1. Introduction
The present article considers the complementation of the verb prevent. It is well known that the verb syntactically selects two types of gerundial complement clauses. Consider the sentences in (1a–b), both from the British English component of the NOW Corpus (see Davies Reference Davies2008–).
(1)
a. A hip injury prevented him from featuring at the beginning of the campaign before suffering a groin injury which […] (2019-12-22 GB)Footnote 1
b. Walking also helps to treat anxiety and depression and prevents these conditions starting in the first place. (2019-12-28 GB)
Prevent occurs as a matrix verb both in sentence (1a) and in sentence (1b). In each case the complementation of prevent involves a verbal gerundive -ing form. It is a commonly accepted view in English grammar that such gerundial constructions are sentential or clausal, and this view is adopted here. One reason for this approach is that it is then possible to represent the argument structure of the lower verb in a straightforward way. The two types of -ing clauses are different in that in the pattern of (1a) the complement construction of prevent is of the type NP from -ing and in (1b) it is of the type NP -ing, without from. For the purposes of discussion, the former pattern may be termed the ‘NP from -ing’ pattern, and the latter, without from, the ‘bare NP -ing’ pattern.
As regards the more detailed analysis of the NP from -ing pattern with prevent, there is a significant difference between the approach in Postal (Reference Postal1974) and that in Sag and Pollard (Reference Sag and Pollard1991). In Postal's Reference Postal1974 On Raising, he argued that a sentence of the type of (1a) involves Subject to Object Raising. (Dixon Reference Dixon, Brugman and Macaulau1984: 593 also proposes a Subject to Object Raising analysis for prevent.) That is, in a Raising analysis the lower subject of sentence (1a), for instance, is raised from the subject position of the lower clause into the object position of the higher clause, and the subject of the lower clause is then what in later frameworks would be called an NP trace. The NP trace is coindexed, and coreferential with the NP raised, that is, him in (1a), with the coindexing being a concomitant of the movement rule.
The other classic analysis in the literature of the NP from -ing pattern selected by prevent is the approach presented by Sag and Pollard (Reference Sag and Pollard1991). They include the verb prevent in their list of object control verbs. In other words, the object of prevent in (1a) is generated by phrase structure rules and the subject of the lower clause is then a PRO, an abstract pronominal NP that is not pronounced. In this approach there is no movement involved.
As far as the bare NP -ing pattern selected by prevent is concerned, scholars have provided careful information on the incidence of the pattern, especially in British English, but the construction has received less analytic attention in the literature than the NP from -ing pattern. One reason may be the circumstance, often remarked on, that the bare NP -ing pattern tends to be virtually absent from current American English, where the NP from -ing construction is clearly predominant. (The bare NP -ing pattern is also found in Australian and New Zealand English (see Mair Reference Mair, Peters, Collins and Smith2009), but these varieties deserve a separate treatment.) Thus Sag and Pollard (Reference Sag and Pollard1991), for instance, appear to pay no attention to the bare NP -ing pattern, restricting their control analysis to the NP from -ing construction. However, Dixon (Reference Dixon, Brugman and Macaulau1984: 59) does put forward the view that the bare NP -ing construction is generated in the same way as the NP from -ing pattern, that is, by Subject to Object Raising. However, he does not offer arguments for his position. In his later grammars of English, he has suggested that what is here called the NP from -ing construction ‘relates to a post-object complement clause’, and what is here called the bare NP -ing construction relates to a ‘complement clause in object function’ (Dixon Reference Dixon1991: 237, Reference Dixon2005: 259). These statements do not necessarily lack insight, but they are enclosed in parentheses in both of Dixon's grammars, and in neither grammar does the author engage in discussing the remarks further. Under these circumstances, there is a gap in the literature, justifying a closer look at the analysis of the two types of gerundial constructions, especially because today it is possible to make use of large electronic corpora of current British and American English.
