Infinitival and Gerund-Participial Catenative Complement Constructions in English World-Wide

Abstract

Previous research on non-finite catenative complementation (for example, start Ving/to V; force NP into Ving/to V) has largely been restricted to BrE and/or AmE. The present study seeks to expand the regional coverage of such research by analysing a set of catenative constructions in two large web-derived corpora, GloWbE and NOW, both of which comprise 20 subcorpora representing different national varieties of English. The implications of the findings for such diachronically relevant phenomena as colloquialisation and grammaticalisation are considered. For example, the dominance of bare infinitivals over to infinitivals with catenative help is suggestive of auxiliarisation, an interpretation supported by the semantically bleached sense of generalised causation associated with help, and historical evidence of support for the bare-infinitival variant in colloquial registers. Notable findings include American English epicentrality—and possibly hypercentrality—in many of the results, with Canadian English and Philippine English in particular sharing the American aversion to from-less “prevent NP Ving” and “help to V”; the occasional conservative tendency of the Outer Circle varieties to resist diachronic trends associated with the reference varieties (such as the rise of “fear Ving” at the expense of “fear to V”); and high scores for the African Englishes, suggested to be attributable to the popularity of “serial verb” constructions in a number of African languages.

1. Introduction

Almost two decades ago, Mukherjee and Hoffmann (2006, p. 148) suggested that verb complementation is an under-researched area in studies of regional variability in world Englishes. Today, this claim remains valid, and particularly so in the case of verb complementation involving subordinate clauses. The present study explores the use of non-finite clauses serving as complements to so-called “catenative” verbs in twenty national varieties of English. Catenative verbs are used in potentially recursive constructions yielding chain-like sequences where each non-finite clause is a complement to a (catenative) verb, as in (1). The subcorpus abbreviation “GH” used in this example is, along with those in all following corpus examples, spelt out and listed in Section 3 below (see

Table 1. Classification and word count of the twenty regional varieties in GloWbE.

WE Classification		Country Subcorpus	Variety	Word Count *
Inner Circle	America	US	American English (AmE)	386,809,355	1,239,817,686
		CA	Canadian English (CanE)	134,765,381
	Europe	GB	British English (BrE)	387,615,074
		IE	Irish English (IrE)	101,029,231
	Oceania	AU	Australian English (AusE)	148,208,169
		NZ	New Zealand English (NZE)	81,390,476
Outer Circle	South Asia (“SA”)	IN	Indian English (IndE)	96,430,888	234,039,410	645,815,287
		LK	Sri Lankan English (LnkE)	46,583,115
		PK	Pakistani English (PakE)	51,367,152
		BD	Bangladeshi English (BDE)	39,658,255
	South-East Asia (“SEA”)	SG	Singaporean English (SgE)	42,974,705	169,095,257
		MY	Malaysian English (MalE)	42,420,168
		PH	Philippine English (PhilE)	43,250,093
		HK	Hong Kong English (HKE)	40,450,291
	Africa (“Afr”)	ZA	South African English (SAfE)	45,364,498	203,016,954
		NG	Nigerian English (NigE)	42,646,098
		GH	Ghanaian English (GhE)	38,768,231
		KE	Kenyan English (KenE)	41,069,085
		TZ	Tanzanian English (TanE)	35,169,042
	Caribbean (“Carib”)	JM	Jamaican English (JamE)	39,663,666	39,663,666
TOTAL					1,885,632,973

* Frequencies are as in https://www.english-corpora.org/glowbe_corpus.asp. URL accessed on 24 December 2024.

).

(1)

All is Love I promise to keep trying to listen to that stillness within. (GH)

Following Huddleston and Pullum (2002, pp. 1176–1177), I shall use the term “catenative complement” to refer to a distinctive type of complement, one that is not covered by other types such as object or predicative, and one that is realised exclusively by non-finite clauses.

There are four kinds of catenative complements—to-infinitival, bare-infinitival, gerund-participial, and past-participial—exemplified in (2) and (3) below.1 The examples in (2) all involve “simple” complementation, the catenative complement in each case being the sole complement of the catenative verb, while those in (3) involve “complex” complementation, with more than one complement (the catenative complement in each case here following an NP object complement).

(2)	a.	I really can’t bear to see them go when they get old (SG) [To-infinitival];
	b.	Most days we helped prepare the food for the children (KE) [Bare-infinitival];
	c.	We do not stop playing because we are old (IN) [Gerund-participial];
	d.	then i got banned from channel for 1 day (SG) [Past-participial];
(3)	a.	I can’t bear him to touch me because I don’t feel ‘complete’ (GB)
		[To-infinitival];
	b.	His dad came rushing down to help him make it to the finish line (PH) [Bare
		-infinitival];
	c.	We do not stop them going to the mosque (IN) [Gerund-participial];
	d.	before the BBC got them banned (GB) [Past-participial].

There is much diversity and complexity in this area of grammar, with more than the merely eight possible complementation patterns with catenative verbs that are exemplified in (2) and (3). For example, an alternative to (3c) is that with a prepositional from complement (“stop them from going to the mosque”). Comprehensive lists of catenative verbs along with the clausal complements which they license are available in Biber et al. (1999, pp. 693–759), Huddleston and Pullum (2002, pp. 1225–1244), and Algeo (2006).

While it is not possible to categorically specify the type(s) of non-finite complement that a particular catenative verb will select, the selection is not entirely random. Verbs with similar meanings tend to enter into similar complementation constructions. Examples include the selection of complex complementation with object and to- infinitive by verbs of coercion (compel, force, cause, allow, etc.) and verbs of saying (warn, urge, instruct, invite, etc.) as in (4a) and (4b), of prepositional phrase complement plus to- infinitival by verbs expressing reliance (rely, depend, appeal, bank, etc.) as in (4c), and of object plus gerund-participial by verbs of depiction (depict, picture, portray, etc.) as in (4d):

(4)	a.	It will shock them and compel them to have a new look at their social
		behaviour (PK);
	b.	The town-crier goes ahead to warn people to put out all lights (GH);
	c.	They depend on others to be more dependable than they are (JM);
	d.	Sigiriya frescoes depict women wearing the cloth gracefully draped like a
		dhoti (LK).

