11.3.1 L1-L2 Contrast and Input Demand

A clear source of non-canonical syntax among non-native speakers of a language is their first language(s) (L1). Examples include a much more frequent use of existing fronting devices (topicalisation, left dislocation) and novel topic-marking devices in many Asian varieties of English (Lange Reference Lange2012; Leuckert Reference Leuckert2019). These can involve constructions fairly directly transferred from the first languages (Can or not?: SgE, syntactic format transferred from Sinitic substrates) or indirect attempts to reconstitute a feature of the first language through constructions available in English (Every year, inflation is there: IndE, partly recreating the Indo-Aryan use of clause-final copula for existential meaning). Bao (Reference Bao2015) describes this as a ‘filtering’ of substrate meanings through superstrate lexicon: semantic functions of the L1 come to be reconstituted in reorganised elements of available L2 syntax. These are not simply errors of course; they have developed their own grammatical rules and speakers have intuitions concerning these constructions (Parshad et al. Reference Parshad, Bhowmick, Chand, Kumari and Sinhad2016).

Some have argued that non-canonical usage in a contact variety arises in part due to complexity within the English system, suggesting an acquisitional source of divergence (Housen Reference Housen, Rafael Salaberry and Shirai2002; Davydova Reference Davydova2011). This can be refined to be a statement about difficulty relative to the structures available in the L1. Sharma (Reference Sharma2023) argues that this interpretation of difficulty – termed input demand, or how much input a speaker of a given L1 needs to acquire an L2 form – may account for why certain non-canonical syntactic constructions persist more over time and become established as a new dialect feature while others do not.

Some syntactic innovations arise repeatedly across non-native English usage. Kortmann and Szmrecsanyi (Reference Kortmann, Szmrecsanyi, Kortmann, Burridge, Mesthrie, Schneider and Upton2004) use the term ‘varioversals’ for such forms, due to their shared origin in the variety type, for example, arising out of acquisitional processes or restricted input in non-native contexts. Some examples are given in (1) below;Footnote ¹ Mauranen et al. (Reference Mauranen, Carey, Ranta, Biber and Reppen2015) cite similar features as potentially arising across ELF situations as well.

1. (1)

   a.	Irregular use of articles (omission):
   	i.	His sister passed message to me.	Singapore
   	ii.	… like mudbrick hut with thatch roof.	West Africa
   b.	Present perfect and simple past levelled (occurrence with punctual adverbials):
   	i.	I have seen him yesterday.	Philippines
   	ii.	Nine supporters of Malawi’s president-elect Mr Bakili Muluzi have been killed when a bus crashed into them as they celebrated their victory.	East Africa
   c.	Wider range of uses of the progressive (with stative verbs):
   	i.	We are knowing each other.	India
   	ii.	Maybe one or two is having HIV/Aids.	South Africa
   d.	Resumptive/shadow pronouns (with subject and object):
   	i.	Some teachers when I was in high school I liked them very much.	Papua New Guinea
   	ii.	My daughter she is attending the University of Nairobi.	East Africa
   e.	Loosening of sequence-of-tense rules (past perfect after present):
   	i.	Never before in the Capital’s history these colonies had faced such a flood threat.	India
   	ii.	Most of these syringes and razors you find that they had been used by someone before.	South Africa
   f.	Invariant non-concord tags (negative):
   	i.	Upili returned the book, isn’t it?	Sri Lanka
   	ii.	He loves you, isn’t it?	West Africa

Some non-canonical innovations in a contact setting arise from more general discourse-driven principles – not strictly grafted from first languages, but drawing on universal or logical tendencies that emerge due to wider mismatches between L1 and L2 and foregrounding of pragmatic functions (Lange Reference Lange2012; Leuckert Reference Leuckert2019). Current theoretical models in bilingualism research such as the Interface Hypothesis can also offer new insights into why we might repeatedly observe discourse-based solutions to the challenge of mismatched grammars in contact (Sharma Reference Sharma2023).

11.3.2 Non-Canonical Alternatives within Standard English

As L1 sources of change are so prominent and noticeable in non-native varieties, loci of variability within English tend to be less of a focus as a source of new usage, but they play a key part in the actuation and spread of non-canonical syntactic usage in second and foreign language varieties (Sharma Reference Sharma2001; Hundt & Vogel Reference Hundt, Vogel, Mukherjee and Hundt2011). Indeed, there is a risk of over-idealising how homogeneous canonical, standard, or unmarked usage is in native varieties. Native and standard English varieties encompass a substantial degree of variability, as attested in the preceding parts of this book. An example noted by Ranta (Reference Ranta, Mauranen and Ranta2009) is lack of plural agreement in existential there-constructions in English: ELF speakers are often penalised for utterances such as There’s people outside, on the false assumption that native speakers do not produce such forms.

As noted, one form of change in non-native speakers may be an increase in the relative frequency of these ‘built-in’ non-canonical constructions. For example, IndE has dramatically higher frequencies of topicalisation and left dislocation constructions than British English, a difference in frequency rather than form. (As previously mentioned, IndE also allows non-canonical constructions that fall outside of the set of constructions licensed in British English at all.)

An example of repurposing an existing non-canonical form within Standard British English for a new non-canonical system in a new dialect is the development of a new topic-marking function of the relativiser who in Multicultural London English (MLE). Cheshire et al. (Reference Cheshire, Adger and Fox2013) compare the use of different relativisers in two varieties of English: traditional London English and the newer MLE. They show that the same canonical set of relativiser forms in English (Ø, that, who) are used in both varieties, but only in the latter variety has a new topic-marking function developed for who (e.g., my medium brother who moved to Antigua, with high topic persistence of that referent in the following discourse). They note that ‘present-day English does not have a specific topic-marking feature, using instead a range of non-canonical syntactic structures such as presentational existential clauses and left-dislocated “presentational” constructions to introduce both a new subject and a new topic into the discourse’ (Reference Cheshire, Adger and Fox2013: 68). By contrast, many of the languages that are in the linguistic ecology of high-contact MLE speakers (e.g., Jamaican Creole, Igbo, Yoruba, Moroccan Arabic, Sylheti, Bengali, Twi, Spanish, and Maltese) have overt topicalisers and focus devices. Cheshire et al. (Reference Cheshire, Adger and Fox2013) argue that the novel development in MLE is in line with Matras’s (Reference Matras2009) observation that, in contact situations, the features most susceptible to transfer are those that convey the speaker’s monitoring and directing of the interaction, due to their routinisation in usage. They conclude that ‘topic marking, then, can be considered to be one of the information structuring strategies that drive innovation in situations of extreme linguistic diversity’ (Cheshire et al. Reference Cheshire, Adger and Fox2013: 67). Their study offers a vivid example of existing (non-)canonical variation in English giving rise to a more elaborated non-canonical system. (It is worth noting that Mauranen et al. Reference Mauranen, Carey, Ranta, Biber and Reppen2015 also observe systematic innovation and change in the relativiser system in ELF, supporting Matras’s speculation that certain parts of the syntactic structure are more susceptible to change in contact.)

Inherent variability in English as a source extends to further subtypes of sources, such as change arising from distributional biases in the input (see van Rooy Reference van Rooy2008 for an example from Black South African English) and universal principles of markedness that are subtly present in native standard varieties and that may get amplified in a non-native variety. For instance, the accessibility hierarchy (Keenan & Comrie Reference Keenan and Comrie1977) affects syntactic rules in both native and non-native varieties.

11.3.3 Pragmatic Innovation in the Usage Environment

Cheshire et al.’s (Reference Cheshire, Adger and Fox2013) example also relates to a further source of novel non-canonical syntax, namely innovation that arises out of particular functions or pragmatic needs in a given group. The social situations of European multiethnolects such as MLE are an intriguing environment to observe these processes at work.

These varieties have developed in major European cities after late twentieth-century migration. Examples include Multicultural London English (London), Kiezdeutsch (Berlin), Citétaal (Belgium), Rinkebysvenska (Stockholm), Straattaal (Netherlands), and further varieties in Copenhagen, Oslo, Helsinki, and other cities (Cheshire et al. Reference Cheshire, Nortier and Adger2015). The varieties share a number of defining features: they originated through late modern migration into working-class areas of European cities, they developed partly out of group second language acquisition in a social mix of non-native and near-native speakers in high-migration zones, they involve rapid change at many levels of the language, and they have been quickly absorbed into native usage by the adolescent population, shifting from learner to native speaker identity.

A number of grammatical innovations in these varieties originate in the pragmatic context of new usage by teenagers of recent migrant heritage background and mixed nativeness, in intensive social interactions that closely resemble the ‘com-sits’ described by Wiese (Reference Wiese2023) earlier. One example of an MLE innovation is a novel causal interrogative construction with the format why … for? An example is Why they looking at me like that for?, a blend of Why … and What … for? One might argue that this is simply a learner error that has taken hold. However, the use of the form has been consistently observed in MLE to express purpose, rather than a general cause. Brookes et al. (Reference Brookes, Hall, Cheshire and Adger2017) show convincingly that the redundant use of for, given its implicit presence in why, has the effect of being a ‘reinforcing’ or ‘expressive’ feature. They observe that the ‘variant typically occurs in pragmatically-charged environments, such as those that involve a confrontation between the participants in the conversational exchange or in direct speech contexts that report an aggressive encounter’ (Reference Brookes, Hall, Cheshire and Adger2017: 6). Its absence in elliptical structures or in negated clauses reinforces this analysis.

It is reminiscent of non-canonical negation structures in Jespersen’s Cycle, whereby in French a redundant emphatic particle (ne … pas) developed out of the pressure for discourse informational clarity, starting out as non-canonical as the former negation device eroded, and later became the canonical negation (Dahl Reference Dahl1979). On analogy with those cases, MLE causal interrogatives could retain their narrow purpose function and register restriction, or they could grammaticalise into a general cause marker, losing their immediacy, emphasis, and stance/affect (confrontation) constraints. This sort of pragmatic bleaching with stabilisation and spread would resemble semantic bleaching found across grammaticalisation, such as the grammaticalisation of determiners out of more contextually constrained demonstratives (Hopper & Traugott Reference Hopper and Traugott2003).

Thus, a new non-canonical construction in MLE involves initial grounding in speaker stance and immediacy, but with the potential to spread and lose its pragmatically narrow and marked functions over time. In the long term, a process of vernacular renewal may continue to regenerate forms of this kind, especially in high-contact urban youth groups.

The influence of immediacy of communicative function on linguistic form and use has been described in a range of research (Bernstein Reference Bernstein1971; Koch & Oesterreicher Reference Koch and Oesterreicher1985; Leuckert & Buschfeld Reference Leuckert and Buschfeld2021). It arises under specific conditions of intensive, vernacular social interaction. If we think back to Asian Englishes, discussed earlier, we can argue that IndE did not historically involve a preponderance of such interactional contexts, because of very limited nativisation and use in young peer groups; this is different to SgE, which was used in informal trade and later nativised more rapidly, with more extensive nativising use within child and teenage peer groups. The two varieties have developed very distinct kinds of innovative, or non-canonical, usage.

12.2.1 Defining the Intro-it

As has been mentioned in the Introduction to this volume, there are two major approaches to non-canonicity that can be distinguished: the frequency-based approach and the theory-based approach. When it comes to the intro-it pattern, this represents a particularly interesting distinction. With regard to the former approach, its enablement of the adherence to common information-structural principles like those of end-weight and end-focus (Behaghel Reference Behaghel1909) make it the cognitively least demanding choice given that, in comparison with its syntactically canonical counterpart, the intro-it reaches a point of grammatical completeness before the notional subject is added. This fact benefits production, comprehension, and processing (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 677)Footnote ² and makes the use of the pattern both the more natural and the more frequent choice.

From a structural or theoretical perspective, the situation is a little more complex: structurally, the notion of non-canonicity adopted by the researchers contributing to the present volume, in a nutshell, entails a deviation from an S_n‑V-X structure, with ‘S_n’ referring to the notional – and thus semantically loaded – subject and ‘X’ operating as a placeholder for what is to follow to make a given construction grammatically complete (see the Introduction to this volume). Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 53), for example, suggest seven basic clause patterns of English, which entail ${\mathrm{S}}_{\mathrm{n}}‐\mathrm{V}$ to be followed by (1) zero, (2) O_d-n, (3) C_S, (4) O_i-n-O_d-n, (5) O_d-n-C_o, (6) A, or (7) O_d-n-A. Deviations from the default ${\mathrm{S}}_{\mathrm{n}}‐\mathrm{V}‐\mathrm{X}$ structure are, of course, not random choices, but are triggered by complex and versatile (information-)structural mechanisms, a circumstance which, as was mentioned above, becomes particularly apparent in case of the intro-it. The pattern is defined by Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999) as the use of ‘the dummy subject it … in the ordinary subject position, anticipating a finite or non-finite clause in extraposition’ (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 155), a definition that already alludes to the structural diversity of the pattern. As is illustrated below, the expletive it, which excludes time, weather, or distance it (Callies Reference Callies2009), may be followed by either a copular or a lexical verb, while, depending on the verbal constituent and what is to be expressed, the postverbal element may take the form of an adjective phrase (AdjP), a noun phrase (NP), or a prepositional phrase (PP), which is then succeeded by the notional subject clause constituting either a that-clause, a to-infinitive-clause, an -ing-clause, a (w)h-clause (e.g., with whether or how), or an if-clause.

However, even if it is the most frequent choice, the role of it as a proxy is not limited to the subject of a construction but may affect the object as well (Kaltenböck Reference Kaltenböck2004: 65; Gentens Reference Gentens2016). As illustrated below, it may, in fact, be used to fill grammatically vacant positions in almost all of the seven basic clause patterns previously discussed, as illustrated in Table 12.1 (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1391–3).

Naturally, the frequencies with which the non-canonical structures are employed vary. While extraposed that-clauses, for example, are generally named as the most frequent realisation (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999), extraposed -ing-clauses are most commonly associated with rather informal spoken discourse (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1393).

Table 12.1Realisations of the intro-it across the basic clause patterns of the English language

Note. _g refers to grammatical constituents while _n refers to notional (or semantically loaded) ones.

Source: adapted from Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1391–3).

Table 12.1Long description

The table displays different clause structures for the usage of "intro-it" in English. It includes three columns: "Canonical Variant," "Non-canonical Variant," and "Example."