The main task of the current article is to examine the syntactic properties of the two types of constructions. A great deal of attention has been paid in recent years to the recent history of the NP from -ing and bare NP -ing patterns with prevent (see for instance Rohdenburg Reference Rohdenburg1995: 85–97, Reference Rohdenburg, Plag and Schneider2000: 36–7; Vosberg Reference Vosberg2006: 149–57), including the virtual demise of the bare NP -ing pattern in American English in the twentieth century (see for instance Mair Reference Mair2002, Ong Reference Ong2011), and to potential semantic differences between the patterns (see Dixon Reference Dixon1991: 236, Reference Dixon2005: 259; Rudanko Reference Rudanko2002: 57–8, Reference Rudanko, Leistyna and Meyer2003; Sellgren Reference Sellgren, Lenker, Huber and Mailhammer2010), but the syntactic analysis of the two constructions in recent English (object control versus NP Movement) has attracted less attention. The advent of large electronic corpora also stimulated the present authors to take a fresh look at the syntax of prevent, with a focus on argument structures that the verb should be associated with.Footnote 2 With respect to each -ing pattern, the key question in any syntactic analysis is whether the pattern involves Subject to Object Raising (NP Movement) or object control. A related question is whether there is a need for a third type of structure in the syntactic analysis of gerundial complements of prevent. The syntactic issue is tackled in section 3. Section 2 offers a descriptive survey of the different types of complements of prevent in a sample of very recent British and American English, and of the incidence of the two types of -ing constructions in very recent British and American English on the basis of the NOW Corpus.
As regards the choice of the NOW Corpus as the main source of data, it comprises over 10 billion words of newspaper data collected from online versions of newspapers and magazines published in 20 different countries. It contains articles beginning in 2010, and is updated daily. The more prominent varieties of English contribute the largest amounts of data towards the corpus, among them British English, whose share of the total is currently a little under two billion words. The choice of the NOW Corpus for the present study was motivated by its size and the consideration that it makes it possible to examine the complementation of prevent in very recent English.
There are other verbs of prevention in English that deserve study, including block, hinder, keep, and stop (for a fuller list of such verbs potentially deserving study, see Visser Reference Visser1973: 2370–2373), but the authors have chosen the eponymous prevent for this study, because it is undoubtedly the central verb in this particular semantic area, and because the findings on prevent can then be expected to be a point of departure, and a point of comparison, in later analytic work on other verbs of prevention.
2. Prevent in a sample of very recent English
To shed light on very recent usage of complements of prevent, the present authors collected a sample of 200 tokens from both British and American English from the end of 2019, going backwards from December 31, 2019. For the data gathering, they used the simple search string ‘[prevent].[v*]’. Not specifying the context of prevent for this survey has the advantage of allowing the different types of complement to emerge. Obvious duplicates were counted only once in the survey.
In the datasets obtained, the number of irrelevant tokens not yielding information on the complementation structure of prevent was small. Examples included This post will be subject to enhanced checks as part of our prevent duty (2019-12-30 GB) and Next, avoid words such as cure, prevent, reduce, treat or stop (2019-12-26 GB).
The remaining tokens provide information on the complementation patterns of prevent. In the present dataset the most frequent type of complement is the NP, with 97 tokens in British English and 110 in American English. Examples are given in (2a–b). Another nonsentential pattern is the NP from NP pattern, with one token in the present dataset, given in (2c).
(2)
a. In fact, a supplement can help strengthen your hair and prevent hair loss. (2019-12-30 US)
b. […] a healthy lifestyle, which can still be effective in preventing progression towards more severe obesity. (2019-12-28 GB)
c. We can not assure you that our insurance coverage is sufficient to prevent us from any loss or that we will be able to successfully claim our losses under our current insurance policy on a timely basis, or at all. (2019-12-30 US)
The bare NP complement is clearly a very frequent pattern with prevent today. As for the NP from NP pattern, only one token was found in the sample, but additional examples are easy enough to find elsewhere in the NOW Corpus, as for instance in This is a calculated, intentional move on their part to prevent us from any further discovery in this case prior to the statute of limitations expiring (2011-03-15 US). While the pattern may not be very common, it has likewise been noted by Herbst (Reference Herbst2004), and, pace Aarts (Reference Aarts2012: 97) and also pace Landau (Reference Landau2002: 485, footnote 19), the pattern should be recognized as possible in current English. The pattern is worth drawing attention to because in it from clearly seems to be a preposition, and the similarity, for instance, of […] prevent us from any further discovery in this case […] and […] prevent us from discovering anything further in this case […] suggests that the from of the NP from -ing pattern may also be analyzed as a preposition.
Proceeding to sentential complements attested in the sample, the main focus of this article is on the NP from -ing and bare NP -ing patterns of the types exemplified in section 1, but the present dataset brings up tokens of -ing complements of the type of those in (3a–d) that deserve to be noted, even if only one of each type comes up in the present dataset.