In cases where a catenative verb licenses more than one type of complement construction, there is often a semantic difference, one that may be quite elusive. The difference may be related to the general modal distinction between potentiality on the one hand (in the case of infinitivals) and actuality on the other (in the case of gerund participials). There is a historical motivation for the two meanings in question (see Huddleston & Pullum, 2002, p. 1241). The infinitival marker to derives from the preposition to, whose goal-oriented meaning is rooted in the notion of direction, and by extension to unactualised future situations (with verbs such as strive, consent, promise, threaten, and the like). By contrast, gerund participles have a historical connection with nominal constructions that depict actual events (compare he finished swimming with he finished his swim; she hates working with she hates work). This distinction elucidates the difference between (5a), where we understand that there was an earlier commitment to put the fiddle in the car, and that the speaker’s failure to act accordingly is projected forward from this commitment, and by contrast (5b), where what is forgotten is an actual rather than merely potential action, namely the insertion of the addressee’s name by the speaker.

(5)	a.	I forgot to put the fiddle in the car (IE);
	b.	I forgot putting your name there (US).

The potential versus actual distinction is, however, more commonly not discernible, or barely so. For example, in (6) and (7), the to infinitive and present participle are readily interchangeable without any evident change in meaning.

(6)	a.	I won’t bother to repeat it here (GB);
	b.	I won’t bother repeating what others have already written (IN);
(7)	a.	Mobil Oil started talking to us locally (IN);
	b.	Firmi and I somehow started to talk again (SG).

This study is limited to cases such as those exemplified in (5), (6) and (7), where a catenative verb licenses more than one type of non-finite complement construction, with the patterns of choice associated with each construction identified across the twenty national varieties represented in the mega-corpora GloWbE (the Global Web-based Corpus of English) and NOW (the News on the Web corpus): see further Section 3. By providing insights into regional variation in catenative constructions in English world-wide, and into some of the diachronic implications thereof, the study aims to contribute to the expanding body of research on morphosyntactic variation in space and time in English world-wide. Most previous studies of catenative constructions have focused on data from British English (BrE) and/or American English (AmE), the two influential and much-studied “reference” varieties that are dubbed respectively “hypercentral” and “supercentral” by de Swaan (2001) and Mair (2013). These include Algeo (2006); Rudanko (2005, 2006, 2017); Rudanko and Luodes (2005); and Vosberg (2009). Exceptions include studies of Australian English (AusE) and New Zealand English (NZE) by Mair (2009), Collins (2015), and Rickman and Rudanko (2018); of Hong Kong English (HKE) by Deshors (2015); of Indian English (IndE) by García Castro (2019); and of sets of varieties by Deshors and Gries (2016) and Romasanta (2017, 2019, 2021, 2022, 2023).2

This paper circumscribes the locus of variation to those patterns of catenative complementation that are found widely throughout the English-speaking world. Accordingly, I have excluded more regionally restricted patterns such as the substrate-shaped use of bare-infinitival complements instead of standard gerund participials in creole-influenced JamE as in (8), the use of to infinitivals for standard bare infinitivals in some African varieties as in (9), bare infinitivals for standard to infinitivals in some South Asian varieties as in (10), and to infinitivals with prepositional for in some South East Asian varieties as in (11).

(8)		Them time dey I was a half believer in the spiritual ting. Things start get worse when I find myself not even wanting to go ova the yard (JM).
(9)		We should make them to assist the police during the day (GH).
(10)		I mentioned him about our desire and asked him give a portion to us from their
		delicious meal being cooked (BD).
(11)		If you wish for to make a striking statement (MY).

The structure of the rest of the paper is as follows. In Section 2, I consider the diachronic development of non-finite catenative complement constructions. Section 3 introduces the two mega web-based corpora used and data extraction procedures. Section 4 presents a number of individual case studies. Section 5 is devoted to concluding observations.

2. Diachronic Background

The nature and distribution of non-finite catenative complement constructions in present-day English are by-products of extensive changes that have occurred throughout the history of English. During the Middle English period to infinitivals (as in [They told him] to leave) began to compete with finite that clauses in subjunctive constructions (as in [They told him] that he should leave), with the former spreading widely and eventually becoming the default option (Los, 2015). In a further development, the to infinitival itself began to be challenged by the rise of gerund participials. This scenario had its origins in the decay of the Old English case system, which paved the way for the increasing use of prepositions to signal grammatical relationships (Fanego, 2004), and in turn for gerunds to be used as prepositional complements (a function not possible for to infinitivals). However, it was not until the early eighteenth century that there was a sharp spike in the frequency of gerund participials, prompted by their use with an increasing number of catenative verbs. This profound reconfiguration of the verb complementation system is commonly referred to as the “Great Complement Shift” (Rohdenburg, 2006; Vosberg, 2009).

A number of diachronic studies concentrating on the past half century have provided robust evidence that in AmE and BrE the competition between non-finite catenative complements has continued to the present day (e.g., Mair, 2009; Leech et al., 2009). In this paper, I identify, analyse and discuss patterns of non-finite catenative complementation that are in some cases parallel to, and in others divergent from, these well-attested developments. The orientation of the present paper is primarily synchronic, the only directly real-time component being that facilitated by the short time span offered in NOW, from 2010 to the present day (see Section 3).

3. The Data

The data for the present study are captured from two large web-derived corpora, GloWbE and NOW. GloWbE comprises 1.9 billion words of texts taken from English websites. Table 1 presents the labels used for the twenty countries represented in GloWbE, as used henceforth in this paper, along with their associated English variety labels. The twenty GloWbE subcorpora represent six high-contact L1 varieties and fourteen indigenised L2 varieties, these two categories corresponding roughly to that between Inner Circle (IC) and Outer Circle (OC) Englishes, respectively, in Kachru (1985). The six L1/IC varieties are here subclassified into three regional groups ((North) American, European, and Oceanian) and the fourteen OC Englishes into four regional groups (South Asian, South-East Asian, African, and Caribbean). Needless to say, the uneven distribution of subcorpora sizes—ranging in size from 387,615,074 words for GB to 35,169,042 words for TZ—necessitates the use of normalised frequencies in our intervarietal comparisons. While the primary focus of the paper is on the “New Englishes” of the OC, comparisons will systematically be drawn with the findings for the IC varieties.