• Row 1: The canonical variant is S-V-zero, and the non-canonical variant is Sg-V-zero-Sn. Example: "It doesn’t matter whether you like it."

• Row 1 (continued): The canonical variant is S-V-zero, and the non-canonical variant is Sg-V-pass-zero-Sn. Example: "It is claimed that she does not know him."

• Row 2: The canonical variant is S-V-Od, and the non-canonical variant is Sg-V-Od-Sn. Example: "It surprised me to see you."

• Row 3: The canonical variant is S-V-Cs, and the non-canonical variant is Sg-V-Cs-Sn. Example: "It is nice to see you."

• Row 5: The canonical variant is S-V-Od-Co, and the non-canonical variant is S-V-Od-g-Od-g-Co-Od-n. Example: "He must consider it interesting working here."

• Row 6: The canonical variant is S-V-A, and the non-canonical variant is Sg-V-A-Sn. Example: "It was on the news that he is guilty."

• Row 7: The canonical variant is S-V-Od-A, and the non-canonical variant is S-V-Od-g-A-Od-n. Example: "She put it into his head that he should resign."

Each row provides a sentence example illustrating different syntactic forms using intro-it across basic clause patterns.

In addition to its formal diversity, the pattern, very generally speaking, also enables the evaluation or assessment of discourse content (Couper-Kuhlen & Thompson Reference Couper-Kuhlen and Thompson2008). It is associated with particular semantic or rhetorical domains which can be derived from the nature of the verb and/or verb-complement structure, which, in a way, ‘set(s) the scene’ for the subsequent clausal content. These include the expression of necessity, importance, ease, or difficulty by means of particular lexical bundles containing, most commonly, combinations of four (e.g., it is clear that, it is possible to, it is difficult to), five (e.g., it may be possible to, it is interesting to note), or even six words (e.g., it should not be forgotten that, it must have been difficult to) (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999). Many previous studies focus on the classification of verb and adjective phrases. Examples include Römer (Reference Römer2009) or Hewings and Hewings (Reference Hewings and Hewings2002), with the latter suggesting a fourfold distinction between hedging (1), emphasising (2), attributive (3), and attitude marking functions (4).

A more comprehensive approach was adopted by Larsson (Reference Larsson2017), who, unlike Hewings and Hewings (Reference Hewings and Hewings2002), whose focus was mainly on AdjPs and passivised verb phrases, also took postverbal NPs into account. Another advantage of her approach is the fact that, while her classification, too, includes attitude markers, hedges, and emphatics (as well as combinations thereof), the latter two functions are only assigned if there is linguistic ‘evidence’, such as amplifiers preceding the adjective (e.g., extremely surprising, very disappointing) as an indicator of an emphatic use or the use of seem or a modal verb of possibility (e.g., could, might) as indicators of hedging, making the analysis less subjective and thus more reliable. Her final category, observations, is reserved for neutral propositional content which, prior to her study, had been mostly disregarded.

Now, what unites research on the semantic or rhetorical functions of the sentence pattern is the fact that there appear to be population- as well as register-specific preferences (see also Zhang Reference Zhang2015). It thus remains for us to determine whether this also holds true for ESL in general and newspaper language in particular.

12.2.2 Theories and Previous Research

To the best of our knowledge, the first researcher to discuss extraposition in his work is Jespersen (Reference Jespersen1927), claiming that ‘words in extraposition may be added as a kind of afterthought after the sentence has been completed; they stand outside it and form, as it were, a separate utterance which might even be called a separate sentence’ (Reference Jespersen1927: §17.12). Now, even though this first perspective seems to extend to structures like dislocations as well (e.g., He is nice, your brother.) and thus encompasses a broader range of phenomena, it-extrapositioning has still received quite a lot of scholarly attention ever since. Research has been dedicated to the description of the structural and functional employments as well as to the discussion of the information-structural mechanisms that may be at play. With the postponement of clausal subjects or objects to the end of a construction while it occupies the grammatical position, the pattern, for example, operates in accordance with the principle of end-weight (Behaghel Reference Behaghel1909). The concept associates the placement of syntactically weighty elements towards the end of a sentence with reduced cognitive efforts for both the production and the comprehension of an utterance (see, e.g., Arnold et al. Reference Arnold, Losongco, Wasow and Ginstrom2000). The applicability of the concept to the intro-it has been confirmed by a number of researchers, including Huddleston and Pullum (Reference Huddleston and Pullum2002), Gómez-González (Reference Gómez-González1997), and Kaltenböck (Reference Kaltenböck2004). Simultaneously, the structure also serves the adherence to the given-new progression (e.g., Clark & Haviland Reference Clark, Haviland and Freedle1977; Prince Reference Prince and Cole1981) and the principle of end-focus (Behaghel Reference Behaghel1909: 138; Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1356–61; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 897–9). Both concepts are strongly connected to the efficiency of information packaging. Given that the dummy it merely fulfils a grammatical function and does not contribute any semantic value to the message that is being produced or perceived, attention is placed on the postponed clausal element, which is commonly associated with new and thus focused information. While the study by Herriman (Reference Herriman2000) supports the view that the structure enforces the principle of end-focus, Kaltenböck (Reference Kaltenböck2004) maintains that this is, indeed, strongly connected to the given-new status. He finds that the majority of instances with discourse-new clausal subjects (roughly 72%) were used in an intro-it construction. This is supported by Huddleston and Pullum (Reference Huddleston and Pullum2002), as well as Miller (Reference Miller2001), who describe discourse-new clausal subjects in non-canonical position as more felicitous or even mandatory.

As far as other language contexts are concerned, quite some research has been devoted to EFL learners already. While the accounts report a general overrepresentation of the phenomenon, it has been found that the structural and functional realisations may vary. One example includes the preference for extraposed to-clauses by some learner populations, including German and Swedish learners of English (Callies Reference Callies2009; Larsson Reference Larsson2016). Likewise, Hewings and Hewings (Reference Hewings and Hewings2002) and Larsson (Reference Larsson2017) have determined that learners tend to use the structure to increase the force of a claim rather than for hedging purposes, with the employment of unusually forceful adjectives, such as amazing or stupid, resulting in non-target-like uses by some learners (Römer Reference Römer2009).

When it comes to the ESL context, non-canonical patterns, and among these fronting and left dislocation in particular, have been described as common features of ‘New Englishes’ (e.g., Alsagoff & Lick Reference Alsagoff, Lick, Foley, Kandiah, Zhiming, Gupta, Alsagoff, Lick, Wee, Talib and Bokhorst-Heng1998; Sharma Reference Sharma, Butt and King2003, Reference Sharma and Hickey2012; Bhatt Reference Bhatt and Mesthrie2008; Mesthrie & Bhatt Reference Mesthrie and Bhatt2008), with a number of corpus-based studies providing detailed accounts of a variety of syntactic structures (Lange Reference Lange2012; Winkle Reference Winkle2015; Leuckert Reference Leuckert2019). One study, Götz (Reference Götz2017), investigates fronting in the newspaper language of the same varieties considered in the present study and finds interesting variety-specific preferences, thus proving that fronting is by no means limited to informal spoken data – as suggested by previous research. As to research into the intro-it – or postponement patterns in general – only a very limited number of studies have been conducted. One of the few exceptions is Dubey’s (Reference Dubey1989) monograph on Newspaper English in India, in which he briefly mentions that, in comparison with the movement of participles or parts of NPs, the cases of postponements employed in newspapers in India were almost exclusively limited to clausal subjects, presumably to achieve end-focus (Reference Dubey1989: 77).

It remains to be seen in what ways this attestation of a preference for extraposed clausal structures can be extended, both on a structural and a functional basis. The precise nature of the research gaps identified and our research questions will be discussed in the following section.

12.3.1 Research Gaps and Research Questions

In the previous description of the status quo of research into the topic, we have identified two major research gaps which the present chapter intends to fill. To the best of our knowledge, no detailed account of the intro-it has been provided in the South Asian English context. Therefore, we pose the following research questions:

1. (1) How frequent is the intro-it in South Asian Varieties of English (SAVEs) (compared to BrE)?

2. (2) Are there different preferences between the individual SAVEs and BrE in using the intro-it regarding its

   1. (a) structural realisations?

   2. (b) semantic functions?

12.3.2 Database and Methodology

In order to avoid overinterpretations of (spontaneous) speech and because we are mainly interested in nativised forms of Englishes, we investigated a corpus of acrolectal newspaper language that had been subjected to an editing process. In doing so, we are able to ascertain that all instances of the intro-it we found are deliberate uses. The database used for the present chapter is the South Asian Varieties of English Corpus (SAVE Corpus; Bernaisch et al. Reference Bernaisch, Koch, Mukherjee and Schilk2011) which, based on two sources each, contains the newspaper language of six ESL varieties in South Asia. It was compiled between 2008 and 2011 and contains approximately 3 million words per subcorpus (i.e., about 18 million words in total). Details on the papers used as well as the time spans considered can be found in Table 12.2.

Table 12.2The individual subcorpora of the SAVE Corpus

Source: Bernaisch et al. Reference Bernaisch, Koch, Mukherjee and Schilk2011: 2; reprinted by permission from Tobias Bernaisch.

Table 12.2Long description

The table lists newspapers from various South Asian countries, along with their respective U R Ls and the time span of their publication. The table is divided into 4 columns, where the headers are the aforementioned. Each country has two papers along with the relevant data. The rows are filled from left to right as follows:

1. 1. For Bangladesh:

   • The corresponding data are Daily Star, New Age, www.thedailystar.net, ww.newagebd.com, 2003 to 2006, and 2005 to2007.

2. 2. For India:

   • The corresponding data are The Statesman, The Times of India, www.thestatesman.net, http://timesofindia.indiatimes.com, 2002 to 2005, and 2002 to 2005.

3. 3. For Maldives:

   • The corresponding data are Dhivehi Observer, Minivan News, www.dhivehiobserver.com, www.minivannews.com, 2004 to 2007/2008, and 2004 to 2008.

4. 4. For Nepal:

   • The corresponding data are Nepali Times, The Himalayan Times, www.nepalitimes.com, www.thehimalayantimes.com, 2000 to 2007/2000, and 2002 to 2008.

5. 5. For Pakistan:

   • The corresponding data are Daily Times, Dawn, www.dailytimes.com.pk, www.dawn.com, 2002 to 2006, and 2002 to 2007.

6. 6. For Sri Lanka:

   • The corresponding data are Daily Mirror, Daily News, www.dailymirror.lk, www.dailynews.lk, 2002 to 2007, and 2001 to 2005.

With BrE as the historical input variety of the varieties under scrutiny, the news section of the British National Corpus (BNC) was employed as a control corpus. The first version of the BNC (i.e., BNC 1994) was compiled between 1991 and 1994 and has a size of about 100 million words, 9 million of which are part of the periodicals section used as part of our database.

We employed a modified random sampling technique to extract the articles to be analysed, making a point to attempt to have each file of the corpus (each of which contains a varying number of articles) feature in the analysis. Files and/or articles were disregarded, however, if they were too short, did not contain neutral newspaper language (e.g., birthday greetings, advertisements), or were used repeatedly in the corpus.

As Table 12.3 illustrates, roughly 8,000 sentences per subcorpus were extracted for the manual annotation process. They were copied into an Excel spreadsheet before identifying the instances of the intro-it, which were then annotated for a number of structural and functional variables, which can be found in Table 12.4.

Table 12.3Overview of analysed sentences per corpus (number of intro-its per thousand sentences/pts)

Table 12.3Long description

The table is sectioned into 5 columns. The columns are labeled as: Country, Corpus, Number of sentences, total, Number of intro-its, total, Number of intro-its, p t s. There are 7 rows for different countries and an eight row for total. The data in the rows are filled from left to right as follows:

• For Great Britain, the corresponding data are B N C underscore News, 8055, 143, and 17.75.

• For India, the corresponding data are S A V E underscore I N, 8289, 165, and 19.91.

• For Bangladesh, the corresponding data are S A V E underscore B D, 8034, 123, and 15.31.

• For Maldives, the corresponding data are S A V E underscore M V, 7951, 280, and 35.22.

• For Sri Lanka, the corresponding data are S A V E underscore S L, 8130, 256, and 31.49.

• For Nepal, the corresponding data are S A V E underscore N P, 8245, 123, and 14.92.

• For Pakistan, the corresponding data are S A V E underscore P K, 8098, 185, and 22.85.

• For total, the corresponding data are 56802, 1275, and 22.45.

Table 12.4Overview of annotated variables

Table 12.4Long description

The table is sectioned into 2 columns. The columns are labeled as: variables and meaning. There are 14 rows for different variables. The data in the rows are filled from left to right as follows:

• The data for S E M A N T I C F U N C T I O N is semantic function based on verb or verb-complement structure.

• The data for S underscore length is length of the sentence of interest in words.

• The data for I N T R O I T underscore T F is indication of whether the given sentence contains an intro-it.

• The data for C O P U L A R underscore T F is indication of whether the verb used is a copular or not.

• The data for P A S S I V E underscore T F is indication of whether the verb is used in the passive or not.

• The data for T H A T underscore C L is indication of whether the postponed constituent is a that-clause.

• The data for Z E R O underscore T H A T underscore C L is indication of whether the postponed constituent is a zero that clause.

• The data for T O underscore I N F underscore C L is indication of whether the postponed constituent is a to-inf clause.

• The data for I N G underscore C L is indication of whether the postponed constituent is an -ing clause.

• The data for I F underscore C L is indication of whether the postponed constituent is an if-clause.

• The data for W H underscore C L is indication of whether the postponed constituent is a wh-clause.

• The data for H underscore C L is indication of whether the postponed constituent is an h-clause.

• The data for I N T R O I T underscore S is indication of whether the intro-it affects the subject.

• The data for I N T R O I T underscore O is indication of whether the intro-it affects the object.

Except for the s_length variable, which was determined using an Excel formula, each of the variables was coded manually by the authors of this chapter. While the majority are based on structural observations, the classification of the semantic function calls for some additional explanatory remarks. Due to the previously outlined advantages of her approach, we adopted Larsson’s (Reference Larsson2017) general distinction between attitude markers, expressing ‘the writer’s affective attitude towards what is stated in the clausal subject’, emphatics, as a means to ‘strengthen the force of the utterance’, hedges, used to weaken the forcefulness of an utterance and to indicate a lack of full commitment, and observations, which may express ‘affectively neutral … propositional content’ (Larsson Reference Larsson2017: 61).