(3)
a. His movement and communications have been monitored and restricted to prevent his fleeing the country and tampering with evidence, […] (2019-12-30 US)
b. Workers in the large barge began to jump into the water to prevent from sinking with the ship. (2019-12-27 GB)
c. I am aware that I do have the option of chemotherapy, but it's just so hard and so battering on the body. I'm trying to prevent going back to that, or at least prolonging the need for it. (2019-12-28 GB)
d. An internet application program provider shall protect user information, and obtain the consent of users while collecting and using users’ personal information in a lawful and proper manner and adopt proper measures, such as warning, limiting functions, suspending updates, and closing accounts, to prevent releasing illegal information content, keep records and report to the competent department. (2019-12-30 US)
Example (3a) is an example of POSS Ing (see Ross Reference Ross, Aarts, Denison, Keizer and Popova2004). This pattern, with POSS as the overt subject of the lower clause, emerged in the seventeenth century (de Smet Reference De Smet2013: 191–2) and was prominent in the eighteenth and nineteenth centuries, but is less frequent in more recent English (see Rohdenburg Reference Rohdenburg, Plag and Schneider2000: 37). Example (3b) is a prevent from -ing construction, which differs from the pattern of sentence (1a) in that in (3b) there is no overt intervening NP between prevent and from. As for (3c) and (3d), they are similar to (3b), except that there is no from.
Each of the lower clauses in (3b–d) has its own understood subject, and the understood subject in each case is PRO, since no movement operation is conceivable for any of the three sentences. A closer look at the sentences in (3b–d) reveals a difference in the way the understood subject of the -ing clauses can be interpreted. In (3b–c) it seems clear that the constructions are straightforward subject control structures, with PRO being coreferential with the subject of the matrix sentence. However, in (3d) PRO is not primarily coreferential with the higher subject, an internet application provider. Instead, the PRO permits a broader interpretation and the label PROArb (see Chomsky Reference Chomsky1986: 124–5) can be applied to the PRO in question.
The patterns of (3c–d) have been noted and illustrated in the comprehensive treatments of prevent by Poutsma (Reference Poutsman.d.) and the OED. However, as far as the present authors are aware, neither Poutsma (Reference Poutsman.d.) nor the OED mention the intransitive prevent from -ing pattern of (3b). Huddleston and Pullum (Reference Huddleston and Pullum2002: 657) also do not include the verb in their list of verbs of abstention such as keep from. Nor is the pattern featured in the list of matrix verbs of the formula ‘Noun-Phrase + Verb Phrase + from + Verb-Phrase + ing’ in Bridgeman (Reference Bridgeman1965: 25–26), which includes verbs such as escape and refrain. Before concluding that the usage represents a new discovery, it is of course necessary to ascertain that it has gained some currency in the language and is not a mere ‘flash in the pan’ in the present dataset. The NOW Corpus suggests itself as a source of further information here, and even if we only consider data from the first half of 2020, it is easy enough to find examples from both British and American English to substantiate the established status of the intransitive from -ing pattern in recent English. Two examples from British and American English are given in (4a–b) and (5a–b).
(4)
a. When you are writing a long series, burnout becomes a big issue. So you have to keep challenging yourself. Brooks avoided this pitfall by writing about something different at times, which helped prevent from falling into some kind of a predictable rut. (2020-05-11 GB)
b. They want you to find out why you did it. That's what's going to help prevent from doing it in the future. (2020-01-10 GB)
(5)
a. An investor will have to sell the stock to prevent from losing any further gains over the next few years. (2020-06-04 US)
b. Make sure that you are up to date on all your vaccinations because they do help prevent from getting additional viruses, and things that can be transmitted […] (2020-01-23 US)
Of the additional examples, (4a) and (5a) are in line with the subject control interpretation of the original example in (3b). A possible analysis for this pattern might involve a construction with a reflexive, and in the case of (5a), for instance, a variant of the type An investor will have to sell stock to prevent himself/herself from losing any further gains over the next few years seems possible. (The authors thank an anonymous reviewer for bringing up this possibility.) Such constructions with reflexives are found easily in the NOW corpus, as in […] I can not prevent myself from peppering my replies with glottal stops if a cabbie asks me […] (2019-12-08 GB), and a sentence such as (3b) would involve the omission of the reflexive pronoun. On the other hand, subject control is not applicable in (4b) and (5b), and a PROarb can be postulated.