All the texts in GloWbE originate from the internet. They are categorised into two sections, “blogs”, and “general” (newspapers, magazines, company websites, and so on), and offer a mixture of informality and formality. The tenor of many of the GloWbE blogs is personal, involved and opinionated, and they feature a high density of linguistic features associated with oral communication (Biber & Egbert, 2018). By contrast, the tenor of the general texts tends more towards the formal end of the informality–formality spectrum.

The NOW corpus contains 20 billion words of data from web-derived newspapers and magazines from the same twenty countries represented in GloWbE, from 2010 to the present time. The online platform allows searches either by country or by date. Web newspapers are professionally written and edited and tend to adhere to an accepted level of quality in maintaining the standards of the genre. Thus, the availability of NOW alongside GloWbE provides us with the opportunity to draw comparisons between the more informal mix of texts in GloWbE and the less informal news reportage texts of NOW. There being insufficient space to present full details of NOW searches to match those for GloWbE, when making GloWbE vs. NOW comparisons I shall simply select relevant frequencies from the NOW findings. Occasional references are also made, where relevant, to frequencies extracted from a third corpus, COHA (Corpus of Historical American English), which comprises 475 million words of AmE texts representing a balance of genres, from 1820 to 2019.

In addition to the large number of national varieties of English represented, the massive size of the GloWbE and NOW corpora is invaluable for the study of low-frequency items and constructions. Admittedly, mega web-derived corpora cannot compete with smaller corpora such as those belonging to the International Corpus of English (ICE) and the “Brown-family” of corpora, in studies where rigorously controlled generic representation and foolproof identification of text origins are essential requisites. However, the generally informal text types in GloWbE make it a particularly suitable resource for a study of historically changeable categories, with such phenomena as colloquialisation and grammaticalisation known to thrive in more informal language (Collins & Yao, 2013). The fact that grammatical changes tend to proceed more rapidly in speech than in the typically more conservative medium of writing is suggested by the findings of Leech et al.’s (2009) investigation of grammatical changes in written and spoken BrE and AmE between the 1960s and the 1990s. A caveat that must be entered here is that such changes cannot be assumed to be directly relevant to developments in the New Englishes. As a prelude to this study, I ran some preliminary searches in GloWbE and NOW for a set of orality/colloquiality-oriented items to determine their distribution in the two corpora: see

Table 2. Orality-sensitive items: GloWbE vs. NOW frequencies.

Category	Form Class	GloWbE (Frequency pmw)	NOW (Frequency pmw)	GloWbE vs. NOW Ratio
Word class	Modal verb	15,231	1204	12.6
	Modal can	3134	2249	1.4
	Auxiliary do	6267	3848	1.6
	Article	98,080	27,684	3.5
	Determiner	31,393	24,755	1.3
	Personal pronoun	50,765	40,475	1.3
	Indefinite pronoun	3602	2167	1.7
	Negator	8688	6508	1.3
Contractions	’d	344	218	1.6
	’ll	439	334	1.3
	’m	744	410	1.8
Punctuation	!	2121	538	3.9
	?	4134	2241	1.8
	…	1387	601	2.3

. The items selected as bellwethers of change were of three types: grammatical word classes, contractions and punctuation marks. The word classes were those whose frequencies in speech in Biber at al. (1999) were significantly greater than those in fiction, news, and academic writing. As Table 2 shows, they were all more frequent in GloWbE than in NOW (by ratios ranging from 12.6:1 for modal verbs to 1.3:1 for determiners, personal pronouns and negators. Contractions—represented here by ’d, ’ll, and ’m—are referred to by Leech et al. (2009, p. 240) as “the paradigm case of colloquialization”, a claim confirmed by the findings of Collins and Yao (2013, 2018). These three contractions were approximately 63% more frequent in GloWbE than NOW. Finally, three punctuation indicators in Table 2—question mark, exclamation mark, and ellipsis—manifest aspects of the interactivity that is an essential feature of conversational speech. They were approximately 44% more frequent in GloWbE than NOW. I hypothesise that if a feature that is known to be on the rise historically yields a relatively high frequency of occurrence in NOW vis-à-vis GloWbE, this may indicate an advanced stage of change wherein it has spread from spoken (or speech-like) genres into a more formal written genre.

Searches of the type “HELP _v?i” and “HELP to _v?i” were conducted in GloWbE and NOW using the online Brigham Young University online platform. For information on this platform, see https://21centurytext.wordpress.com/introducing-the-1-9-billion-word-global-web-based-english-corpus-glowbe/, accessed on 16 December 2024. In some cases where complete searches were not possible, it was necessary to restrict them. For example, in the case of complex constructions, it was not possible to capture all the possible expressions that could manifest the direct object NP (since the corpora are POS-tagged but not syntactically parsed), so searches were limited to personal pronouns instead of full NPs. For example, for constructions of the type “force him into leaving”, it was not possible capture relevant NP tokens using the query “VERB+ * into _v?g”—due to mishits of the type “had gone into hiding”—so “VERB+ _pp into _v?g” was used. The pmw frequencies yielded by the searches for the individual varieties, for the regional groupings thereof, and for the IC vs. OC were then compared across the alternants for each construction type using ratios and percentages, in the spirit of recent “onomasiological” research (e.g., Collins, 2023a, 2023b).

4. Individual Case Studies

This section investigates four types of catenative complement constructions: help NP (to) V; stop/prevent NP (from) Ving; V NP into Ving/to V; V Ving/to V.