Note, however, that we did not code for combinations of the former three categories because, after close consideration of the data at hand, we came to the realisation that particular categories override others. As the examples below illustrate, both emphatics (5) and hedges (6) exert a domineering influence on the perception of the meaning expressed by means of the attitude markers.Footnote ³

In case of a combination of all three (7), the instance was considered a hedge as well, given that, irrespective of the strong terms in which a claim might be expressed, as soon as it is preceded by a hedge, its force is still weakened.

Finally, given that our database is comprised of newspaper language, generally characterised by neutrality and formality, we considered it reasonable to slightly extend the observational category. We propose a distinction between observation-neutral, observation-attitude, observation-hedging, observation-emphatics, and observation-reporting. While the former category comprises factual information without any kind of assessment (e.g., It is the government’s responsibility that the laws are enforced.), observation-attitude, observation-emphatics, and observation-hedging – as opposed to the simple attitude, emphatics, and hedging categories – are not concerned with the attitude of the author of the article (or with making their statement more or less forceful) but with that of, for example, the people they are writing about (e.g., It is the manager’s declared belief that he will win the election [attitude]; It has been criticised immensely that the minister did not attend the event. [emphatics]; The chairman maintained it seemed reasonable to take action. [hedging]). Finally, observation-reporting predominantly contains verbs that are employed in passivised verb structures and which are commonly used in the news sector to report events neutrally (e.g., It is reported that fifty people were wounded during the attack.)

Following the annotation process, all analyses were conducted using the software package R Studio. In order to attend to research question (1), the data were submitted to a regression modelling process using the R package dplyr that predicts the probability for an intro-it to occur based on the independent variables. To answer research questions (2a) and (2b), we zoomed in on the formal, functional, and semantic contexts of the intro-it by applying conditional inference trees (ctree) to our dataset (Hothorn et al. Reference Hothorn, Hornik and Zeileis2006). In a nutshell, ctree analysis recursively performs univariate splits of the dependent variable based on values of an independent variable or a set of covariates. As pointed out by Timofeev (Reference Timofeev2004) and Phelps and Merkle (Reference Phelps, Merkle, Eichhorn and Kovar2008), adopting such a classification tree approach presents a number of advantages that are relevant for our dataset: (i) it can handle small datasets, (ii) as a nonparametric statistical approach, this method only comes with a limited number of statistical assumptions, (iii) it can be applied to various data structures and is exceptionally efficient in handling categorical predictors, (iv) the output provides rankings of variable importance, and (v) the output of a classification tree analysis can easily be understood and interpreted. Please note that, in order to be in a position to reveal the effects of the individual variables on the complementation patterns of the intro-it and to thus provide a comprehensive account of the behaviour of these patterns in the individual corpora, we ran individual trees for each variable. The results of the analysis are presented and discussed in the following.

12.4.1 The Likelihood of Using the Intro-it across SAVEs

When predicting the likelihood of using an intro-it to occur across SAVEs and BrE, we fitted a generalised linear model and took S_Length (logged) and Corpus as predictors. The interaction between the two variables turned out not to be significant, which means that the final model consisted only of the two main predictors S_Length (logged) and Corpus. The model is highly significant and explains 4.4% of the variance in the data (R² = 0.044). Although this might not seem to be a high value, it performed significantly better than the baseline model, and visual inspection of the residual plots suggests that the model performed decently. The two main effects are illustrated in Figure 12.1.

Figure 12.1 Effect plots illustrating the main effects Corpus (panel 1) and S_Length (logged) (panel 2) predicting the likelihood for an intro-it to occur in the SAVE Corpus and the BNC_NEWS

Figure 12.1Long description

Plot a: corpus effect plot

This line graph is titled corpus effect plot and illustrates how the variable corpus affects the dependent measure I N T R O I T T F. The graph is designed to show categorical differences across various corpora.

• The vertical axis is labeled I N T R O I T T F, measuring relative frequency or weight of the I N T R O I T T F function. It ranges from 0.000 to 0.030, with tick marks at regular intervals.

• The horizontal axis is labeled corpus and displays seven discrete corpus labels:

  B D, B N C, I N, M V, N P, P K, and S L. Each corpus label corresponds to a point on the graph, indicating the value of I N T R O I T T F, for that dataset. The points are connected by a continuous line to help visualize trends across corpora. Each point also has a vertical error bar, reflecting the confidence interval or variability for that data point.

The pattern of the graph is as follows:

• It begins at a low value at B D, possibly the baseline corpus.

• The line remains relatively flat through B N C and I N, suggesting minimal change in I N T R O I T T F.

• A sharp peak occurs at M V, where I N T R O I T T F reaches its highest value across all corpora.

• The value then drops at N P, indicating a decrease in frequency.

• A slight rise follows at S L, though it does not reach the height seen at M V.

Overall, the plot shows non-linear variation in I N T R O I T T F across corpora, with some datasets contributing significantly more than others. The presence of error bars helps interpret the reliability of these observed differences.

Plot b: S length effect plot

This plot is titled S length effect plot, showing the relationship between S length likely sentence length or a sentence-length-related measure and the outcome variable I N T R O I T T F.

• The vertical axis is labeled I N T R O I T T F, ranging from 0.00 to 0.10 in increments of 0.02.

• The horizontal axis is labeled S length and spans from negative 2 to 4, suggesting that this variable is standardized e.g., z scored for comparability.

A single smooth curve is plotted on the graph to depict the effect of S length on I N T R O I T T F.

• The curve starts near the bottom-left corner, with values close to 0.00 when S length is around negative 2, indicating that shorter sentences are associated with very low I N T R O I T T F values.

• The line then gradually increases, becoming steeper as S length moves toward the upper end of the scale.

• At around S length equals 4, the curve reaches close to 0.10, indicating that longer sentences are predicted to have higher I N T R O I T T F values.

Error bars or confidence intervals are visible along the line, growing wider at the extremes of the S length range, which indicates greater uncertainty in predictions for very short or very long sentences.

A rug plot is shown at the base of the graph along the x-axis, consisting of small vertical tick marks that indicate the density or distribution of actual data points. Most of these ticks cluster between negative 1 and 2, suggesting that most sentences in the data fall within this range of length.

The overall trend suggests a positive, nonlinear relationship: as sentence length increases, so does I N T R O I T T F, with more substantial increases for longer sentences.

It becomes immediately obvious that the use of the intro-it is not generally more likely to occur across SAVEs than in BrE. In fact, the likelihood of using this pattern is not significantly different from BrE in BgE, IndE, NpE, or PkE. However, we do see that the pattern is significantly more likely to occur in Sri Lanka and on the Maldives.

The second main predictor that increases the likelihood of using the intro-it is S_Length (logged). Since this effect does not engage in an interaction with Corpus, it seems that sentence length has a robust effect across varieties in the sense that the longer a sentence is, the more likely it is to make use of an intro-it. While an increased sentence length obviously generally increases the chance for non-canonical patterns to occur (e.g., cf. Götz Reference Götz2017; cf. Kircili, Chapter 13 in this volume; Kircili, Reference Kirciliforthcoming), it also makes sense from an information-packaging standpoint to make use of this pattern (e.g., when there are very long subjects; see Section 12.2.2). Here, the intro-it enables the writer to finalise the main structure of a sentence (i.e., S_g-V-C_s-S_n) and thus to be in a position to pack the main information (contained in the notional subject) into a structure that is in line with the principle of end-weight and thus more easily accessible to and processed by the reader, as in example (8) from BgE.

While this does not seem to be a particularly surprising finding, we can note that the cognitive effect of easing the processing load for the reader (and writer), which has been attested for other weight-sensitive structures, including the placement of subordinate clauses, and which has mainly been researched in the context of ENL varieties of English (e.g., Arnold et al. Reference Arnold, Losongco, Wasow and Ginstrom2000; Diessel Reference Diessel2005; Junge et al. Reference Junge, Theakston and Lieven2015), also seems to hold true for the intro-it and SAVEs.

12.4.2 Comparing the Use of the Intro-it across SAVEs

We will now turn to our second research question and zoom in on the cases in which intro-its were used across SAVEs by taking a look at their overall frequencies, their formal complementation patterns, and the semantic functions in which the construction is used.

Frequency of the Intro-it across SAVEs

Figure 12.2 illustrates the frequencies of the intro-it per thousand sentences (pts) across SAVEs as compared to the BNC_NEWS.Footnote ⁴

Figure 12.2 Normalised frequencies of intro-its per thousand sentences across SAVEs and the BNC_NEWS

Figure 12.2Long description

The horizontal axis labeled I N T R O I T P T S, or frequency of intro-its per thousand sentences, ranges from 0 to 35, in increments of 5. The horizontal axis labeled C O R P U S marks different labels of S A V E or South Asian variety of English corpus and B N C or British national corpus. The labels include B N C-N E W S, S A V E-B D, S A V E-I N, S A V E-M V, S A V E-N P, S A V E-P K, and S A V E-S L, and the bars associated with each corpus vary in shades. The frequency is more for S A V E-M V. All others are around the range of around 18 to 30.

As Figure 12.2 illustrates, we can observe considerable variation in the frequencies of the intro-it between the varieties, which is highly significant (χ² = 132.86, df = 6, p < 2.2e-16). The highest frequencies are found in the newspaper language of the Maldives with 35.22 pts and Sri Lanka with 31.49 pts, followed by Pakistan with 22.85 pts, India with 19.91 pts, and Great Britain with 17.75 pts. The lowest frequencies are found in Bangladesh with 15.31 pts and Nepal with 14.92 pts. These frequencies are very much in line with the predicted probabilities of the pattern to occur (see Section 12.4.1). Compared to BrE, the historical input variety, we can observe that it is Bangladesh and Nepal that show similar – although slightly lower – frequencies, while India and Pakistan make use of the pattern somewhat more frequently. Yet, it is only writers from Sri Lanka and the Maldives who use the pattern significantly more frequently than the BrE ones.

Structural Preferences of the Intro-it across SAVEs and British English

Turning to the structural preferences across SAVEs, we will first focus on complementation patterns. In doing so, we look at the different clause types that the intro-it may occur in. The findings are illustrated in Figure 12.3.

Figure 12.3 Ctree illustrating the clause types used to complement the intro-it ~ Corpus + Passive in the SAVE Corpus and the BNC_NEWS

Note. TH = that-clauses, ZT = zero that-clauses, TO = to-infinitive clauses, IN = infinitive clauses, IF = if-clauses, WH = which-clauses, HO = how-clauses.

Figure 12.3Long description

Decision tree diagram analyzes clause types used to complement the intro-it, with particular focus on how they are influenced by the variable passive T F. At the top of the tree, there is a node labeled passive T F with a significance level of p less than 0.001, indicating a strong relationship between this variable and the outcome. The tree splits into two branches based on a numerical threshold.

Left Branch:

First corpus Node: This node is labeled corpus with a significance of p = 0.002, indicating a significant effect of the corpus on clause types.

B N C Node: This leads to a node labeled B N C with n = 115, meaning there are 115 observations from the B N C corpus. The corresponding bar graph here shows higher frequencies of clause types T H and T O.

B D, I N, M V, N P, P K, S L Node: This node represents other corpora, aggregating data from B D, I N, M V, N P, P K, and S L, with n = 825. The graph for this node shows elevated values for Z T and T O, suggesting that these clause types are more frequent in these corpora compared to T H and T O.

Right Branch:

Second corpus Node: This right branch leads to another corpus node, where the significance is p = 0.024. This is still statistically significant but weaker than the first corpus node.

B D, I N, M V, P K, S L Node: This node, with n = 288, displays a clear preference for the clause type T H, which has the highest bar in the graph, showing that these corpora use T H more frequently than others.

B N C, N P Node: This node includes data from B N C and N P corpora, with n = 44. The bar graph for this node shows elevated frequencies for both T H and Z T, indicating a higher usage of these clause types in the B N C and N P corpora.

In conclusion, the decision tree analyzes how different corpora affect the choice of clause types used with intro-it. The nodes highlight significant variations between corpora and show clear preferences for certain clause types within those corpora, with T H and T O being more prominent in some cases, and Z T and T O being more frequent in others.

The ctree illustrates that the most important and significant factor triggering particular complementation patterns is voice of the verb across all varieties (see node 1; p < 0.001). When we follow the left branch of the tree (i.e., the cases where there is a main verb that is not used in the Passive), node 2 illustrates significantly different complementation patterns (p = 0.003) between the BNC on the one hand (see final node 3) and the SAVE data on the other, which all cluster together (see final node 4). Albeit significantly different, the overall distribution patterns follow the same order regarding the three most frequent complementation patterns: the most frequent complementation pattern for active constructions in newspaper language is to-infinitives, with the SAVEs making use of this construction in almost 60% of the cases, compared to ca. 45% in the BNC. These are illustrated by example (9) from BrE and example (10) from PkE.

(9)

Now [it]^Sg [is]^V [up to the aid agencies]^Cs [to make life as comfortable and easy as possible for the refugees.]^Sn

(BNC_NEWS CBC)

(10)

[It]^Sg [is]^V [difficult]^Cs [to quantify the gains and losses],^Sn but unsuspecting investors ought to have lost heavily.

(PK_DA_2007–06-24)

That-clauses figure as the second most frequent complementation pattern (cf., however, Section 12.2.2), with ca. 40% of all cases in BrE and ca. 36% in the SAVEs, followed by zero-that clauses, which BrE newspapers make use of in ca. 6% of the cases and SAVE newspapers in ca. 4% of the cases. There is another noteworthy finding among the low-frequency complementation patterns, namely that, in BrE, intro-it in active constructions is complemented using how-clauses (see (11)), which is something that we do not find in SAVEs at all, which, in turn, complement the intro-it using wh-clauses, as in (12), which is not observed in the British data at all.