Proceeding to the NP from -ing pattern and the bare NP -ing patterns in the dataset, the findings in the sample are presented in Table 1.
NP from -ing and bare NP -ing complements of prevent in a sample of very recent British and American English
The overall findings are in accordance with expectations in that the bare NP -ing pattern is much less frequent in the American English sample. However, the proportion of the bare NP -ing pattern turned out to be much lower in the present dataset of British English than what would have been expected on the basis of some other fairly recent work, where the proportion of the bare NP -ing pattern was almost equal to the NP from -ing pattern (e.g., see Mair Reference Mair2002, Reference Mair2006). In the present sample, the transitive NP from -ing pattern turned out to be almost three times as frequent as the bare NP -ing pattern. At the same time, it should be added that the figures in Mair's (Reference Mair, Yáñez-Bouza, Moore, Bergen and Hollman2019: 359) very recent article, based on the Brown Corpora, also give some indication of a decrease in the proportion of the bare NP -ing pattern in British English in relation to the NP from -ing pattern, compared to the proportion in the 1990s. That the bare NP -ing pattern might indeed be receding in very recent British English is a possible hypothesis, but it cannot be fully substantiated on the basis of the present sample, since the particular text type might also be a factor.
In view of the rarity of the bare NP -ing complement in very recent American English, it seems reasonable to draw mainly on the British English dataset when discussing the syntactic structures of that pattern.
3. Corpus data and the syntax of NP from -ing and bare NP -ing complements of prevent
As far as the syntactic analysis of NP from -ing and bare NP -ing complements of prevent is concerned, there is a noteworthy difference between the approaches in Postal (Reference Postal1974) and Sag and Pollard (Reference Sag and Pollard1991). Postal's comments also cover some other verbs that are semantically similar to prevent, but his focus is on prevent, and he considers such examples as I prevented Jack from kissing the gorilla (Postal's sentence 159a), reaching the conclusion that for constructions exemplified by his sentence, a Raising analysis is the ‘only one that does not run into obvious difficulties’ (Postal Reference Postal1974: 163). This means essentially that the NP Jack in Postal's sentence is generated by phrase structure rules as the subject of the lower clause and is then raised by the rule of Subject to Object Raising (NP Movement in later frameworks) into the object position of the matrix sentence. (Postal's approach was conceived before trace theory, but in later frameworks with trace theory an NP trace is left behind in the subject position of the lower clause.)
Regarding Sag and Pollard (Reference Sag and Pollard1991), they present their view of the analysis of prevent as part of a wide-ranging discussion of different types of control constructions in English, and it comes under their order/permit type of object control verbs. They write:
Verbs of the order/permit type all submit to a semantic analysis involving states of affairs (soas) where a certain participant (the referent of the object) is influenced by another participant (the referent of the subject) to perform an action. […] The influencing participant may be an agent (as in Kim persuaded Sandy to leave) or a nonagent (as in Ignorance of thermodynamics compelled Pat to enroll in a poetry class). The semantics of all verbs in this class thus involves a soa whose relation is of the influence type. With respect to such soas, we may identify three semantic roles, which we will refer to as influence (the possibly agentive influencer), influenced (the typically animate participant influenced by the influence) and soa-arg (the action that the influenced participant is influenced to perform (or, in the case of verbs like prevent and forbid, not to perform). (Sag and Pollard Reference Sag and Pollard1991: 66)
Sag and Pollard (Reference Sag and Pollard1991: 66) suggest that some verbs, including allow and permit, that have uses with three arguments of the type described in the quotation, also have a Raising analysis (with two arguments). However, they do not mention prevent among these, and therefore it is reasonable to think that they regard prevent as a verb involving control, that is, object control, and not Raising. In the quotation provided, prevent is mentioned without from, but they also provide a list of verbs of the order/permit type, and in that list they include the verb with from in parentheses (Sag and Pollard Reference Sag and Pollard1991: 65). Therefore, it seems reasonable to conclude that their reference to prevent in the quotation is meant to apply to the pattern of (1a), where from is present.