4.1. Help NP (to) V

Catenative help—here restricted to that used in complex constructions—licenses both bare-infinitival and to-infinitival complements, as exemplified in (12):

(12)	a.	I would have helped them get a little boost (KE);
	b.	The country helped them to get a job (IN).

Diachronic developments in these two variants are investigated by Leech et al. (2009, p. 188), using parallel British and American corpora of the 1960s and the 1990s. Their study indicated that AmE favours the bare infinitive more than BrE, and that this variant is growing in popularity in both varieties at an equally strong pace. By contrast, the to-infinitive variant, whose popularity is undergoing a mild decline in both varieties, is favoured more in BrE. Using spoken corpus data, Leech et al. (2009) suggest that the choice of a bare-infinitival or to-infinitival complement after help is mainly stylistically motivated, the former preferred in informal styles and the latter in formal styles. Possible iconic and semantic motivations have also been suggested for this alternation. For example, Dixon (2005, p. 201) proposes that the to infinitive encodes more indirect causation or support than does the bare infinitive, claiming that John helped Mary to eat the pudding suggests that he did so in a roundabout way, for example ‘by guiding the spoon to her mouth’, while John helped Mary eat the pudding actually means that he himself ate part of it.

Mair’s (2009) study yields similar findings to those of Leech et al., except that he extends the regional coverage to include AusE and NZE, via comparisons of FLOB and Frown with the Australian Corpus of English (“ACE”) and the Wellington Written Corpus of NZE (“WWCNZE”). In all four corpora, the bare infinitive is more frequent than the to infinitive, but proportionately more so in AmE than the others (prompting Mair to refer to the bare infinitive with help as a “statistical Americanism”). Mair also looked at spoken English (for which no comparable AmE data were available), reporting that AusE and NZE were less conservative than BrE.

Collins’s (2015) corpus-based diachronic study of AusE, BrE, and AmE over the 19th and 20th centuries confirms the leading role of AmE in the complementary rise of help V and fall of help to V, developments that are not apparent in the British data until the second half of the 20th century. AusE is found to lie in between the innovative American pattern and the conservative British, with a level of support for the bare infinitive in the 20th century that is not far behind that in AmE.

Two initial observations arise from these studies. Firstly, in view of the considerably greater popularity in speech of the bare infinitive with help than that of the to infinitive, we may surmise that colloquialisation is playing a role in the rise of the former (Leech et al., 2009, pp. 190–192; Mair, 2009, p. 273; Collins, 2015, p. 32). Secondly, as Mair (2009) further argues, the dramatic rise of the bare-infinitival variant is a case of incipient grammaticalisation. More specifically, the absence of the to—a property associated with canonical modal auxiliaries—is suggestive of auxiliarisation. Further evidence of grammaticalisation/auxiliarisation is to be found in the semantic bleaching that help + bare appears to be undergoing, with the notion of assistance giving way to merely generalised causation (similar to the sense of make and have as in I made/had her proofread the manuscript).

Turning now to the findings of the present study presented in

Table 3. The “help NP V” construction in GloWbE: frequencies pmw.

Total
92.00 (75.5%)
Inner Circle						Outer Circle
90.66 (79.3%)						93.58 (74.2%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
101.80		78.14		92.06		85.28				105.14				92.74					70.73
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
90.07	113.52	80.07	76.20	88.66	95.45	110.75	56.16	70.75	103.46	107.09	88.66	118.52	106.29	94.24	83.43	87.86	107.57	90.59	70.63

NOTE: Bracketed percentages, as in subsequent tables, represent proportionalities of the alternants for the construction in question (in this case “help NP (to) V” in Table 3 and Table 4).

and

Table 4. The “help NP to V” construction in GloWbE: frequencies pmw.

Total
29.91 (24.5%)
Inner Circle						Outer Circle
23.60 (20.7%)						32.61 (25.8%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
16.99		28.65		25.16		35.30				25.47				31.03					36.16
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
14.15	19.83	26.06	31.24	24.28	26.03	41.44	35.57	23.44	40.75	30.23	31.00	26.52	36.37	31.06	28.63	28.37	29.54	35.67	36.16

(for which the search queries used were “HELP _pp _v?i” and “HELP _pp to _v?i”, respectively), we find confirmation of the ascendancy of bare-infinitival over to-infinitival complements with transitive help. Interestingly, the relevant ratio in GloWbE (3.08:1) is overwhelmingly surpassed by that in NOW (11.50:1). Note that in Table 3 and Table 4, and all subsequent tables, the proportionalities expressed here as ratios are expressed as percentages (in brackets) for the totals and IC/OC subgroups of each alternant. It may be surmised that, across the twenty varieties in the two corpora, the bare-infinitival variant, having undergone a colloquialisation-driven increase, has become firmly established in the more formal written genre of news reportage.

As Table 3 shows, the findings for the two IC reference varieties (AmE 90.07; BrE 80.07) are in line with the findings of previous studies, reflecting the leading role played by AmE in the rise of the bare-infinitival variant. However, it would appear that AmE is not leading the way in English world-wide, with CanE and NZE in the IC having a higher frequency, along with IndE, BDE, Sing, PhilE, HKE, SthAfrE, KenE and TanzE in the OC. What this perhaps suggests is that, historically, while Americanisation might have exerted hypercentral influence on other varieties, a number of them have caught up with AmE, and in some cases overtaken it. In fact, it is notable that the varieties with the highest frequencies are specifically those that historically have been most strongly subject to AmE influence: CanE overall in the IC, PhilE overall in the OC, and SingE in SEA (see further Deterding, 2007). The strong results for AmE and PhilE in GloWbE were similarly attested in NOW.

Turning to the to-infinitival variant, Table 4 suggests that—if we can take the frequencies in this table, and Table 3, to reflect diachronic trends—its decline is more advanced in the IC than the OC, and particularly in AmE whose dispreference for the to variant is almost twice as strong as that of BrE. A similar dispreference, albeit weaker than that found in the IC, is displayed in the OC by PakE and PhilE relative to the other varieties. The most conservative IC group is Europe, and—aside from Carib—SA in the OC. A similar pattern of (dis)preferences was observed in NOW.