(11)

But [it]^Sg [has been]^V [a mystery]^Cs [how this and spongiform encephalopathies, such as mad cow, kill brain cells.]^Sn

(BNC_NEWS AJS)

(12)

No shots were fired during the raid and [it]^Sg [is not known]^V [where the men are being hauled up.]^Sn

(PK_DT_2002–07-19)

With verbs in the Passive voice, if we follow the tree to the right, there is a significant split between the varieties again (see node 5, p = 0.028), and this time Nepal clusters together with Great Britain (see final node 7), whereas the other SAVEs form a cluster again (see final node 6). While both groups use that-clauses most frequently, the former generally shows more variety by also using different complementation patterns than the SAVEs, which almost exclusively use that-clauses, namely in over 80% of the cases (see (13)). This is followed by to-infinitive-clauses (ca. 8%, see (14)), and zero-that clauses (ca. 4%, see (15)), and very low proportions of the other complementation types (each under 2%). While British and Nepalese English speakers also complement the introductory-it pattern most frequently by that-clauses (ca. 65%, see (16)), they also show a high proportion of zero-that clauses (over 20%, see (17)), and also make use of how-clauses (ca. 5%, see (18)). The remaining patterns also only occur with very low frequencies, albeit still higher than in the SAVEs.

(13)

[It]^Sg [was announced]^V [that the party would look after the victim’s children.]^Sn

(IN_SM_2002–09-02)

(14)

During the Oslo meeting on June 26, [it]^Sg [was agreed]^V [to put pressure on Sri Lanka in this regards as well.]^Sn

(LK_DM_2007–07-23)

(15)

[It]^Sg [was reported]^V [ø he may have been manhandled.]^Sn

(MV_MN_2005–05-03)

(16)

[It]^Sg [was agreed]^V [that the wife was the chatelaine par excellence.]^Sn

(BNC_NEWS A59)

(17)

[ø It should be able to accumulate enough water by 2050,]^Sn [it]^Sg [is hoped.]^V

(NP_NT_2006–06-30)

(18)

[It]^Sg [remains to be seen]^V [how much Biggs has got left and even Duff has made mistakes.]^Sn

(BNC_NEWS A2E)

In addition to the verb’s voice, the Verbtype and whether a subject or an object is referred to by the intro-it have an effect on the complementation type. This is illustrated in Figure 12.4.

Figure 12.4 Ctree illustrating the clause types used to complement the intro-it ~ Corpus + Verbtype + Function in the SAVE Corpus and the BNC_NEWS

Figure 12.4Long description

A diagram illustrates a conditional inference tree used to analyze clause types in relation to verb type and corpus. The diagram consists of nodes connected by lines, and it presents how these nodes break down into different categories based on probabilities.

Top Node:

• Verb type node: The conditional inference tree begins with a top node labeled verb type, which contains a probability value that indicates the likelihood of a certain verb type influencing the outcomes in subsequent nodes. This node is connected to two lower nodes, each labeled corpus with different associated probabilities. This suggests that the selection of a corpus or the type of corpus plays a significant role in determining the pattern of clause types used with different verb types.

Branching Nodes:

• First corpus node: The left branch leads to the first corpus node. This node further branches out into additional nodes.

• Second corpus node: The right branch leads to another corpus node, which connects to further nodes as well. These branching nodes break down the data further, leading to specific groupings of clauses or categories.

Bar Graphs:

At the bottom of the diagram, five bar graphs are presented, each corresponding to a different node with a specific sample size n.

• Node 3 where n = 27: The first bar graph is labeled node 3 where n = 27. The y-axis ranges from 0 to 1, representing the frequency of different clause types, and the x-axis labeled with various clause types: T H, Z T, T O, I N, I F, W H, H O. The bar graph displays varying heights of bars corresponding to each clause type, indicating their relative frequency within this node.

• Node 5 where n = 7: The second bar graph is labeled node 5 where n = 7. It follows the same y-axis and x-axis as Node 3, displaying the distribution of clause types across the data, with each bar corresponding to one of the clause types listed. The heights of the bars reflect how frequently each clause type occurs in this specific node, to-infinitive clauses being most frequent.

• Node 6 where n = 365: The third bar graph is labeled node 6 where n = 365. This graph also shows the frequency of the seven clause types for the second largest subset of the data with 365 observations. The bar heights indicate the frequency distribution of different clause types in this data segment, that-clauses being most frequent.

• Node 8 where n = 106: The fourth bar graph is labeled node 8 where n = 106. Like the previous graphs, it shows the frequency of clause types for a dataset with 106 observations. The bar heights show the distribution of different clause types in this segment of the data.

• Node 9 where n = 767: The fifth bar graph, labeled node 9 where n = 767, represents the largest subset of the data. The bar heights in this graph indicate how the various clause types are distributed in this final node. To-infinitive clauses are most frequent, followed by that-clauses.

T O has the highest value in all the nodes except node 6 where T H has the highest value.

The ctree in Figure 12.4 illustrates that, across all corpora, complementation patterns in intro-it sentences differ significantly depending on the Verbtype used (see node 1, p < 0.001). If a lexical verb is used, there is another significant split between the corpora (see node 2, p = 0.002), with NpE showing significantly different complementation patterns (see node 3) than the other SAVEs, however, this time plus the BrE speakers. While in NpE, for lexical verbs there is an almost even preference of complementing the intro-it with that-clauses, as in (19), or to-infinitive-clauses, as in (20), for the other varieties, there is another interesting split concerning the functional element that is referenced by the intro-it (see node 4, p = 0.04)): if the it-pattern refers to an object, the six varieties show a clear preference for selecting a to-clause to complement it in ca. 70% of the cases (see node 5), as illustrated in (21), whereas subjects are complemented by that-clauses in almost 80% of the cases (see node 6), see example (22). As far as copular verbs are concerned, the SAVEs behave significantly differently from BrE (see node 7, p = 0.003), with BrE speakers showing an almost even distribution between that- and to-clauses (see node 8), illustrated in (23) and (24), and SAVE speakers choosing to-clauses in over half of the cases (see node 9), as in example (25).

(19)

[It]^Sg [was carefully instilled]^V [in her]^A [that all girls grew old before they grew up.]^Sn

(NP_NT_2002–12-13)

(20)

[It]^Sg [took]^V [divers from the Gent Fire Brigade]^Od [two days]^A [to retrieve Prem Prasad’s body.]^Sn

(NP_NT_2004–06-04)

(21)

But a senior banker, Shaukat Tarin, told Dawn that [it]^Sg [was]^V [best]^Cs [to leave it to shareholders to assess the merits and demerits.]^Sn

(PK_DA_2006–07-27)

(22)

[It]^Sg [saddens]^V [me]^Od [that educated people in the government could do such things.]^Sn

(MV_DO_011)

(23)

[It]^Sg [has become]^V [increasingly apparent]^Cs [that Thomson belonged to the American experimental tradition.]^Sn

(BNC_NEWS_A1H)

(24)

[It]^Sg [was]^V [very difficult]^Cs [for a court to say what a reasonable mother would have done in circumstances which were almost hypothetical.]^Sn

(BNC_NEWS_A30)

(25)

But [it]^Sg [will be]^V [difficult]^Cs [to say that the dead has been mourned in a befitting way by the nation itself.]^Sn

(BD_DS_2003–07__pt17)

After having looked in detail at the different structural preferences across the SAVEs and BrE, we will now turn to the semantic functions that are associated with the intro-it.

12.4.3 Semantic Function of the Intro-it across SAVEs and the BNC_NEWS

We will now turn to the semantic functions in which the intro-it is used. The ctree in Figure 12.5 illustrates a number of variety-specific preferences. After the first significant split between the British, Nepalese, and Maldivian varieties on the one hand, and BgE, IndE, PkE, and SLE on the other hand (see node 1, p < 0.001), if we first follow the branch to the left, the most noteworthy difference from the latter group is that the former group shows the highest proportions of the pattern used in an attitudinal function. Further down the tree, there is another significant split that clusters the British and the Nepalese speakers together (see final node 3) and separates them from the Maldivian speakers (see final node 4). British and Nepalese speakers show a clear preference for using the intro-it in an attitudinal function in ca. 36% of the cases, see (25), followed by an almost equal proportion of observation-neutral and observation-attitudinal cases with ca. 14% each (see final node 3), as illustrated in (26) and (27), respectively. In comparison, apart from a similarly high proportion of ca. 30% of all cases in the attitudinal function, MvE (see final node 4) also shows a relatively high proportion of intro-its that are used either in an observation-reporting function (see example (28)) or observation-neutrally (as in (29)).

Figure 12.5 Ctree illustrating the semantic functions in which the intro-it is used by the different SAVEs vs. the BNC_NEWS

Note. A = attitude, E = emphatic, H = hedging, OA = observation-attitudinal, OE = observation-emphatics, OH = observation-hedging, ON = observation-neutral, OR = observation-reporting

Figure 12.5Long description

The tree diagram analyzes the semantic functions of the introductory it across different corpora. The diagram consists of a hierarchical structure of nodes connected by lines, illustrating the decision-making process and categorization of data. Here’s a detailed description of the diagram:

Top Node:

• Corpus node: At the top of the tree structure is a node labeled corpus, which contains various statistical p-values indicating the significance of different corpora in the analysis. These values help determine how strongly each corpus influences the results in subsequent branches of the diagram.

Branching Nodes:

The corpus node branches into several additional nodes based on specific corpora, each representing a different set of data sources or variables. The branches are labeled as follows: B N C, M V, N P, S L, B D, I N, and P K.

Each branch represents a corpus or a specific set of data from the given categories. These nodes lead to further breakdowns of the data based on the semantic functions of the introductory it.

Bar Graphs:

Underneath each branch, there are bar graphs labeled node 3 through node 9, representing different subsets of the data, categorized according to the various corpora. These nodes display the distribution of semantic functions across the following categories: A, O A, O E, O H, O N, and O R.

Each bar graph corresponds to a specific node and shows the frequency of each semantic function for that subset of data. The Y-axis of each bar graph ranges from 0 to 1, indicating the relative frequency of each category. The X-axis contains the aforementioned categories, each representing a different semantic function that the introductory it may take.

The highest value is for A in nodes 3 and 4 and O A for all the other nodes. The graphs follow an overall increasing trend except nodes 3 and 4.

When we follow the right branch of the tree, an overall trend that Sri Lanka, India, Pakistan, and Bangladesh have in common is the high proportion of using the intro-it in observation-attitudinal and -reporting functions, as compared to the other group. More specifically, we can observe that SLE splits off from the remaining varieties first (see node 5, p < 0.001) with the highest proportions of observation-attitudinal (see (30)) and -reporting (see (31)) functions (see final node 6).

Further to the right of the tree, node 7 (p = 0.01) splits off IndE from BgE and PkE, with IndE showing the highest proportions of the intro-it to occur in observation-reporting (as in (32), -attitudinal (33), and -neutral (34) functions (see final node 8). Pakistan and Bangladesh, finally, use the pattern most frequently in an observation-attitudinal (see (35)) function, followed by an observation-reporting (see (36)) and -attitudinal (37) function (see final node 9).

It is clearly visible that the major split between the two variety groups shows different preferences for using the intro-it in an attitudinal function (in the case of the British, Nepalese, and Maldivian speakers), whereas the remaining SAVEs prefer using the pattern in observational functions of different subtypes.

13.2.1 Defining (Adverbial) Fronting Phenomena

As a traditional, and, in fact, ‘one of the most consistent rigid SVO [subject-verb-object] languages’ (Givón Reference Givón2001: 235), the canonical and thus standard syntactic structure of the English language involves a subject that precedes its verb and, depending on whether the latter is transitive or not, its object(s) (Givón Reference Givón1993; de Bleser Reference Bleser, Karnath and Thier2012: 426) as well as other obligatory or optional sentence elements, such as complements or adverbials (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 899). In contrast, non-canonical sentence patterns, from the theory-based or structural perspective discussed in the Introduction to this volume, deviate from this S_n-V-X core structure.Footnote ² One pattern that meets this structural criterion of non-canonicity and, as the following discussion will hopefully prove, simultaneously poses a very interesting case with regard to the frequency-based approach to non-canonicity as well, is fronting. It is commonly defined as ‘the initial placement of core elements which are normally found in post-verbal position’ (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1377; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 900). In fact, as the examples in (1) illustrate, in addition to the employment of the mentioned standard postverbal constituents as initial elements of a sentence, sometimes even the predication is affected (Greenbaum & Quirk Reference Greenbaum and Quirk1990: 163):

1. (1)

   a.	${\underset{¯}{\mathrm{O}}}_{\underset{¯}{\mathrm{d}}}‐\mathrm{S}‐\mathrm{V}$
   	Whether I can attend, I don’t know.Footnote 3	(preposed direct object)
   b.	${\underset{¯}{\mathrm{C}}}_{\underset{¯}{\mathrm{S}}}‐\mathrm{S}‐\mathrm{V}$
   	A strange person he is.	(preposed subject complement)
   c.	${\underset{¯}{\mathrm{C}}}_{\mathrm{O}}‐\mathrm{S}‐\mathrm{V}‐{\mathrm{O}}_{\mathrm{d}}$
   	John they called the boy.	(preposed object complement)
   d.	$\underset{¯}{\mathrm{V}}‐{\mathrm{O}}_{\mathrm{d}}‐\mathrm{Aux}‐\mathrm{S}$
   	Drawing a picture was the little girl.	(preposed predication)
   e.	$\underset{¯}{\mathrm{A}}‐\mathrm{S}‐\mathrm{V}‐{\mathrm{O}}_{\mathrm{d}}$
   	On the table she put the book.	(preposed obligatory adverbial)
   f.	$\underset{¯}{\mathrm{A}}‐\mathrm{S}‐\mathrm{V}‐{\mathrm{O}}_{\mathrm{d}}$
   	Last week I called him.	(fronted optional adverbial)

In this chapter the focus is on the initial placement of one constituent in particular, namely the adverbial. While research has commonly focused on the discussion of instances like (1e) above, where a grammatically demanded adverbial is moved to the front, I propose an extension of the consideration of ‘core’ constituents to optional adverbials (as in (1f)), reserving the notion of preposing for the commonly marked movement of obligatory constituents and that of fronting for the common – and thus unmarked – sentence-initial placement of optional adverbials. It should be noted here that the term markedness is defined in various ways (see Haspelmath Reference Haspelmath2006 for an overview). In this chapter, and in line with the frequency-based approach to canonicity discussed in the Introduction to this volume, the term refers to the uncommonness or rarity of a feature in language use, as in this case, while the other approach to non-canonicity concerns the deviation from an expected theory-based or structural norm, as suggested by Halliday (Reference Halliday1994) in relation to his marked theme. This makes the topic of this chapter all the more interesting because its focus on fronting phenomena implies that, if a given sentence contains a sentence-initial optional adverbial, it is unmarked or canonical from a frequency-based perspective while the deviation from the S-V-X norm makes it structurally non-canonical. In case of preposing, however, both characteristics of non-canonicity – the frequency-based and the theory-based – are met.