The research task here is then to question whether data from a very large corpus might shed light on aspects of the syntactic analysis of the two types of NP -ing clauses selected by prevent. The corpus consulted is the NOW Corpus, up to the end of June 2020, and the method of analysis is to consider such special NPs in their idiomatic uses which may distinguish between control and Raising constructions. One such NP is the NP cognizance, found in the idiom take cognizance, but the disadvantage in this case is that this NP is very rare, and relevant combinations with prevent can also be expected to be extremely rare. The present authors also considered the NPs advantage, part of the idiom take advantage, heed, part of the idiom pay heed, and tabs, part of the idiom keep tabs, but searches for relevant tokens, of the type They prevented tabs (from) being kept on our movements, did not yield examples for British or American English. However, existential there is a more promising target for investigation, because it is a frequently occurring NP, and also among the standard arsenal of diagnostics that can be used to separate control and Raising constructions from each other (see Davies and Dubinsky Reference Davies and Dubinsky2004: 7–8). The search string used was ‘[prevent].[v*] there’, and it retrieves 64 tokens. Not all of them are relevant. For instance, consider (6), where there is not dependent on prevent.
(6) If this is not prevented there is a danger of the Super 6 teams being a warehouse system attracting players […] (2018-05-09 GB)
However, most of the 64 tokens are relevant. There are some very isolated tokens, of the magnitude of one or two, in some non-core regional varieties of English in the corpus, but the figures from British and American English are given in Table 2. They are supplemented with tokens from two other core varieties, Australian and Canadian English.
From -ing and bare -ing complements of prevent in environments of existential there
The figures are so low for Australian and Canadian English that it is probably best to regard them as having curiosity value only, and to set them aside with that, but illustrations from British and American English are given in (7a–b) and (8a–b).
(7)
a. It is the last substantial strip left which prevents there from being an urban sprawl from London to Crawley […] (2010-02-13 GB)
b. We would love to be accessing those patients to prevent there being a problem in the future. (2012-07-30 GB)
(8)
a. […] it seems like a mistake to let the need for weepy in-memorium strings prevent there from being a guitar anywhere on stage. (2016-05-23 US)
b. “France is doing everything to prevent there being a lot of people coming to pray,” (2019-11-07 US)
The figures from Great Britain and the United States are not very high overall, taking into account that the segments of the NOW Corpus in question are very large. However, even these fairly low figures offer a surprise. In spite of the very clear predominance of the from -ing pattern in relation to the bare -ing pattern in the British English data that was observed in section 2 of this study, it turned out that in the particular diagnostic environment being considered here, there is an unexpectedly clear preference for the bare -ing complement in the British English data. Sentence (7a) is the only token of the from -ing type in this diagnostic environment, and the figure for the bare -ing pattern is as high as 14. As for the American English data, the figures are almost even, but in view of the well-known general predominance of the from -ing pattern in relation to the bare -ing variant with prevent in American English, which was also confirmed for very recent data above, the finding still shows an unexpectedly high frequency of the bare -ing variant in this particular syntactic environment.
It is also appropriate here to consider passive sentences with prevent and with NPs in idiomatic constructions. Sentences with the NP there are hardly suitable for this purpose (see Postal Reference Postal1974: 159, footnote 55), but Postal, as reported by Landau (Reference Landau2002: 487–488), has pointed to idiom-like constructions that may be considered. The constructions in question are extremely rare, but Postal has drawn attention to idiomatic sentences such as those in (9a–b):
(9)
a. Strings were prevented from being pulled.
b. The rug was prevented from being pulled out from under Mary.
There are speakers who may be hesitant or dubious about the status of one or both of the sentences in (9a–b), but taking Postal's intuitions, not contested by Landau (Reference Landau2002: 487–488), into account, the present authors have not marked them as ill formed. They may then be compared with the versions without from, given in (10a–b). (Postal does not consider the versions without from.)
(10)
a. ??Strings were prevented being pulled.
b. ??The rug was prevented being pulled out from under Mary.
The sentences in (10a–b) appear less well formed than those in (9a–b). The Complexity Principle, established by Rohdenburg (Reference Rohdenburg1996), may be a factor, with the more complex (passive) environment favouring the more explicit variant with from and contributing to the ill-formedness of (10a–b) but it is questionable whether it is sufficient to account for it by itself (see the penultimate paragraph of this section).
Regarding the theoretical significance of the results emerging from Table 2 and the comments on the passives, it is not easy to propose final conclusions, but as regards the active patterns without from, the numbers of bare -ing complements are noticeable in Table 2, especially in British English. Such sentences cannot be analyzed as control constructions, since the NP there cannot bear a theta role. That is, the NP there must be syntactically generated in the lower clause in sentences such as (7b) and (8b). This finding is solid and it is compatible with a Raising analysis.