4.2. Stop/Prevent NP (from) Ving

“Preventive” catenative verbs such as stop, prevent, bar, block, forbid, and inhibit are used with a gerund-participial complement, either without the preposition from as in the (a) examples below or with from as in the (b) examples. In this study stop and prevent are selected for analysis on the grounds that, as preliminary searches of GloWbE confirm, they are by far the most commonly used verbs in this construction.

(13)	a.	Why would I stop her doing something she loves (GB);
	b.	how can we stop her from doing this (ZA);
(14)	a.	They believe this will prevent them from being killed (GH);
	b.	we Kiwis are free to take action when required to prevent others being killed
		(NZ).

According to Leech et al. (2009, p. 193), the from-less variant was attested in AmE as well as BrE in the 18th and 19th centuries. However, in the 20th century its popularity declined in AmE, at the same time as increasing in BrE, to such an extent that it has virtually become a Briticism.

The distribution of prevent in Collins’s (2015) study largely supports this historical picture. He found that the from-less version was largely spurned in AmE, in contrast with BrE where its percentage rose from 45.7% in the first half of the 19th century to 72.7% in the second half of the 20th century. Meanwhile, AusE was found to diverge from BrE, its steadily decreasing endorsement of the from-less version—possibly influenced by the American dispreference for this variant—contrasting with the continuing British endorsement of it.

4.2.1. Stop NP (from) Ving

The overall frequencies presented in

Table 5. The “stop NP Ving” construction in GloWbE: frequencies pmw.

Total
4.25 (37.2%)
Inner Circle						Outer Circle
5.37 (44.3%)						1.56 (17.5%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
0.91		8.93		6.27		1.54				1.44				1.76					1.04
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
1.17	0.65	11.47	6.39	6.17	6.36	1.53	2.15	1.50	0.99	1.12	2.19	0.92	1.53	2.12	1.50	1.75	1.68	1.76	1.04

and

Table 6. The “stop NP from Ving” construction in GloWbE: frequencies pmw.

Total
7.17 (62.8%)
Inner Circle						Outer Circle
6.76 (55.7%)						7.34 (82.5%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
6.46		6.42		7.41		6.52				8.22				7.61					5.71
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
7.24	5.68	7.06	5.78	8.56	6.25	7.30	3.65	9.95	5.17	9.94	9.97	8.21	4.77	6.41	12.08	7.35	6.99	5.23	5.71

below reveal the from variant with stop to be almost twice as popular as the from-less variant (7.17:4.25, or 1.69:1). The frequencies in Table 5 largely support the historical picture presented above, with BrE strongly outscoring AmE in GloWbE (by a ratio of 9.8:1), a situation that was found to be even more pronounced in NOW (18.0:1). AusE and NZE align themselves with neither the American nor the European subgroups. Notably small frequencies are recorded in the OC by PhilE and JamE, in line with the situation in AmE. The comparative smallness of the overall GloWbE frequency (4.25 pmw) is surpassed by that for NOW (1.47) where it would appear to be out of keeping with the relative formality of the genre of news reporting.

Table 6 suggests that AmE may have lost the dominance with the from variant that diachronic studies suggest it enjoyed in the 20th century. In GloWbE, AmE has been overtaken by AusE in the IC, and by IndE, PakE, SingE, MalE, PhilE, NigE, and GhanE in the OC. The situation is similar in NOW where the AmE frequency is smaller than those of BrE, AusE, NZE, IndE, PakE, BDE, SingE, MalE, PhilE, SthAfrE, NigE, GhanE, and KenE. Yet the influence of the AmE preference for the from variant would appear to linger on, with the OC varieties in particular supporting the from-less variant strongly, and even BrE doing so (despite its continuing and largely unshared support for this variant as well).

Comparison of Table 5 and Table 6 reveals some striking differences between the distribution of variants in the IC and the OC. The from variant is only mildly more frequent than the from-less variant in the IC (GloWbE: 1.3:1; NOW: 1.2:1), by comparison with the OC (GloWbE: 4.7:1; NOW: 5.2:1). The strong predilection for the from variant in the OC perhaps points to the strength of AmE’s influence in the OC in the early 21st century. There may of course also be other factors at play in the OC results—such as L1 influence, internal variety-specific patterns of evolution, and colonial lag—which it would be desirable for future researchers to explore.

4.2.2. Prevent NP (from) Ving

Prevent has a similar distribution to the semantically similar stop. As

Table 7. The “prevent NP Ving” construction in GloWbE: frequencies pmw.

Total
11.36 (61.3%)
Inner Circle						Outer Circle
1.99 (17.2%)						0.77 (6.9%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
0.50		3.07		2.41		0.58				0.78				0.90					0.81
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
0.57	0.43	3.35	2.79	2.44	2.38	0.60	2.58	0.84	0.66	0.51	0.94	0.49	1.19	1.45	0.52	0.72	0.78	1.02	0.81

indicates, the results for prevent—as for stop—support ascription of the title “Briticism” to the from-less variant, with BrE evidencing the strongest frequency in both corpora and, as with stop, AmE (along with PhilE) showing a strong dispreference for this variant. Furthermore, as with stop, the from-less variant with prevent is more common in the IC than the OC, suggesting perhaps the prevalence of AmE’s influence on the latter.

Finally, consider the from variant with prevent. As

Table 8. The “prevent NP from Ving” construction in GloWbE: frequencies pmw.

Total
7.17 (38.7%)
Inner Circle						Outer Circle
9.56 (82.8%)						10.31 (93.1%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
10.73		8.78		9.20		11.48				10.10				9.33					11.30
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
11.03	10.43	8.94	8.62	8.45	9.94	10.04	11.81	14.70	9.37	10.47	10.71	11.63	7.59	9.41	10.25	11.97	7.57	7.45	11.30

shows, AmE is dominant in the IC. Furthermore, by contrast with the from-less variant, the OC outscores the IC (by a ratio far exceeding that for stop, 13.4:1, in both GloWbE and NOW), a finding that it is tempting to attribute to the strength of AmE’s influence in the OC in the past decade or so.