In fact, it has been stressed repeatedly in the literature that preposing (as opposed to fronting) is a comparatively rare pattern (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985; Sinclair Reference Sinclair1990) while the optionality and innate flexibility of the majority of adverbials make them a fruitful but, in the context of non-canonicity, rarely discussed structure. The reason for the proposed extension is that previous research has proven that there are (discourse-)pragmatic, (information-)structural, semantic, and cognitive effects associated with the allocation of an adverbial (be it optional or obligatory) to the first slot within a construction (Sinclair Reference Sinclair1990; Diessel Reference Diessel2005; Wiechmann & Kerz Reference Wiechmann and Kerz2013; Junge et al. Reference Junge, Theakston and Lieven2015), and both preposing and fronting are subject to comparable if not identical (information-structural) implications, a topic that will be discussed later on in this section.

First, however, it is necessary to take a look at the diverse structural and functional peculiarities of adverbials in general. With regard to the former, as briefly indicated in the introduction, adverbials may be realised in various ways and in varying degrees of complexity, ranging from individual adverbs (such as luckily, perhaps, or frankly in (5a–c) below)Footnote ⁴ over adverb, noun, and prepositional phrases (such as slowly but surely in (3c), last Monday in (3a) or on the table in (3b) below) to different kinds of dependent clauses, illustrated under (2) (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 767–8).

1. (2)

   a.	Verbless:	If in doubt, the example is out.
   b.	Finite:	Although she likes Jon, she did not invite him.
   c.	Non-finite (to-inf.):	To be honest, I don’t know how to react.
   d.	Non-finite (-ing):	Working tirelessly for hours, she immediately fell asleep.
   e.	Non-finite (-ed):	Compared to last year, they did a good job.

As far as the functional characteristics are concerned, over the years, various classification schemes have been suggested, including Quirk et al.’s (Reference Quirk, Greenbaum, Leech and Svartvik1985) differentiation between adjuncts, conjuncts, disjuncts, and subjuncts or Halliday’s (Reference Halliday1994) and Huddleston and Pullum’s (Reference Huddleston and Pullum2002) distinction between different kinds of adjuncts. The taxonomy adopted here is that by Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999), who distinguish between so-called circumstance, linking, and stance adverbials. The former type is employed ‘to add circumstantial information about the proposition in the clause’ (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 762), which may be temporal or spatial in nature or provide details about the process or manner, contingency, the extent or degree, or the recipient (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 776–81). They are exemplified in (3):Footnote ⁵

1. (3)

   Adverbials of circumstance
   a.	Time:	Last Monday she cancelled her meeting.
   b.	Place:	On the table she placed a heavy book.
   c.	Process/manner:	Slowly but surely, the project is evolving.
   d.	Contingency:	Because he lied repeatedly, he lost her trust.
   e.	Extent/degree:	To a certain degree, I think she is right.
   f.	Recipient:	For his family he would move mountains.

The second type, linking adverbials, is used to establish a connection between a given ‘clause (or some part of it) … [and] some other unit of discourse’ (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 762) and it therefore contributes significantly to the cohesion and coherence of a given output. Again, various subtypes, listed in (4), can be distinguished (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 875):

1. (4)

   Linking adverbials
   a.	Addition/enumeration:	Firstly, I want to discuss the unemployment rate.
   b.	Apposition:	In other words, their conclusions were well-founded.
   c.	Contrast/concession:	However, this shouldn’t stop you.
   d.	Result/inference:	Hence, you shouldn’t rely on his help.
   e.	Summation:	All in all, the event was a success.
   f.	Transition:	Incidentally, you never know what life has in store for you.

Finally, stance adverbials express the speaker’s or writer’s attitude, point of view, or opinion as well as their assessment of, for example, the certainty or reliability of the proposition of a given message (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 762). Three major subtypes, listed in (5), can be identified (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 854):

1. (5)

   Adverbials of stance
   a.	Attitude:	Luckily, they returned home safe and sound.
   b.	Epistemic:	Perhaps, I should leave now.
   c.	Style:	Frankly, I don’t give a damn.

The fact that adverbials are (more commonly than not) optional goes hand in hand with their flexibility, that is, with the possibility to employ the constituent in varying slots within a given sentence, ranging, according to Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985: 490), from Initial, Medial, and End (for the first, post-operator, and final position) to four intermediate slots.

Which position is ultimately chosen may depend on both their structural make-up – with individual adverbs being considered particularly flexible (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 491) – and the semantic categories discussed above. While linking adverbials have been found to be most commonly used initially, stance and circumstance/place adverbials tend to be placed in medial or final position, respectively (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999). The situation may differ, however, when several adverbials are used in juxtaposition. Hasselgård (Reference Hasselgård2010), drawing on Quirk et al.’s (Reference Quirk, Greenbaum, Leech and Svartvik1985) distinction, investigated adjunct adverbials and their placement in British English and, in doing so, also took a look at common patterns of adverbial clustering in fronted position. She determined that a combination of two adjuncts, one of time and one of space, was by far the most frequent and that these two are generally the types that are most likely to be found in combinations with other adverbials. Focusing adverbs (e.g., especially), considered as subjuncts in Quirk et al.’s (Reference Quirk, Greenbaum, Leech and Svartvik1985) taxonomy, are also common, but, if they are used, they always appear in cluster-initial position, modifying the following adverbial (Hasselgård Reference Hasselgård2010: 92–3).

However, the semantic class poses just one factor that exerts an influence on the positioning of adverbials. Even greater is the influence of powerful mechanisms that are at play when structuring a message, namely those of information structure (Halliday Reference Halliday1967), defined as ‘the complex interaction of numerous phenomena and principles that govern the organization of information in discourse’ (Callies Reference Callies2009: 10; see also Féry & Ishihara Reference Féry and Ishihara2016). These involve aspects like the thematic structure, the information status progression, syntactic weight, focus placement, or other stylistic reasons, some of which will be briefly discussed in the following.

The thematic structure, as the first information-structural principle, is concerned with the organisation of a clause as a message (Halliday Reference Halliday1994: 37). From a Hallidayan perspective, the theme, as the starting point of the message, which is further developed in what follows (i.e., the rheme), commonly constitutes the subject. If this criterion is not met, however, and the position is occupied by a different constituent, the theme is inevitably marked and thus receives special attention or emphasis (Reference Halliday1994: 44). This view is supported by numerous researchers, including Quirk et al. (Reference Quirk, Greenbaum, Leech and Svartvik1985), who consider the thematic constituent ‘the first thing that strikes the speaker … [while] the rest is added as an afterthought’ (Reference Quirk, Greenbaum, Leech and Svartvik1985: 1377) or Virtanen (Reference Virtanen and Virtanen2004: 81), who equates thematicity with informational foregrounding. This makes apparent the special status the first constituent occupies within a given sentence and thus provides evidence for the relevance of optional adverbials in relation to the topic of non-canonicity. Diessel (Reference Diessel2005), in his investigation of different types of adverbial clauses, also recognises the importance of the first position and makes clear that adverbials can be particularly beneficial in that they may be ‘used to organize the information flow in the ongoing discourse … [and] function to provide a thematic ground or orientation for subsequent clauses’ (Reference Diessel2005: 459).

The topic of information flow directly leads to another important principle that is strongly affected by fronting phenomena, namely the information status progression. It is concerned with the newness and givenness of information (Clark & Haviland Reference Clark, Haviland and Freedle1977; Prince Reference Prince and Cole1981) and thus with facts which are ‘new and at the centre of the … communicative interest’ (Bache & Davidsen-Nielsen Reference Bache and Davidsen-Nielsen1997: 114) and those which are ‘assume[d] to be known’ (Bache & Davidsen-Nielsen Reference Bache and Davidsen-Nielsen1997: 113) or are retrievable ‘from the preceding discourse’ (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 896). In the English language, the so-called given-new contract is assumed to hold between interlocutors (Clark & Haviland Reference Clark, Haviland and Freedle1977: 3), implying that the standard progression of the information status is from what is known to what is newly introduced to an interaction. As previous research has shown, this sequence may, in fact, be adhered to by means of both fronted and preposed adverbials. While this observation may seem natural, for example, for linking adverbials, which establish a connection between the previous and the upcoming discourse, it has also been confirmed for sentence-initial clausal adverbials which, according to Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 835), frequently entail known information and, in a way, prepare the recipient for the new information that is to follow in the remainder of the message.

Two major principles that are also frequently associated with discourse-new information are those of end-focus and end-weight. The former claims that special focus, and thus emphasis, is placed on the final ‘lexical item of the last element in the clause’ (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 897), which commonly coincides with information that is newly introduced to the discourse. Likewise, new information is commonly provided in a ‘longer, “heavier” structure’ (Greenbaum & Quirk Reference Greenbaum and Quirk1990: 398), which, in accordance with the principle of end-weight (Behaghel Reference Behaghel1909: 139), is likely to be placed at the end of a sentence and after providing the necessary context. This reduces cognitive efforts on the part of both the producer and the recipient of a message, given that lengthy constituents do not have to be retained in the working memory until the point of grammatical completeness of a given construction is actually reached.

The way adverbials interact with these principles is twofold. On the one hand, researchers like Virtanen (Reference Virtanen2008) have found that adverbials of manner are, for example, frequently moved to the front in order to allow more complex constituents to be placed finally, thus enabling the adherence to the principle of end-weight. On the other hand, however, previous studies, in particular into the positioning of clauses, have determined that the clause semantics may, in fact, override (information-)structural as well as cognitive factors (Diessel Reference Diessel2005; Wiechmann & Kerz Reference Wiechmann and Kerz2013), that is, that lengthier and more complex constituents, which naturally entail greater processing efforts, or those that contain discourse-new and focused information, are put first due to their semantic class. Diessel (Reference Diessel2001) developed a sequence of adverbial clauses based on their likelihood to be encountered before or after their juxtaposed main clause. According to his findings, the most frequent fronted adverbial clauses (all of which are circumstance adverbials in Biber et al.’s Reference Biber, Johansson, Leech, Conrad and Finegan1999 framework) are conditionals, followed by temporal and causal clauses, while result and purpose clauses are least commonly found sentence-initially.

However, as has been established, such an intentional deviation from the canonical word order, and thus from common information-structural principles, brings with it the possibility to establish emphasis, which is also at play when dramatic or rhetoric effects are to be achieved (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1378) or when, in line with Givón’s (Reference Givón and Haiman1985) principle of task-urgency, which highlights ‘the main purpose of the utterance’ (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 904), the element that is placed initially is ‘contextually most demanded’ (Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985: 1377).

The final two functions of fronting phenomena that will be discussed here are concerned with the relation between sentences in discourse. By using determiners or ‘anaphoric deictic markers such as this, that, these and such’ (Callies Reference Callies2009: 39), by repeating given information (Callies Reference Callies2009: 39), or by the use of particular types of linking adverbials, both preposing and fronting pose a fruitful means to establish cohesion or, as illustrated below in (6), to express a contrasting relationship between units of discourse:

However is used here to mark the contrast between the expected experience expressed in the first and the actual experience referred to in the second sentence. As becomes clear, this is typically achieved by providing different alternatives that are contrasted by means of links acting as contrasting connectives (Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999: 900–1). In fact, according to Callies (Reference Callies2009: 39), this ‘contrastive emphasis’ is actually the most common function of fronting phenomena. However, the function of establishing cohesion and coherence is not exclusive to linking adverbials. As Hasselgård (Reference Hasselgård1996) makes clear, both temporal and spatial adverbials may also contribute to cohesive discourse (cf. Reference Hasselgård1996: 110).

Having discussed the diverse structural and functional realisations of adverbials as well as the information-structural implications their initial placement may have, one final aspect that needs to be mentioned is that of inversion, without which, according to Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 900), a true understanding of fronting phenomena is not possible, since the preposing of certain elements, among them particular types of adverbials, triggers the phenomenon. This is illustrated in the examples under (7):

1. (7)

   a.	[On the wall]A [hung]V [a nice picture.]S	(subject-verb inversion)
   b.	[Next]A [came]V [the Queen.]S	(subject-verb inversion)
   c.	[Rarely]A [have]Aux [I]S [seen]V [this.]O	(subject-operator inversion)

As indicated, the initial placement of various types of adverbials may lead to the inversion of the subject and the entire verb phrase or its operator. These include opening adverbials of place, as in (7a), or time, illustrated in (7b), as well as negative or restrictive elements such as only, never, scarcely, or, as in (7c), rarely.

13.2.2 Previous Research into Adverbial Fronting Phenomena in Learner Language

Previous research on fronting phenomena as such has been versatile. When it comes to preposing, quite a bit of research has been dedicated to both English as a Native Language and English as Second Language contexts (e.g., Quirk et al. Reference Quirk, Greenbaum, Leech and Svartvik1985; Birner & Ward Reference Birner and Ward1998; Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan1999; Esser Reference Esser, Fischer, Tottie and Lehmann2002; Huddleston & Pullum Reference Huddleston and Pullum2002; Ward & Birner Reference Ward, Birner, Horn and Ward2006 on ENL; Lange Reference Lange2012; Winkle Reference Winkle2015; Götz Reference Götz2017 on ESL, to name but a few),Footnote ⁶ all of which have greatly contributed to the understanding of the structure, its characteristics, and its underlying mechanisms, also discussed by Götz and Kircili (Chapter 12 in this volume).

