations of the collexemes (Martin and White, 2005, p. 65), while the latter indicates the connotations of the collexemes as the modal meanings are harmonically assimilated into the collexemes/the main clause predicate heads (Lyons, 1977; Bybee et al., 1994, p. 214). Therefore, by referring to the two formal clues, the semantic connotations of collexemes can be deciphered. Thus, when used as collexemes of the INTRODUCTORY IT PATTERN, necessary possesses weaker obligatoriness than essential or imperative because only the latter are accompanied with modal verbs of obligations (must, have to), and hope entails weaker volition than desire as it is not modified by obligation modal verbs but co-occurs with modal verbs of possibility/permission (e.g., can, could, and might). The scalar difference in the expression of obligation and volition is in line with findings by Van linden and Verstraete (2011) and also by Dixon (2005).

In addition, our results show that collexemes with lower obligatoriness/volition (e.g., necessary and hope) have higher CLS values than those with higher obligatoriness/volition. According to Table 3, among Obligation collexemes, the CLS of necessary is 91.96 while those of imperative and essential are considerably lower: 10.22 and 7.36, respectively. Likewise, between the two Volition collexemes, the CLS of hope, which entails lower volition, is also higher than desire. Thus, the fair conclusion is that the INTRODUCTORY IT PATTERN favors collexemes with lower obligatoriness/volition for the expression of the Deontic stance. This lower obligatoriness/volition subsequently connotates less desirability on the part of the writer-speakers. The preference of the INTRODUCTORY IT PATTERN for lower obligation/volition aligns with its generally impersonal and objective nature (Collins, 1994; Gómez-González, 1997; Herriman, 2000a; Berman, 2004; Biber, 2006; Hamawand, 2007), resulting in a reserved or a negotiable expression of stance.

The expression of the Dynamic stance builds on five collexemes that denote the subtype of Ease-of-Performance, and four collexemes that deliver the subtype of Circumstance. These Dynamic collexemes (ranked in descending order by their CLSs) are possible, difficult, easy, hard, and impossible for Ease-of-Performance, then take, cost, safe, and dangerous for Circumstance. Among the Ease-of-Performance collexemes, possible and impossible can express both Ease-of-Performance [see (15a)] and Truth [see (15b)]. Between them, 76 out of 100 occurrences of possible express Ease-of-Performance, and 26 out of 28 occurrences of impossible denote the same stance subtype. Thus, these two are primarily associated with Ease-of-Performance and are only discussed as Ease-of-Performance collexemes in this section.

Observations on the use of Ease-of-Performance collexemes reveal how a limited set of collexemes can designate Ease-of-Performance. These collexemes form three natural meaning groups, presented in Table 4 that respectively assessed the evaluated events/propositions as easy, potential, and difficult/impossible. Reported together are the negation statistics, as negation also influences the groupings.

According to Table 4, the collexeme possible, categorized under the “potential” group, primarily (81.08%) assesses events/propositions as potentially realizable, but it can also be negated (18.92%) to denote difficulty/impossibility. Meanwhile, the three collexemes, difficult, impossible, and hard, are in the “difficult” group, and most of their occurrences (98.26, 100, and 100%, respectively) denoted the difficulty of performance. The “easy” group only contains easy. Although most of its uses as an Ease-of-Performance collexeme denoted the ease of performance, in a small fraction of its occurrences (8.11%), easy is negated to denote the difficulty of performance.

Therefore, although Ease-of-Performance collexemes are predominantly associated with one Ease-of-Performance assessment result (“easy,” “potential,” or “difficult”), they can all express the difficulty of performance. The total count of collexemes denoting the difficulty of performance, including the negated use of possible and easy, and the affirmative use of collexemes from the “difficult” group amounts to 147, constituting 60% occurrences of Ease-of-Performance collexemes. Hence, the prevalent meaning of the Ease-of-Performance subtype is difficulty-of-performance.

Meanwhile, for the subtype of Circumstance, no degrees or scalar changes in the stance expression are observed in the use of Circumstance collexemes. However, each Circumstance collexeme is associated with one major stance meaning. The two verbal collexemes, take and cost, are idiomatically used to depict the financial cost, labor, time, and other resources needed to realize the events/propositions denoted by the extraposed subject clauses. Moreover, 75.76% occurrences of take comment on the amount of time [see (16)], while cost is invariantly used to specify the financial costs [see (17)]. The two collexemes, take and cost, are, thus, supplementary for the expression of Circumstance. A much different aspect of Circumstance is shown by the two adjectival Circumstance collexemes, safe and dangerous, used to describe the general environments or the atmosphere related to realizing events/propositions.