4.3. V NP into Ving/to V

Verbs such force, fool, will, pressure, push, manipulate, coerce, bribe, tempt, coax, and manipulate, which denote various types of coercion by force or pressure, can be used transitively with either a prepositional phrase complement (with into + gerundial participial) as in (15a), or a to-infinitival complement as in (15b). See further Rudanko (2005) and Kim and Davies (2015).

(15)	a.	i was really angry and i tried to force them into going to the police (US);
	b.	I had to force them to go home at night (HK).

Is there a semantic distinction here? Cognitively, the actual distinction that has previously been suggested as applicable to a number of gerund-participial versus to-infinitival alternations is reinforced in the case of the into construction by the sense of containment associated with the preposition into, which in turn contributes a strong sense of immediacy to the actuality that distinguishes it from its to-infinitival counterpart.

The into construction is noted to be particularly innovative by Rickman (2024) who, in his study of NZE, attests to the use of hard-talk, peer pressure and turbo-shock as creative catenatives in the construction. The validity of Rickman’s claim is suggested by a random search by the present author for expressions used as coercive verbs in the into construction in just one GloWbE subcorpus (NG). The resultant shortlist points to a high level of open-ended creativity: sucker, groom, shame, hoodwink, church, blind, hypnotize, mesmerise, brainwash, rechannel, contort, panic, bait, scam, demoralize, stampede.

For the into construction, the query “VERB+ _pp into _v?g” was used to capture tokens in GloWbE and NOW (with the direct object limited to personal pronouns). However, for the to-infinitival construction, the query “VERB+ _pp to VERB+” does not restrict the catenative verb to coercives (with non-coercive verbs dominating the output, such as request, need, want, like, expect, pay, help, trust, and send). Unfortunately, this search routine difference means that no direct quantitative comparisons between the into Ving and to-infinitival constructions could be made.

The first thing to notice is that the overall frequency for the into construction in GloWbE (4.15 pmw), as reported in

Table 9. The “V NP into Ving” construction in GloWbE: frequencies pmw.

Total
4.15
Inner Circle						Outer Circle
4.67						3.93
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
4.94		4.52		4.55		3.18				3.90				4.62					3.56
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
6.04	3.84	4.87	4.17	4.63	4.46	3.52	2.53	3.64	3.01	5.21	3.82	3.79	2.79	4.32	6.31	4.95	4.04	3.50	3.56

, is almost twice as common as that for NOW (2.34 pmw), as detailed in

Table 10. The “V NP into Ving” construction in NOW: frequencies pmw.

Total
2.22
Inner Circle						Outer Circle
2.26						2.21
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
2.07		2.38		2.33		1.64				2.04				2.88					1.89
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
2.34	1.80	2.64	2.11	2.37	2.28	1.73	1.26	1.74	1.81	1.96	1.83	1.81	2.56	2.65	3.84	3.22	3.04	1.63	1.89

. This is an unsurprising finding, given that its historical rise and association with creative innovation will be likely to correlate with currency in informal/colloquial genres.3

The results for the IC, not only in GloWbE but also in NOW, are suggestive of epicentral influence, with AmE strongly outscoring CanE, BrE outscoring IrE, and AusE outscoring NZE. In the results for the OC, in both GloWbE and NOW, perhaps the most remarkable finding was the extremely high score for Africa, a finding that is comparable to those for other constructions in the study (“stop NP Ving”, “prevent NP Ving”, “start Ving”, “fear Ving”, and “fear to V”). Curiously, similarly high African scores are not reported for a number of GloWbE studies on a range of grammatical phenomena conducted by the present author (Collins, 2023a, 2023b). A plausible explanation for this finding is to be found in the popularity in a number of African languages—including pidgins and creoles—of so-called “serial verb” constructions. Verb serialisation, or “stacking”, is a syntactic phenomenon in which two or more verbs are strung together in a single clause, similar in various respects to catenative complementation in English. It is a common feature of many African languages, and in particular Western African languages (see Aikhenvald & Dixon, 2006; Haspelmath, 2016), so it is not surprising that NigE and GhE often have the two highest frequencies in the African subgroup within GloWbE (q.v. “stop NP from Ving”, “prevent NP from Ving”, “V NP into Ving”). Serial verb constructions are a feature of languages such as Igbo and Esan in Nigeria and of Dagaare, Ewe and Akan in Ghana. They are also common in Nigerian Pidgin and Ghanaian Pidgin (for example, À ték bág kári gàri kóm rích háws chop, “I brought gari home in a bag to eat” (F152 Nigerian Pidgin: https://ewave-atlas.org/valuesets/42-152. URL accessed on 13 December 2024); ì go make … kɔfi … kam giv mì se mek à nak àm smɔ nɔ, “He went to make coffee for me, came back and said I should drink it in sips” (F151 Ghanaian Pidgin: https://ewave-atlas.org/valuesets/40-152, accessed on 13 December 2024)). It is conceivable that, over time, serial verb-promoted catenative constructions may have diffused from the West African Englishes of Africa to regionally disparate Englishes in Eastern and Southern Africa.

As a final observation, I suggest that the similarity of the findings across the two corpora may be explained in terms of the tendency noted in previous studies for OC varieties to exhibit less stylistic variability across formal and informal contexts than IC varieties, for example, in JamE (Shields, 1989), in PhilE (Gonzalez, 2005), in NigE (Unuabonah & Gut, 2018), and in SingE, TanzE, JamE, GhanE, and MalE (Mazzon, 2019). A possible influencing factor here is the tendency in OC countries for English to be learned primarily in the formal context of the classroom.

4.4. V Ving/to V

A large number of catenative verbs accept both gerundial-participial and to-infinitival complements. They can be grouped into at least four semantic subclasses—including aspectuality, emotional states, intention, and conation—and some of these have further subclasses. Aspectuality catenatives may be subdivided into those that indicate the beginning of an activity such as start in (16), its continuation such as continue in (17), or its cessation such as cease in (18). Catenatives expressing emotional states may be either negative, such as fear, dread, and hate as in (19), or positive, such as like and love as in (20). Catenatives that express intention include intend, propose, and plan, as in (21). Conation catenatives include try and attempt, as in (22).