Research into learner language, however, has been scarce, with two laudable exceptions being Plag and Zimmermann (Reference Plag, Zimmermann, Börner and Vogel1998) and Callies (Reference Callies2009), both of whom investigated German learners of English. Callies (Reference Callies2009), using a multi-method approach, investigated multiple non-canonical sentence patterns. His results on preposing confirm that learners, just like native and second-language speakers of English, rather shy away from employing the construction, with the corpus analysis yielding only five instances of preposing of an obligatory constituent, one of which had been produced by a native speaker and four by learners of English (Reference Callies2009: 194–5). In relation to the use of sentence-initial optional adverbials, Callies (Reference Callies2009: 178) remarks that they tend to be overrepresented in the L2 output of German learners, which may be due to the fact that they are not familiar with the previously mentioned principles of end-weight or end-focus.

Other studies on the use of adverbials and adverbial fronting by learners have paid a lot of attention to the use of connectors/linking adverbials and, in this connection, especially their overrepresentation in learners from different L1 backgrounds, which is, in fact, so common that it has been referred to as a potential ‘universal feature of interlanguage’ (Gilquin et al. Reference Gilquin, Granger and Paquot2007: 328). Granger and Tyson (Reference Granger and Tyson1996), for instance, compared their employment by native speakers with that by French learners. They determined that there is no general overrepresentation but solely an over- or underrepresentation of connectors with particular functions. An overrepresentation was determined, for example, for connectors that either corroborate and thus support the argument (e.g., in fact, indeed) or provide an example (e.g., for instance, namely) or additional information (e.g., moreover) (Reference Granger and Tyson1996: 20). Connectors which serve to establish contrast (e.g., though, however) or cohesion (e.g., thus, therefore) were found to be less frequently used by learners. This implies that, while connector use for emphasis or exemplification seems common, using them to ‘change the direction’ of arguments or develop them logically does not (Granger & Tyson Reference Granger and Tyson1996: 20). Reasons include the insufficient knowledge of students concerning the semantic restrictions and the syntactic behaviour of individual adverbials as well as their ‘inexperience in manipulating connectors within the sentence structure’ (Granger & Tyson Reference Granger and Tyson1996: 25). They conclude that, instead of providing them with lists of words and phrases, students should learn to use them not as stylistic enhancers but as higher-level discourse units with a variety of semantic functions and a high degree of syntactic flexibility. This increased awareness of their semantic, stylistic, and syntactic properties, achieved, for instance, by studying authentic texts, might help students to attain the ultimate goal behind the use of connectives, namely cohesion in discourse (Granger & Tyson Reference Granger and Tyson1996: 25–6). In fact, this connection between curricula activities, designated teaching materials and learner productions has been proven in a number of studies; so much so that van Vuuren and Berns, in their Reference Vuuren and Berns2018 study on Dutch and Francophone learners of English, referred to it as a ‘teaching-induced interlanguage feature’ (Reference Vuuren and Berns2018: 457) (see also Field & Oi Reference Field and Oi1992 on Cantonese; Altenberg & Trapper Reference Altenberg, Tapper and Granger1998 on Swedish; Milton Reference Milton and Granger1998 on Chinese; or Leńko-Szymańska Reference Leńko-Szymańska2008 on Polish, Russian, Spanish, French, Swedish, Finnish, and German learners of English).

Another common focus of previous research endeavours has been on transfer-related performances. For Dutch learners, for example, it has been found that they use more circumstantial and linking adverbials, while the number of stance adverbials in fronted position is smaller than that used by native writers. With increasing proficiency, learners’ means develop in the direction of native writing, however. The only exception are, in fact, linking adverbials, in case of which the learners clearly exhibited a continuous overuse. This fact is attributed to both the teaching materials and language transfer (van Vuuren Reference Vuuren2017; van Vuuren & Laskin Reference Vuuren and Laskin2017), given that linking adverbials (as well as local anchors) as cohesive devices were not only overrepresented in the English L2 output of their Dutch participants but also in their L1 writing (van Vuuren & de Vries Reference Vuuren, de Vries, de Haan, de Vries and van Vuuren2017; see also Larsson Reference Larsson2017 on the use of probably and possibly by Swedish and Norwegian and of maybe and perhaps by Spanish learners of English in Larsson et al. Reference Larsson, Callies, Hasselgård, Laso, van Vuuren, Verdaguer and Paquot2020).

13.3.1 Research Gaps and Research Questions

Following the account of previous research endeavours into the topic, two research gaps were identified:

1. (1) There is a lack of research into the formal and functional peculiarities of adverbial fronting phenomena in German learner language going beyond the consideration of particular kinds of adverbials.

2. (2) There is a lack of research involving predictive modelling that might contribute to the understanding of the use of the structures by learners of English.

Consequently, this study intends to answer the following research questions:

1. (1) Can significant differences be determined between the frequencies of the different semantic and/or syntactic types of adverbials in the native speaker vs. the learner data and/or do the findings support previous research (particularly on conjuncts/adjuncts)?

2. (2) Can information-structural variables (information status or the syntactic complexity of the initial element) be determined as predictors for the use of adverbial fronting in the two corpora?

13.3.2 Databases and Methodology

This study makes use of the German sub-component of the International Corpus of Learner English (ICLE) and the Louvain Corpus of Native English Essays (LOCNESS; Granger Reference Granger and Granger1998). ICLE is a corpus of argumentative essays produced by advanced (university-level) students of English, compiled with the aim of enabling the investigation of ‘the interlanguage of advanced learners from various mother-tongue backgrounds’ (Granger & Tyson Reference Granger and Tyson1996: 18). The version employed for this chapter contains a total number of approximately 3.7 million words from 16 different first-language backgrounds. The German component contains about 234,000 words (Granger et al. Reference Granger, Dagneaux and Meunier2009).

LOCNESS was compiled at the Université catholique de Louvain in order to provide control data for ICLE (Granger & Tyson Reference Granger and Tyson1996: 19). The corpus contains approximately 320,000 words and it, too, is mostly made up of argumentative essays of 500+ words on a variety of topics, written by British A-level as well as British and American university students (‘LOCNESS description’ n.d.). For the present study, it was decided to use the American English argumentative essays in spite of the fact that the variety of instruction in Europe (and thus also in Germany) tends to be British English (Leńko-Szymańska Reference Leńko-Szymańska2008) because the former is commonly considered ‘the most important and influential dialect’ due to ‘films, television, popular music, the Internet and the World Wide Web, air travel and control, commerce, scientific publications, economic and military assistance, and activities … in world affairs’ (Algeo Reference Algeo2010: 183; cf. also Schneider Reference Schneider, Kachru, Kachru and Nelson2009). Consequently, the assumption is that the language of university students in their early twenties is likely to be shaped more by the continuous influence of and exposure to the language in these areas than by what they were taught at school and thus predominantly in the first years of language acquisition. Table 13.1 provides an overview of the material used.

Table 13.1Overview of the analysed data from ICLE and LOCNESS

Table 13.1Long description

The table is divided into 4 columns. The first column header is empty and the others are labeled as I C L E, L O C N E S S, and Total. The rows are filled from left to right and the data are filled as follows:

• For the number of essays, the data are 74, 48, and 122.

• For the number of sentences, the data are 1619, 1630, and 3249.

• For excluded sentences, the data are 193, 142, and 335.

• For the total, the data are 1426, 1488, and 2914.

As indicated in Table 13.1, 74 out of 437 ICLE-Germany essays were randomly selected and compared with 48 out of 188 American English argumentative essays from LOCNESS, with the topics ranging from experiences with friends, pros and cons of cars, and environmental issues (in ICLE) to important discoveries, capital punishment, or feminism (in LOCNESS). Although the difference in the number of essays is not ideal, it was deemed more important to have a comparable number of sentences and to analyse complete essays (as opposed to the consideration of, e.g., only a particular number of sentences per essay). This has the advantage that, for instance, summative linking adverbials, which are naturally used towards the end of a text, also feature in the analysis.

Turning to the annotation process itself, Table 13.2 shows the parameters for which each sentence was manually annotated. It is based on an annotation scheme used by Götz (Reference Götz2017) for her study on fronting phenomena in South Asian varieties of English, which was adopted with minor adjustments. Other than the given-new status (category gn) and the semantic function (front_function), the majority of variables are dedicated to the syntactic form and function of the first constituent of every sentence (e.g., front_el, front_form, front_obl_tf). While the explanations provided in the table are sufficient for the majority of the individual categories, a more detailed discussion is in order when it comes to the given/new distinction. According to Prince’s (Reference Prince and Cole1981) taxonomy of assumed familiarity, a threefold distinction can be made between new information, information that is inferable, or evoked (or given) information (see also Pham, Chapter 9 in this volume). In a first step of the analysis, this distinction was generally followed (with a slightly more extensive consideration of inferable entities than originally suggested in Prince’s work) to be able to obtain a more fine-grained picture of the information status in the essays, but, for the sake of comparability with other studies, it was then decided to conflate the inferable and the given categories to be able to proceed with just a twofold distinction between given and new information. For the statistical evaluation, different statistical tests and methods were employed, including descriptive statistics, conditional inference trees (cf. Hothorn et al. Reference Hothorn, Hornik and Zeileis2006), and logistic regressions using the R package dplyr in R Studio.

Table 13.2Annotation scheme used for the corpus analysis with explanations

Table 13.2Long description

The table is sectioned into 2 columns. The columns are labeled as category and explanation. There are 13 rows for different variables. The data in the rows are filled from left to right as follows:

• The data for C O R P U S is construction at hand belongs to I C L E or L O C N E S S.

• The data for F I L E N A M E is filename of the essay the construction is taken from.

• The data for C or C N is Sentence is canonical, C or non-canonical, N C.

• The data for S underscore length is number of words the sentence at hand consists of.

• The data for F R O N T underscore I F is sentence-initial element constitutes subject, F or any other sentence constituent T.

• The data for G N is first element constitutes given or inferable, G or new information, N.

• The data for A underscore F R O N T underscore T F is first element constitutes adverbial T or not F.

• The data for F R O N T underscore E L is first element constitutes subject S, object O, complement C, verb V, or adverbial A.

• The data for F R O N T underscore F O R M is Syntactic realisation as one-word expression x, phrase (N P, P P, A D V P, A D J P) or clause C L.

• The data for F R O N T underscore F U N C T I O N is Semantic function of the adverbial – circumstance C, stance s, or linking L.

• The data for F R O N T underscore O B L underscore T F is Sentence-initial element is obligatory T or optional F.

• The data for L E N G T H underscore E L underscore I is number of words the first element consists of.

• The data for C L U S T underscore T F is the sentence at hand begins with adverbial cluster T or not F.

13.3.3 Results and Discussion

To establish a basis for the following results, Figure 13.1 provides an overview of the frequency distributions of the placement of different sentence elements in sentence-initial position.

Figure 13.1 Frequency distribution of different sentence elements in fronted position (percentage)

Figure 13.1Long description

The vertical axis marks numbers that range from 0 to 80 in increments of 10. The horizontal axis marks the different annotation schemes used to explain the corpus I C L E or international corpus of learner english and L O C N E S S or louvain corpus of native english essays. The annotations used are subject or s, verb, or v, object or o, complement or c, and adverbial or a. There are two vertical bars where the dark shaded one represents I C L E and the light shaded one represents L O C N E S S. The subject and adverb use is predominantly seen in L O C N E S S and I C L E. Other annotations are much smaller compared to them. A data table is provided at the bottom with the distribution of the annotations given from left to right for I C L E and L O C N E S S, respectively. The values are as follows: For I C L E, the corresponding values are 54.56, 0.00, 1.47, 0.28, and 43.69. For L O C N E S S, the corresponding values are 69.15, 0.07, 0.67, 0.20, and 29.91.

As can be seen in Figure 13.1, objects, complements, and verbs were hardly found sentence-initially in the data, while the canonical subject was the most commonly chosen sentence-initial constituent in both populations, constituting 54.56% (n = 779) of the 1,426 sentences taken from ICLE and 69.15% (n = 1029) of the 1,488 sentences from LOCNESS. Consequently, adverbials were, in fact, the second most frequent sentence-initial element in both corpora, employed in 29.91% (n = 445) of the cases in LOCNESS compared to 43.69% (n = 622) in ICLE. Instances of adverbial preposing (and thus the use of obligatory adverbials as sentence-starters) were, as suggested by previous research, exceptionally rare, constituting only 2.02% (n = 9) of the instances identified in the native speaker and as little as 0.48% (n = 3) of the cases in the learner data. In both populations, see (8) and (9), it was either the use of negative or restrictive adverbs or the use of spatial references, both of which trigger inversion, that resulted in the employment of preposing:Footnote ⁷

This observation already illustrates the crucial role of adverbials in relation to fronting (as the initial placement of optional constituents) since they are, by far, the most common non-canonical fronted element.

Zooming in on their different functions, Figure 13.2 provides an overview of percentage distributions of the different semantic types (circumstance, linking, stance) of adverbials in the two corpora.

Figure 13.2 Frequency distribution of adverbial functions in ICLE and LOCNESS (percentage)

Figure 13.2Long description

The vertical axis marks numbers that range from 0 to 60 in increments of 10. The horizontal axis marks the different front functions used to explain the corpus I C L E or international corpus of learner english and L O C N E S S or louvain corpus of native english essays. The annotations used are circumstance or c, linking or l, and stance or s. There are two vertical bars where the dark shaded one represents I C L E and the light shaded one represents L O C N E S S. I C L E is lower in c, higher in l and s. A data table is provided at the bottom with the distribution of the annotations given from left to right for I C L E and L O C N E S S, respectively. The values are as follows: For I C L E, the corresponding values are 39.65, 41.57, and 18.78. For L O C N E S S, the corresponding values are 51.24, 32.81, and 15.96.