Thus, the two subtypes of the Dynamic stance, Ease-of-Performance and Circumstance, emphasize different aspects of the empirical laws governing the realization of the events/proposition denoted by the extraposed subject clause. The subtype of Ease-of-Performance expresses the degrees of ease in actualizing events/propositions, leaning toward an overall difficulty-of-performance assessment. In contrast, no degrees of circumstantial requirements are observed in the expression of Circumstance, which is in line with the general non-scalar nature of circumstantial stance subtype (Nuyts, 2014). Instead, Circumstance collexemes, like take and cost, are conventionally used to assess the different resources required to actualize events/propositions, reflecting the highly idiomatic use of the INTRODUCTORY IT PATTERN (Mair, 1990, p. 23; Kaltenböck, 2005).

The range of collexemes that express the Evaluation stance is reported in Figure 2. For better visualization of Figure 2, (i) the CLS values of collexemes were logged and amplified [5*log₂(N-2)], and (ii) the names of stance subtypes were abbreviated to the initial three to five letters (e.g., “Appro” for “Appropriateness”).

According to Figure 2, Evaluation collexemes are rich in lexical variety, totaling 57 collexemes. Among all six subtypes of the Evaluation stance, Appropriateness features the highest diversity, constituting 42.11% of all the Evaluation collexemes. In comparison, other subtypes, especially the Frequency subtype, has considerably fewer lexical varieties (only four collexemes). Besides, the subtype of Responsibility is not expressed by any adjectival or verbal collexeme and, thus, is not shown in Figure 2.

Figure 2 also shows that some adjectival collexemes have extraordinarily high CLSs. They are important, good, nice, helpful, surprising, and interesting, with important scoring the highest. Due to space limitations, our discussion on Evaluation collexemes will be illustrated through those with the highest CLSs.

The most attracted Evaluation collexeme, important, is a Significance collexeme. While it invariantly comments events/propositions as significant, the significance assessment affords multiple behavior implications. When important is followed by infinitival clauses (65.63%), it obliges the reader-hearers to follow the instructions expressed in the infinitival clauses [see (a) and (b) in (18)]. In addition, contrary to the claim that important only obliges reader-hearers to actualize the content of infinitival clauses (Van linden and Verstraete, 2011), our results show that the deontic meaning of important is equally detectable in almost all cases (95.24%), where the extraposed subject clauses are that-clauses/zero that-clauses (32.81%). In (19a), for example, important is the deontic as it obliges the reader-hearers to actualize the events/propositions denoted by the extraposed subject clause. This deontic meaning is supported by the fact that important is compatible with obligation/necessity modal verbs, should and must, as in (19a″), but is unacceptable with the possibility/permission modals can/could/might, as in (19a″). Thus, important is primarily (96.88%) deontic and delivers deontic assessment toward the events/propositions expressed by both infinitival and that/zero that clauses. In the two cases where important do not deliver deontic meaning, it comments on events/propositions that do not need to be realized because they are past events [see (21(a)], or the answer to a yes-no question [see 21(b)]. In both cases, the important is evaluative like another Significance collexeme significant [see (21c)]. Nonetheless, the collexeme is primarily a Deontic collexeme.

Similarly, the verbal Significance collexeme, note, invariantly used in the form of should/must be noted also delivers deontic meaning [see (20)] as it obliges reader-hearers to note the events/propositions encoded in the extraposed subject clauses. Hence, the two Significance collexemes, closely associated with the deontic meanings afforded by the obligation modal verb should, deliver Deontic stance.

As opposed to important and note, a sense of deontic meaning is only vaguely observable but equally deniable with the use of other Significance collexemes. For example, the noteworthy in (22b) comments events/proposition as worthy of notice rather than obliging reader-hearers to note. For other Significance collexemes, including verbal matter [see (22a)], significant [see (21c)], and pointless [see (22c)], the stance meaning is also more evaluative than deontic. Therefore, apart from important, Significance collexemes generally comment on the significance of events/propositions but do not oblige reader-hearers with regards to the assessed events/propositions.