(16)	a.	There are also affiliates who have started to sell these supplements (PK);
	b.	Several other local retailers have started selling inexpensive cloth bags (JM);
(17)	a.	The cost of living, particularly housing, continues to rise (IE);
	b.	China continues rising as a dominant and sometimes belligerent global
		economic player (US);
(18)	a.	In 20th century American even this old truism would cease to be true (US);
	b.	it is necessary to cease being a Negro, cease being true to history and himself
		(PK);
(19)	a.	I dread to think what will happen should I develope an ulcer (GB);
	b.	I dread thinking how our kids will negotiate the desolation (NZ);
(20)	a.	This can be a great idea for someone who likes to grow their own food (HK);
	b.	we just mentioned that fungus likes growing in warm and wet environment
		(BD);
(21)	a.	she intended to give her life to God (ZA);
	b.	she intended giving the boat to you (BD);
(22)	a.	I tried to use the best of my vocabulary in English to describe my feelings
		(BD);
	b.	I tried using it twice and still failed (SG).

In some cases, the distinction referred to in Section 1 between the actuality of the gerund participle and the future projection of the to infinitive is applicable/perceptible. Compare (23a), where the referent of he anticipates the future consequences of waking up “some sleeping thing”, with (23b), which describes the actuality of everyone’s past experience of finding Everton hard to beat.

(23)	a.	He spoke in an undertone, as though he feared to wake some sleeping thing (LK);
	b.	Everton are a team everyone fears playing because we all know their quality
		(GB).

Another distinction between to-infinitival and gerund-participial complements, known in the literature as the “Choice Principle” (Rudanko, 2017; Rickman & Rudanko, 2018), is that in the former the implicit subject is likely to exercise agentivity/choice over the event, but not so the subject of the gerund participial. This hypothesis is at best a tendency. It is clearly in evidence in (20a) where grow is transitive and (20b) where it is intransitive, but not the other examples above.

Unfortunately, there is insufficient room in this paper for exploration of all four of the semantic subtypes proposed, so I shall restrict the analyses to just the first two—aspectuality and emotional states—and in simple constructions only (no direct object).

4.4.1. Aspectuality

While aspectual catenatives of initiation and continuation allow both to-infinitival and gerund-participial complements, cessation catenatives generally do not (cease as in (18) above being an exception). Stop governs gerundial participials, as in (24), but not to infinitives, a restriction that is presumably semantically motivated, in so far as to infinitives project into another (temporal) location and are hence incompatible with the notion of cessation (e.g., * She stopped to go …).

(24)

she stopped going because nurses kept telling her to come back (NG).

As noted in Section 1, the English language has experienced a long-term drift in clausal verb complementation away from finite and infinitival constructions towards gerund-participial constructions. This trend has seen many of the catenative verbs involved emerging as candidates for grammaticalisation, all the more so as gerund-participial patterns are extended to include transitive uses as in (25).

(25)	a.	That maybe should have started them thinking in new directions (US);
	b.	they actually stopped them thinking about important political issues (PH).

One manifestation of grammaticalisation is semantic—and often phonological—weakening. For example, the earliest attested meaning of start in English was narrowly lexical (“leap”, or “jump”, a sense incompatible with any type of non-finite clausal complement), with the current sense of “commencement” emerging during the eighteenth century.

As

Table 11. The “START Ving” construction in GloWbE: frequencies pmw.

Total
163.0 (64.8%)
Inner Circle						Outer Circle
147.3 (59.4%)						165.9 (66.6%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
163.0		135.1		143.9		169.5				153.4				178.6					137.5
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
172.5	153.4	151.9	118.3	148.6	139.1	204.9	116.8	177.6	178.8	180.1	162.7	154.4	116.4	180.9	228.3	153.7	183.3	146.6	137.5

and

Table 12. The “START to V” construction in GloWbE: frequencies pmw.

Total
88.4 (35.2%)
Inner Circle						Outer Circle
100.7 (40.6%)						83.1 (33.4%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
94.8		102.9		104.4		66.9				105.9				77.0					86.9
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
87.1	102.5	115.4	90.3	108.4	100.4	61.4	61.9	62.2	82.0	124.8	106.5	91.3	100.8	97.7	61.2	55.2	76.8	94.2	86.9

show, gerund participials with start outnumber to infinitivals by around two to one (1.84:1 in GloWbE, compared with 2.13:1 in NOW). The OC varieties evidence marginally more support for gerund participials, while the IC varieties do so for to infinitivals, especially in NOW. Within the OC, it is Africa (and particularly NigE) that shows the strongest preference for gerund participials.

4.4.2. Emotional States

The catenative selected to represent this category is fear. A comparison of

Table 13. The “FEAR Ving” construction in GloWbE: frequencies pmw.

Total
2.13 (54.2%)
Inner Circle						Outer Circle
2.30 (65.5%)						2.06 (50.1%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
2.65		2.18		2.09		2.00				1.72				2.39					2.00
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
2.94	2.26	2.35	2.00	2.38	1.79	1.72	2.06	2.88	1.32	2.05	2.38	1.27	1.19	2.07	1.78	1.96	4.43	1.73	2.00

and

Table 14. The “FEAR to V” construction in GloWbE: frequencies pmw.