Based on the percentages, we can derive clear preferences. While both groups least frequently opted for stance adverbials, circumstance and linking adverbials were preferred by native speakers and German learners, respectively, with both the stance and linking adverbials being significantly more frequently employed by the latter group (stance: χ² = 10.829, df = 1, p < 0.001; linking: χ² = 31.05, df = 1, p < 0.001). In particular the results concerning linking adverbials thus seem to confirm previous research in that these adverbials tend to be overrepresented in learner data. In fact, when taking a more qualitative approach and taking a closer look at individual concordance lines, it becomes apparent that the results corroborate some of the findings reported by Granger and Tyson (Reference Granger and Tyson1996). Just like in their study on French learners, German students seem to rely on fronted adverbials with a linking function to expand on information previously provided, using, for example, moreover, therefore, or so, but refraining from sentence-initial also, a potential false friend in German, which is found only twice in the learner but 16 times in the native speaker data. Interestingly, the establishment of contrast, commonly fulfilled, for example, by however in the American data, was, in fact, instead achieved by the use of the subordinating conjunction but, used as many as 81 times in ICLE (compared to 13 instances in LOCNESS). Given the fact that, even in their mother tongue, starting a sentence with but is considered stylistically rather clumsy, it is surprising that German learners still felt compelled to use the conjunction for this purpose in their L2 output. In both corpora, linking adverbials were also identified as the semantic type that was most commonly attested in adverbial clusters, as exemplified below. Both combinations, illustrated in (10) and (11), contain linking as well as circumstance adverbials (roughly corresponding to the adjunct category investigated by Hasselgård Reference Hasselgård1996).

Moving on to the structural distribution, Figure 13.3 shows the frequency of the structural realisations of sentence-initial adverbials (i.e., CLAuse, PHRase, SINgle adverb) in percent.

Figure 13.3 Frequency distribution of syntactic realisations in ICLE and LOCNESS (percentage)

Figure 13.3Long description

The vertical axis marks numbers that range from 0 to 60 in increments of 10. The horizontal axis marks the different sentence-initial adverbials used to explain the corpus I C L E or international corpus of learner english and L O C N E S S or louvain corpus of native english essays. The annotations used are clause or C L A, phrase or P H R, and single or S I N. There are two vertical bars where the dark shaded one represents I C L E and the light shaded one represents L O C N E S S. I C L E is higher in C L A and P H R, as shown in the graph. A data table is provided at the bottom with the distribution of the annotations given from left to right for I C L E and L O C N E S S, respectively. The values are as follows: For I C L E, the corresponding values are 21.22, 30.71, and 48.07. For L O C N E S S, the corresponding values are 29.44, 35.73, and 34.83.

What can be seen here is that the native speakers exhibited a tendency to use more complex syntactic realisations (i.e., phrases or clauses) more frequently, namely in 35.73% and 29.44% of the cases, while learners only used them in 30.71% and 21.22% of the instances, respectively. Still, the only statistically significant difference was, in fact, determined for single constituents (χ² = 45.674, df = 1, p < 0.001), used in 48.07% (n = 299) of the learner and 34.83% (n = 155) of the native speaker cases. The tendency for the American students to express themselves in more complex structures is exemplified in (12) to (14) below. While the first two sentences contain phrasal structures that, in ICLE, would have predominantly been expressed using a single constituent (e.g., therefore), the third example is particularly interesting. The reason is that, as is frequently the case in this subcorpus, this participant fronts a dependent clause, obviously in order to particularly emphasise the reason behind the points that were made in the preceding discourse.

This final example is also indicative of the observations made for the syntactic preferences of each type of adverbial, with circumstance adverbials being the only case for which clauses were preferred by native speakers while learners rather opted for phrase structures. For each of the other semantic categories, the single-word expression was, in fact, the most commonly chosen alternative.

Regarding the second research question, enquiring whether certain information-structural variables can be identified as predictors for adverbial fronting, the ctree in Figure 13.4 already alludes to the fact that, overall, there are significant differences between the two corpora. Likewise, both the given-new variable and the sentence length proved to be of relevance to the nature of the constituent that is found in sentence-initial position.

Figure 13.4 Ctree: Distribution of front_el~corpus+s_length+gn

Figure 13.4Long description

The conditional inference tree shows a hierarchical classification based on corpus type, sentence length, and givenness, G or N, ending in bar charts. At the top, Node 1 is labeled C O R P U S, branching into two categories: L O C N E S S and I C L E.

The L O C N E S S branch leads to Node 2 labeled S_L E N G T H, which splits into two paths:

• Less than 16, leading to Node 3 labeled G N, which divides into:

G leading to Node 4,

N leading to Node 5.

• Greater than 16, leading to Node 6 labeled G N, which splits into:

N leading to Node 8, which continues to Node 7 labeled S_L E N G T H, dividing into:

Less than 34 to Node 9

Greater than 34 to Node 10

G leads directly to Node 10.

The I C L E branch from Node 1 leads to Node 11 labeled S_L E N G T H, which divides into:

• Less than 22, leading to Node 13, then to Node 12 labeled G N, branching into:

N to Node 14

G to Node 15

• Greater than 22 connects directly to Node 15.

All terminal nodes (Nodes 4, 5, 9, 10, 14, 15) include small bar charts.

Each bar chart shows relative frequency values from 0 to 1 across five categories labeled A, C, O, S, and V.

In general, ctrees (Hothorn et al. Reference Hothorn, Hornik and Zeileis2006) are based on recursive partitioning algorithms, providing a means to meaningfully split the values or manifestations of the dependent variable based on the independent variables that are assumed to exert an influence. One major advantage is the fact that ctrees also detect behavioural patterns in fairly small datasets, making this method a perfect choice for the present chapter.

As can be seen, even though the first significant split separates the learner from the native speaker data, they also share certain behaviours, such as the fact that the next important splits (Node2 (p < 0.001), if the left branch of the tree is followed, and Node11 (p < 0.001), which is reached to the right) are both based on the length of the sentence at hand. Focusing on the learners, and thus on the right part of the tree, sentence-initial adverbials, compared to other sentence-initial constituents, are particularly commonly chosen in sentences that are longer than 22 words (cf. Node15). If they are shorter than that, the given-new distinction is of relevance (Node12; p = 0.02) and they are more frequently employed if they contain given information (Node14). When it comes to native speakers, and thus the left part of the tree, Node2 separates sentences from each other that are either longer or shorter than 16 words and in both cases, the next important split is the given-new status of the first constituent (Node6 (p = 0.008) for sentences that exceed 16 words, and Node3 (p = 0.034) for those that are shorter). If the branches of Node3 are followed, it can be seen that the percentages of canonical sentences (commencing with a subject) are much higher than those of sentence-initial adverbials (cf. Node4 and Node5). Interestingly, in Node5, which provides the frequency distributions for new information, the percentage of fronted adverbials is slightly higher than when the information is given (visible in Node4). This is striking because, as mentioned previously, the standard progression of the information status in an English sentence would commence with given and proceed to new information. This observation also holds true for particularly long constructions. If the left branch of Node6 is followed, which, again, refers to new information in sentence-initial position, there is another split (Node7; p = 0.025) between sentences that are up to 34 words long and those that exceed 34 words, with the percentage being particularly high for the latter instances (Node9; note, however, that this figure is only based on 34 cases).

However, to determine whether the three variables corpus, gn, and s_length also have a predictive power, the glm function (allowing for two-way interaction) was used and yielded the results depicted in Figure 13.5.

Figure 13.5 Effect plot of glm a_front_tf~log(s_length)+gn+corpus+gn:corpus

Figure 13.5Long description

Two graphs side-by-side effect plots showing how syntactic variables relate to A F R O N T dash T F values. The left plot, titled S_L E N G T H effect plot, displays a single line graph with a shaded confidence band. As sentence length increases from 0 to 80 along the x-axis, the A F R O N T dash T F value on the y-axis rises from 0.1 to 0.6, showing an upward trend. The right plot, titled G N by C O R P U S effect plot, has two smaller panels. The first is labeled C O R P U S equals I C L E and shows a downward sloping line from G to N. The second is labeled C O R P U S equals L O C N E S S and shows an upward sloping line from G to N.

Even though the model ‘only’ explains 5.9% of the variance in the data (R² = 0.059) it turned out to be highly significant and to perform significantly better than the baseline model. As the effect plots in Figure 13.5 make clear, the observations made for the above ctree are, in a way, confirmed and the variable Corpus was returned as a significant predictor (p < 0.001). The plot on the left indicates that, for both varieties, the likelihood for adverbial fronting to occur increases significantly with an increasing sentence length (p < 0.001), an observation that can be explained by the fact that additional/explicatory information or temporal references, as they are frequently provided by circumstantial adverbials, are commonly provided in longer structures as in (15):

The plot on the right, on the other hand, implies that there is a significant interaction between the given-new structure and the corpus. While the learners show a tendency for adverbial fronting to occur when the constituent contains given information, in LOCNESS it is significantly more frequently used with new information (p < 0.05) as illustrated in (16) and (17).

At first sight, this seems like a surprising result given that it implies that, while German learners structure their sentences in conformity with the given-new contract, native speakers frequently opt to deviate from the standard progression when using sentence-initial adverbials. However, considering the suggestion made by Diessel (Reference Diessel2005), the finding may confirm that, under certain circumstances, particular information-structural principles can override each other and that a deviation is acceptable, for example, to enable an adherence to the principles of end-weight or end-focus or in order to achieve balanced weight, as in (16) above, where the lengths of the preverbal and the postverbal constituents (in number of words) are roughly the same.

14.2.1 Emergent Grammar in ELF

ELF interactions come in manifold forms, which makes it hard – or maybe even impossible – to define what ‘ELF grammar’ actually looks like. In consequence, recent research has moved away from regarding ELF as a stable code and rather conceptualises it as ‘a series of more or less demanding communicative situations where speakers come with whatever their language skills to tackle the communicative tasks at hand’ (Ranta Reference Ranta, Jenkins, Baker and Dewey2017: 247). According to this view, ELF grammar is in a state of fast change, strongly depending on the linguistic backgrounds and proficiency levels of its speakers, who are engaging in specific interactional encounters, which can be temporary and clearly task-based (such as asking for directions) but might also be more stable and general (such as roommates communicating through a lingua franca) (see Mauranen Reference Mauranen2018: 107–8). Of course, this does not imply that morphosyntactic aspects cannot be investigated in ELF contexts; however, it highlights the necessity of suitable methods and an awareness of the transient nature of ELF when working in the field of ELF grammar. As Pitzl highlights, interactants in ELF situations bring their own individual multilingual resource pools, consisting of ‘all the linguistic resources … [they] have at their disposal’ (Reference Pitzl2016: 298), but they also have access to the shared multilingual resource pool which emerges during the encounter itself and is constantly modified and adapted (Reference Pitzl2018: 34–6; see also Hülmbauer Reference Hülmbauer, Mauranen and Ranta2009: 325). In other words, ELF speakers tend to ‘bridge lingua-cultural boundaries, resulting in code-mixing and foreign language use, and the use of forms or form-function relationships which are non-codified’ (Osimk-Teasdale & Dorn Reference Osimk-Teasdale and Dorn2016: 373).

Thus, rather than merely listing features which seem to be diverging from native speaker ‘standard English’ – whatever this term is supposed to mean – recent research in ELF primarily focuses on the strategies interactants use to establish mutual understanding, that is, on how they ‘effectively employ language in communication by drawing on a set of resources and … strategically use them in interaction’ (Vettorel Reference Vettorel2019: 184). This includes, for instance, accommodation processes in general but also very specific strategies like restructuring or reformulating on the side of the speaker and repair initiations, such as requests for clarification, on the side of the hearer(s) (e.g., Wong Reference Wong2000: 247; Cogo Reference Cogo, Mauranen and Ranta2009; Vettorel Reference Vettorel2019). However, there has been considerably less research on the question why a certain strategy is preferably selected in specific interactional contexts, and, in particular, larger quantitative investigations of this topic are still missing (but see Neumaier Reference Neumaier2023a, Reference Neumaier, Wilson and Westphal2023b, who looks at the interface of conversational patterns and grammar in Southeast Asian interactions).

Although World Englishes research typically focuses on long-term situations of language contact, the question of how grammatical conventions come into being through language use is central to its research agenda, and the methodological approaches it has developed might be fruitfully applied to ELF contexts. When describing the emergence of creoles, Mufwene lists several aspects which seem to influence whether a feature is likely to be selected from a common pool of multilingual repertoires, including its ‘statistical frequency, semantic transparency, regularity, salience, [and] social status of the model speakers’ (Reference Mufwene2008: 19). Mufwene’s concept of the linguistic feature pool strongly resembles the idea of a shared multilingual resource pool advocated in current ELF research (as discussed previously). In fact, as Thompson puts it, ‘in view of the highly diverse interaction environments that embed ELF, we might expect that not only the extent but the capacity for exploitation is also greater – the virtual language, or resource pool, is bigger – than it is in non-lingua-franca contexts’ (Reference Thompson2017: 214). That this is the case has already been illustrated in previous studies on minus-features in ELF conversations, for instance, on copula usage by (South)-East Asian speakers (see Leuckert & Neumaier Reference Leuckert and Neumaier2016).

Accepting that ELF grammar is inherently adaptive and in a process of constant restructuring means, of course, that any investigation of syntactic patterns must be able to take this fluid nature into account. How, indeed, can one analyse something which is influenced not only by various linguistic inputs but also by the context-specific contingencies of the very specific communicative situation, where the focus is – in many cases – not predominantly on grammatical correctness but rather on intelligibility? The concept of ‘emergent grammar’ (Hopper Reference Hopper, Auer and Pfänder2011) as well as the related idea of ‘probabilistic grammar’ put forward by Szmrecsanyi et al. (Reference Szmrecsanyi, Grafmiller, Heller and Röthlisberger2016) have been proposed as ways to account for this complexity and will also form the underlying framework for the present study. These conceptualisations of grammar as something dynamic go together well with the framework developed by Pham et al. (Reference Pham, Leuckert, Dreschler, Götz, Günther, Kircili, Lange, Mycock, Neumaier and 329Rüdiger2024), as they also regard non-canonical syntactic constructions as logical products of language usage and acknowledge their theoretical potential to develop into unmarked constructions in specific contexts. Emergent grammar has been described as ‘an alternative to the standard lexical-item-and-rules model of linguistic description’ (Hopper Reference Hopper, Auer and Pfänder2011: 23). In communication, speakers resort to their individual linguistic feature pool which consists of both individual items but also formulaic constructions and fragments, and the appliance of grammatical rules is subject to ‘different degrees of adaptation to meet syntactic constraints and the requirements of context’ (Widdowson Reference Widdowson1989: 135). Importantly, this does not imply that grammar cannot be investigated systematically; on the contrary, it is regarded as essential to establish a basis of analysis. The concept of emergent grammar acknowledges, however, that

Although Hopper’s statement does not refer to interactions between non-native speakers per se, we suggest that the concept of emergent grammar is well suited to investigate the ‘syntactic borderlands’ researchers enter when dealing with ELF conversations.