Following the distinctive uses of important and note as opposed to other Significance collexemes, the natural conclusion is that important and note should be Obligation or Deontic collexemes, as opposed to a Significance collexeme, as proposed by Herriman (2000a) and Zhang (2015). In addition, note is associated with stronger obligatoriness than important as the former is constantly used for deontic meaning, while the latter can be used for evaluative purposes like other Significance collexemes [see 21(a) and 21(b)]. The relatively weaker deontic meaning connotated in important is also mentioned by Van linden and Verstraete (2011). Furthermore, our study shows that important possesses lower obligatoriness than necessary, with the latter observed to demonstrate a constant deontic use in our corpus just like note. Additionally, important has stronger attraction to the INTRODUCTORY IT PATTERN than necessary and note (101.21 > 91.96 > 5.44). Hence, with important and note categorized as Obligation collexemes, it still holds true that Obligation collexemes with weaker obligatoriness (important < necessary < essential/imperative/note) have stronger attractions to the INTRODUCTORY IT PATTERN.

The group of collexemes with the second highest CLSs are good [see (23a)], nice, and helpful, denoting General Evaluation. Among them, good has an observable deontic flavor (47.17%), when used in the comparative [see (23b)] or superlative form [see (23c)]. In comparison, good used in the base form is more evaluative than deontic [see (23a)], whereas, for most other General Evaluation collexemes, a hint of deontic meaning seems acceptable but equally deniable. For example, nice in sentences (24a) and (24c) is evaluative for commenting on events that already have happened or are happening. Alternatively, nice is deontic in (24b) and (24d), when the situational setting (a couple was discussing about refurbishing the room) can provoke the deontic meaning, as in (24b), or when the events/propositions are hypothetical or desired, as in (23d). Thus, the stance meanings of General Evaluation collexemes vary between favorability and deontic meaning.

The collexemes with the third highest CLSs are Emotive Reaction collexemes, including surprising and interesting. The use of surprising together with the verbal collexeme, surprise, is primarily (71.43%) negated (e.g., not surprising and hardly surprising), assessing an event/proposition as expected, whereas interesting is used in the affirmative form to arouse the attention of reader-hearers (Hewings and Hewings, 2002) to the unexpected events/propositions. Thus, surprising and interesting are considered “expectedness” adjectives (Groom, 2005). However, expectedness is simply the most typically expressed Emotive Reaction. Other emotions also include sad/dismaying/disconcerting and fascinating/amazing/tempting.

The collexemes with the fourth highest CLSs are the three Appropriateness collexemes: adjectival right, sufficient, and strange. The use of right expresses polarized appropriateness judgment based on general conventions inferable from the context. For example, in (25a), these actions denote the outrageous misconduct by several police officers. In this case, the general conventions for appropriateness judgment are those of social justice. In (25b), the general conventions are principles or laws of economics/finance. As for sufficient, the convention is the general agreement on the appropriate amount or degree; for strange, it is the general conventions on normality.

The group of collexemes whose CLSs are the lowest among all Evaluation collexemes is the Frequency collexemes. Frequency collexemes assess events/propositions as the standard or non-standard practice. For example, (26a) described a situation where symbolic representations were publicly acceptable, which is different from the latter period under discussion. In cases like (26b), a sense of inappropriateness is observable from the use of unusual, evaluating events/propositions as disobeying the commonly accepted standard.

Evaluation collexemes can be about various value judgments that result in deontic meanings, or they can simply deliver evaluative meanings. However, the boundary between these two meanings is vague. For example, the General Evaluation collexeme, good, when used in the forms of better or best, is deontic (Collins, 1994), but these forms can sometimes be evaluative (Herriman, 2000a). The same vague deontic reading can be perceived in the use of some Appropriateness collexemes, which is also observed by Van linden and Verstraete (2011).

The reasons behind the expression of vague deontic meanings have raised some interest in scholars. A possible explanation involves some “face work,” which modulates the expression of deontic meanings (Brown and Levinson, 1987; Nuyts et al., 2010), thus resulting in the blurry boundary between purely evaluative and deontic meanings. In this study, writer-speakers assess propositions/events as significant, favorable, acceptable, or standard practices as a polite alternative to suggest that members in the community of writer-speakers should follow the propositions, or rather actualize the assessed events to meet various general conventions. In this sense, the Evaluation stance demonstrates more politeness than the Deontic stance, as its deontic connotation is only vaguely perceivable. Hence, the overwhelming prevalence of the Evaluation stance is of pragmatical significance.