Total
1.80 (45.8%)
Inner Circle						Outer Circle
1.21 (34.5%)						2.05 (49.9%)
Americas		Europe		Oceania		South Asia				South-East Asia				Africa					Carib
1.26		1.25		1.14		2.26				1.44				2.54					1.19
US	CA	GB	IE	AU	NZ	IN	LK	PK	BD	SG	MY	PH	HK	ZA	NG	GH	KE	TZ	JM
1.52	0.99	1.24	1.25	1.40	0.87	1.85	3.26	1.93	2.00	1.33	1.68	1.83	0.91	1.23	2.06	1.93	4.27	3.21	1.19

shows that in GloWbE the gerund-participial variant outstrips the to infinitival by a ratio of 1.18:1 (one less pronounced than that in NOW, 1.74:1). Another notable finding for the gerund-participial variant was that—as found with a number of other catenative constructions—there was a correlation between IC ascendancy over the OC and AmE leadership within the IC. Also notable, and aligning with the findings for other constructions (e.g., “help NP V”, “V NP into Ving”), is the strong result for Africa in the OC.

5. Conclusions

Previous research on non-finite catenative complementation has largely been restricted to BrE and/or AmE. The present study has sought to expand the regional coverage of such research by analysing a set of catenative constructions (help + NP + (to) + V; “preventative” V + NP + (from) + Ving; “coercive” V + NP + into + Ving/to V; “aspectual/emotive” V + Ving/to V) across the twenty national varieties represented in GloWbE, with a second mega-corpus, NOW, regularly drawn upon for comparative purposes. The selection of these two large web-derived corpora is defended on the grounds of their capacity to capture low-frequency items and constructions, and of the potential insights offered by the register contrasts between them (as established by an analysis of the distribution of a set of colloquiality-oriented items, including contractions and punctuation marks, across the two corpora). Wherever possible, findings are related to historical trends attested in previous studies, and to such diachronically relevant phenomena as colloquialisation and grammaticalisation.

American influence is widely in evidence in the findings, with the Englishes of countries that are geographically close to the USA (CanE, JamE) and/or historically closely related (PhilE) tending to follow AmE patterns. For example, the American predilection for “prevent NP from Ving” is shared by all three of these varieties, while the American aversion to the alternative from-less variant “prevent NP Ving” is paralleled in CanE and PhilE. A similar result is attested with the declining to-infinitival variant with help, with the American distaste for this variant matched by that in CanE and PhilE. Curiously, in some cases, there were other varieties that shared an AmE preference or dispreference for a particular construction, but more strongly so than AmE itself. A case in point is help with bare-infinitival complements, where the AmE frequency (90.07 pmw) was strongly overtaken by that for CanE (113.52) and that for PhilE (118.52). While AmE epicentrality—and possibly hypercentrality—is clearly a factor in many of the study’s findings, it is not the only variety to enjoy epicentral status. The results for the IC across a range of constructions, both in GloWbE and in NOW, not only see AmE tending to outscore CanE but also BrE outscoring IrE and AusE outscoring NZE.

Another notable finding was that, occasionally, the OC varieties, as a group, conservatively buck real-time trends associated with the reference varieties. For example, the to infinitival with help is attested in a number of studies to be losing ground to its bare counterpart, and in the present study it is the IC whose frequencies suggest that it is leading the way in this development, its dispreference for the to-infinitival variant (23.60 pmw) being markedly stronger than that of the OC (32.61). Similarly, with catenative fear, the variant with a to-infinitival complement, which this study shows to be less frequent in GloWbE than that with a gerund-participial complement, is far more strongly endorsed by the OC (2.05) than by the IC (1.21).

The most remarkable finding in the results for the OC, however, is arguably the dominant scores for Africa. It was suggested that this finding is due to the popularity of “serial verb” constructions in a number of African languages, and, in particular, in Western African languages, including pidgins. It was also hypothesised that over time serial verb-promoted catenative constructions may have diffused from the Western African Englishes to regionally disparate Englishes in Eastern and Southern Africa.

The findings of the study are compatible with previous claims (by Leech et al., 2009; Mair, 2009; Collins, 2015) that historical changes in catenative complementation are driven in some cases by the stylistic factor of colloquialisation, and by the structural factor of grammaticalisation. Consider, for example, the current and increasing domination of bare infinitivals over to infinitivals with catenative help, a trend more advanced in the IC (with a bare-infinitival vs. to-infinitival ratio of 3.84:1) than the OC (2.86:1). The finding of the study that the overall ratio in GloWbE (3.08:1) is overwhelmingly surpassed by that in NOW (11.50:1) suggests that the bare-infinitival construction has undergone a colloquialisation-driven increase, which has become firmly established in the more formal written genre of news reportage. Furthermore, the decline of the infinitival marker to in this construction is undoubtedly ascribable to grammaticalisation. This development has parallels in the historical process that saw the canonical modal auxiliaries of contemporary English transition, between OE and ME, from lexical verbs to auxiliaries that select a bare infinitival (see further Hopper & Traugott, 2003). Furthermore, the semantic bleaching that is another common feature of grammaticalisation is in evidence with catenative help, with the notion of assistance giving way to merely generalised causation (similar to catenative make). Another uncontroversial case of grammaticalisation is that suggested by the dominance of gerund participials over to-infinitival complements with aspectual start. This situation, which is more pronounced in the OC with a Ving vs. to V ratio of 1.99:1, than in the IC (1.46:1), reflects the historical drift in catenative complementation in English towards gerund-participial constructions. As with help, one manifestation of grammaticalisation with aspectual start is semantic weakening, in this case involving a shift from the earlier meaning of “leap/jump” to the current sense of “commencement” which emerged during the eighteenth century.

It was noted that in the results reported for the two corpora, GloWbE and NOW, the similarities tended to outweigh the differences. It was suggested that a possible explanation for this finding is to be found in the tendency noted in previous studies for OC varieties to exhibit less stylistic variability across formal and informal contexts than IC varieties, and further that this tendency might be influenced by that in OC countries for English to be learned primarily in the formal context of the classroom.

It remains to comment on potential future directions. Researchers should seek to avail themselves of the most advanced statistical methods relevant to the study of complementation. Importantly, moreover, researchers exploring the grammar of World Englishes in general need to engage in the preparation of suitable large diachronic corpora comprising not merely web-derived texts but those from other sources, and not merely written but also spoken texts. Finally, it may be suggested that the present study could be used as a suitable reference point to explore catenative constructions in the EFL varieties of the Expanding Circle.