Probabilistic grammar models embrace this usage-based approach but also ‘incorporate statistical regularities derived from experience [of the speakers, and] … associate these quantitative patterns not (only) with surface forms or lexical items … but with abstract features or constraints’ (Grafmiller et al. Reference Grafmiller, Szmrecsanyi, Röthlisberger and Heller2018: 3). This gives rise to three predictions:

To date, probabilistic approaches have typically been concerned with grammatical patterns in second-language varieties of English. Heller et al. (Reference Heller, Bernaisch and Th. Gries2017), for instance, compare the use of the genitive in Asian Englishes with British English. Recently, however, researchers have started to acknowledge the enormous potential of probabilistic models for the study of ELF (Deshors Reference Deshors2020), revealing that ‘ELF is not only a discourse-driven and fuzzy phenomenon’ but exhibits underlying – and statistically measurable – systematicity while at the same time ‘ELF users not only react passively to ongoing processes but also contribute to create patterns within the core grammar of English’ (Laitinen Reference Laitinen2020: 440).

14.2.2 Minus-Plurals in ELF

In this chapter, we suggest using a quantitative approach to investigate some of the intra- and extra-linguistic factors influencing the use of a specific non-canonical syntactic construction, the lack of overt morphological plural marking in nouns, in ELF contexts. Examples (1) and (2) illustrate the phenomenon.

Thus, while in (1) the head of the noun phrase (NP), group, is clearly semantically marked for plural by the preceding quantifier many as well as the copula, it is not morphologically marked as such. Similarly, in (2) the head noun coffee also remains morphologically unmarked, even though it is modified by a numeral quantifier indicating plurality – and even though the speaker explicitly marks plural in the preceding NPs (five beers, three beers). In the description of the selected feature, we rely on Rüdiger’s (Reference Rüdiger2019: 48) classification, illustrated in Table 14.1. The main reason for using the term ‘minus-plurals’ instead of a more common one is its relative neutrality, since other labels, such as ‘omission’, ‘lack’, or ‘underuse’, are, to a degree, associated with more prescriptive ideas (consciously or subconsciously). Our decision to focus on minus-plurals is directly connected to our research questions: minus-plurals are a well-known feature of many varieties of English (see, for instance, eWAVE feature 57) and may result from language contact or represent simplification tendencies. In research on Asian ELF, minus-plurals have been reported regularly (e.g., Thompson Reference Thompson2017; Ji Reference Ji2016; Kirkpatrick Reference Kirkpatrick2010).

Table 14.1Plus- and minus-terminology

Source: adapted from Rüdiger (Reference Rüdiger2019: 48). Rüdiger also describes cases of ‘swap’-phenomena, which are not relevant for plural marking in our data.

Table 14.1Long description

The table is divided into 4 columns with the labels Superordinate term, Subordinate terms, Other labels, and example. The data is filled in a row, where the subordinate term has two subdivisions: plus and minus. The data from left to right is filled as follows:For the term non-canonical use:

• The relevant data for plus plural are superfluous items, overuse, and two childrens.

• The relevant data for the minus plural are omission, lack, underuse, and two coffee.

14.3.1 ACE and VOICE

In order to trace minus-plurals in spoken ELF conversations from different contexts, we analysed the files containing non-institutionalised, spontaneous language in the Asian Corpus of English (ACE) and the Vienna-Oxford International Corpus of English (VOICE). ACE and VOICE each represent 1‑million-word corpora of naturally occurring face-to-face interactions of ELF speakers, with VOICE predominantly featuring European and ACE featuring Asian speakers from countries that are part of the Association of Southeast Asian Nations (ASEAN)Footnote ¹ as well as China, Japan, and South Korea. The corpora can be accessed freely via http://corpus.ied.edu.hk/ace/ and https://voice.acdh.oeaw.ac.at/, respectively. ACE has been modelled to match the compilation procedure and structure of VOICE, with both corpora containing conversations in educational contexts (ca. 25%), leisurely contexts (ca. 10%), professional business (ca. 20%), professional organisation (ca. 35%), and professional research/science (ca. 10%). The dataset we used is summarised in Table 14.2, which also gives the figures for the tokens of overtly marked plurals and minus-plurals (see Section 14.3.2 for further details on token extraction and annotation). To reach roughly equal word counts, we extracted conversations from two interactional contexts in ACE (leisure and education) and one interactional context in VOICE (leisure).

Table 14.2Dataset used in the analysis of minus-plurals and overtly marked plurals

Table 14.2Long description

The table is divided into 3 columns with the labels Asian Corpus of English or A C E, and Vienna-Oxford International Corpus of English or V O I C E. The data is filled in 4 rows from left to right as follows:

• The relevant data for word count is c a. 76000 words and c a. 72000 words.

• The relevant data for the number of speakers is 36 and 63.

• The relevant data for overtly marked plurals is 1029 and 917.

• The relevant data for minus-plural is 256 c a. 20% of all cases and 31 c a. 3% of all cases.

It is important to note that we did not normalise frequencies since we focus on the relative proportions of minus-plurals to overtly marked plurals, which means that frequency normalisation would not provide further insights.

ACE and VOICE are compiled with the goal of accessing natural ELF conversations, meaning that L1 speakers of English and speakers at all levels of proficiency are generally included. Consequently, it is important to stress that neither ACE nor VOICE are corpora of learner English. Just like Osimk-Teasdale and Dorn point out in their paper on Part-of-Speech tagging in VOICE, the ‘goal [is] not to tag “errors” or to register degrees of systematicity or conformity in reference to established norms of conventional usage’ (Reference Osimk-Teasdale and Dorn2016: 374) but to find out to what extent variation in plural marking in ELF contexts can be explained by language contact and by relevant interactional strategies.

14.3.2 Token Retrieval and Annotation

After deciding on a dataset for analysis, we manually identified and tagged all relevant constructions by reading through the corpus files and tagging each regular plural noun that is morphologically marked in Standard English, that is, with the inflectional {-s}-ending. In order to be able to make sound statements about the relative frequency of minus-plurals, we tagged both overtly marked plural forms as well as minus-plurals. Focusing on variation this way allows investigating ‘alternate ways of saying “the same” thing’ (Labov Reference Labov1972: 188). We annotated all relevant tokens yielded by the process described above for the predictor variables listed in Table 14.3.

Table 14.3Predictor variables in the statistical analysis

Table 14.3Long description

The table is divided into 2 columns with the labels predictor variable and levels. The data is filled in 7 rows from left to right as follows:Sociolinguistic variables:

• The data for the age group is young, mid, old, and n or a.

• The data for gender is male, female, and n or a.

• The data for Speaker’s L 1 is language code according to I S O 639-3.

• The data for language family is given in Table 14.4. Intra-linguistic variables:

• The data for quantifiers in N P is quantifier or no quantifier.

• The data for animacy of the head noun in N P is animate or inanimate.

• The data for corpus is A C E or V O I C E.

We included four variables that can be described as sociolinguistic in nature. Age is potentially relevant because of apparent-time change (e.g., Chambers Reference Chambers, Chambers, Trudgill and Schilling-Estes2004) that may be visible in the data, that is, there might be differences between generations due to younger speakers adopting new features or features disappearing or changing over time.Footnote ² Gender, in turn, is relevant because female speakers are known as innovators in language development, which means that minus-plurals could possibly be more frequent in their language use. Corpus-linguistic methods have been identified as highly useful to investigate gender in World Englishes contexts as they contribute to ‘obtain[ing] a sociolinguistically and empirically valid picture’ (Bernaisch Reference Bernaisch and Bernaisch2021: 5).

Both speaker L1 and language family are crucial in that they allow us to discuss the potential impact of language contact on occurrences of minus-plurals. Terassa, for instance, found that ‘speakers of HKE [Hong Kong English] might omit the plural suffix because Cantonese does not inflectionally mark its nouns for plural’ (Reference Terassa2017: 139), although she identified low rates of minus-plurals in Singapore English (SgE) and Indian English, whose substratum languages do have inflectional plural marking. Language families may also provide insight, since language areas can have significant effects in statistical modelling (see, for instance, Bentz Reference Bentz2018: 121).

We would like to point out that we used the language codes according to the ISO 639-3, which is organised by the Summer Institute of Linguistics (SIL). The SIL, which also publishes the Ethnologue, has been criticised for its missionary activities and at times questionable language naming practices, but the ISO language codes offer a tool to quickly refer to languages with clearly identifiable codes. The language families, the languages, and the language codes relevant for our data are listed in Table 14.4. We decided to use sub-groups for the Indo-European languages, since we otherwise would have had to exclude language family from the statistical analysis due to an extreme skew towards Indo-European (IE) in the dataset.

Table 14.4Language families and languages (including ISO 639-3 codes) in the analysed ACE and VOICE sections

Table 14.4Long description

The table is divided into 2 columns with the labels Language families, 17 branches, and Languages and language codes I S O 639-3, 31 languages. The data is filled in several rows from left to right as follows:

• The data for the branch Afro-Asiatic or A f r is Maltese, m l t.

• The data for the branch Albanian, A greater than I E,) is Albanian, a l b.

• The data for the branch Lolo-Burmese, B is Burmese, b u r.

• The data for the branch Dravidian, D is Tamil, t a m.

• The data for the branch Finno-Ugric, F is Finnish, f i n.

• The data for the branch Germanic G greater than I E, is Danish, d a n, Dutch or d u t, English or e n g, German, g e r, Norwegian, n o, Swedish, s w e.

• The data for Indo-Aryan, I n greater than I E, is Hindi, h I n.

• The data for Indo-Iranian, I r greater than I E, is Iranian, i r a.

• The data for Japonic, J is Japanese, j a p.

• The data for Korean, K, is Korean k o r.

• The data for Malayo-Polynesian, M is Cebuano c e b, Filipino f i l, Indonesian Malay i n d), Tagalog t g l.

• The data for Romance R greater than I E is Catalan c a t, Italian i t a, Spanish s p a.

• The data for Sinitic S is Cantonese or Yue y u e, Mandarin c m n, Putonghua p u t.

• The data for Slavic S l a greater than I E is Czech c z e, Polish p o l, Serbian s r p.

• The data for Tai T a is Thai t h a.

• The data for Turkic T is Kyrgyz k I r.

• The data for Vietic V is Vietnamese v i e.

In addition to the sociolinguistic variables, we considered the presence or absence of a quantifier in the NP and the animacy of the head noun in the NP as potential intra-linguistic predictors. The role of quantifiers in SgE in relation to plural inflection, for instance, ‘has been examined in a number of accounts … with contradictory findings’ (Terassa Reference Terassa2017: 119). We decided not to include usage frequency of the nouns as a variable (as Terassa Reference Terassa2017 does in her study on plural inflection in Asian Englishes), since our dataset was relatively small while still allowing for inferential modelling. Animacy has been included since ‘human nouns are more likely to have plural marking than non-human (especially inanimate) nouns’ (Haspelmath Reference Haspelmath, Dryer and Haspelmath2013 in WALS) across all languages and animacy may therefore be a relevant predictor.

The final predictor in our model was corpus, that is, if the token occurs in ACE or VOICE. This predictor was included to find out if this essential distinction causes any significant splits in the model and since the two corpora, while compiled with comparability in mind, are different in some fundamental ways. First, they have been compiled in different timeframes – VOICE data were recorded between 2001 and 2007, while data collection for ACE started in 2009, with the corpus being released in 2014. This means that the corpora do not represent language use in their respective regions at identical periods of time. Second, and more importantly, the dominant language ecologies differ in the corpora and, by extension, in the regions they were created in. These differences are accounted for to an extent by the language families and speaker L1s, but the corpora are better suited to represent Asia and Europe at large (although ELF is by nature multilingual and there are no clear-cut boundaries). Third, and again in spite of the intended comparability, the two corpora are two different datasets created by different teams and in different situations.

14.3.3 Statistical Approach: RePrInDT

We subjected the variables described in the previous section to a statistical analysis using conditional inference trees. Tree-based methods, such as conditional inference trees and random forests, have gained in popularity in recent years and their increasing use is symptomatic of what has been called the ‘quantitative turn’ in linguistics (see Kortmann Reference Kortmann2021). Such methods ‘function by repeatedly splitting data sets up in two parts such that the split leads to the best increase in terms of classification accuracy or in terms of some other statistical criterion’ (Gries Reference Gries2020: 618; see also Bernaisch et al. Reference Bernaisch, Th. Gries and Mukherjee2014; Lohmann Reference Lohmann2013; Tagliamonte & Baayen Reference Tagliamonte and Harald Baayen2012). Gries (Reference Gries2020) addresses the widespread conception that tree-based methods are ‘easy to interpret’ even though they offer various pitfalls. Key issues in this context are that ‘there can be patterns in data that make trees underperform considerably when it comes to accuracy, variable importance/parsimony, and effects interpretation’ (Gries Reference Gries2020: 644). An advancement of the conditional inference trees available, for instance, as part of R’s partykit package (Hothorn & Zeileis Reference Hothorn and Zeileis2015), is Weihs and Buschfeld’s (Reference Weihs and Buschfeld2021) repeated undersampling in PrInDT (RePrInDT), which is itself a further developed version of PrInDT. RePrInDT improves conditional inference trees in R by undersampling the larger class in the dependent variable (in our case: 15% of overtly marked plurals) in order to achieve better prediction rates in the trees (cf. Weihs & Buschfeld Reference Weihs and Buschfeld2021: 5). This function is particularly useful in such cases as ours, in which one class is (much) bigger than the other and, as a result, prediction would always favour the larger class.

The script runs numerous trees and provides information on all results as well as the three trees with the highest accuracy. Furthermore, a big advantage is the in-built calculation and plotting of the balanced accuracy, that is, the accuracy for both the larger and smaller class in the predictor variable. We decided to work with RePrInDT since, just like in Weihs and Buschfeld’s (Reference Weihs and Buschfeld2021) case study on zero and realised subject pronouns in SgE, there is a strong imbalance in our dataset due to the much higher frequencies of overtly marked plural forms. It is important to note that, at an earlier stage, we had included speaker as a variable in the tree but decided to discard it since it did not appear as an important predictor variable in any of the best-performing trees.