Whether the evidence of sentences (7b) and (8b) is a convincing argument for a Raising analysis of the pattern without from is a separate question. It was pointed out above that the passives in (10a–b) sound unlikely to speakers, while the variants with from sound better. This state of affairs brings up a third possibility. This is that prevent, when it selects a bare -ing complement, may involve an ACC Ing complement. That type of construction was argued in Postal (Reference Postal1974: 105, note 16) to be appropriate for a class of verbs in English, including resent. He pointed out that while an active sentence of the type They resented it happening to Bob is well formed, a passive version of the type *It was resented happening to Bob is ill formed. This is explained under the assumption that the NP it is generated in the subject position of the lower clause and that it then stays in that position. In other words, verbs of this class do not permit Subject to Object Raising (with the proviso that the lower subject is in the oblique (or non-nominative) form even though remaining in situ (see Postal Reference Postal1974: 105)). The underlying assumption is that an NP raised by Subject to Object Raising can be expected to permit passive variants. This is a reasonable assumption to make since there are quite a number of matrix verbs with objects that undergo Subject to Object Raising that even prefer the passivization (or some other movement operation) of the raised NP, including allege (Postal Reference Postal1974: 304) to leaving it in situ.
A resent type analysis for the derivation of bare -ing constructions of prevent would account for the occurrence of sentences such as those in (7b) and (8b), and the ill-formedness of passivized versions without from, as in (10a–b). Accepting that passivized versions are better with from -ing, as illustrated in (9a–b), it appears that a Raising analysis is needed in the case of the from -ing complement.
The bigger picture to emerge from this discussion of the syntax of prevent is that the verb resists simple categorization as regards the distinction between object control and Subject to Object Raising. With respect to the from -ing pattern, it can be said with confidence that control structures involving object control are needed for that type of complement. For instance, consider sentences such as that in (11) for British English and that in (12) from American English from the present dataset.
(11) Their being outside the government did not prevent the voters from holding them accountable on December 12. (2019-12-28 GB)
(12) […] if he came in and found Turner or if Pinkney prevented him from coming in, she would be arrested, according to documents. (2019-12-30 US)
In the sentences of (11) and (12), the verb selects three arguments and has an interpretation based on influencing, spelled out by Sag and Pollard (Reference Sag and Pollard1991: 66) as typical of object control; the surface subject represents the object of prevent in each case.
It is also possible to associate the object control analysis with a particular sense of the verb, namely, that of ‘stop, keep, or hinder (a person or thing) from doing something’ (OED, sense 14a). It can be added that in the case of sentence (1a) it is also easy to form a passive of a type where the derived subject is linked to prevent, of the type He was prevented from featuring at the beginning of the campaign.
Keeping the focus on the from -ing pattern, the discussion also suggests that an object control approach is not the only analysis of the pattern with prevent. If we can accept Postal's view (see above) that (9a–b) are well formed, it is plausible to think that they involve Subject to Object Raising. In these sentences the surface subjects are the NPs Strings and The rug, idiomatically used, and they are linked to the predicates of the lower clauses. As a consequence, they must have been generated in the lower clauses and then raised into the higher clauses by Subject to Object Raising.
Going beyond the small datasets and idiomatic uses, it is possible to strengthen the case for a Raising analysis of the NP from -ing pattern, because it is easy enough to find passivized sentences not involving special NPs where it seems clear that the surface subjects originate in a lower clause and do not represent the original direct objects of prevent. In such configurations the verb thus selects two arguments, not three. Consider the sentences in (13a–b), from the British English segment of the NOW Corpus.
(13)
a. Two children, aged seven and nine, were prevented from being prosecuted for a knife-related offence because of their age, […] (2003-03-01 GB)
b. As a result of this personal conflict, the nine-pounder gun was prevented from being re-loaded, and was eventually captured by the Infantry, […] (2010-04-08 GB)
In (13a–b) prevent does not have the sense of ‘influence the referent of NPO not to perform some action,’ to hark back to part of the formulation provided by Sag and Pollard (Reference Sag and Pollard1991: 66). Instead, in (13b), for instance, the NP the nine-pounder gun represents the underlying object of the verb re-load and is raised from the lower clause by Subject to Object Raising, and the sense of prevent is causative, along the lines of ‘bring it about that not S2.’ Or, to give a more elegant definition, it is possible to refer to a particular sense of the verb in the OED: ‘To preclude the occurrence of (an anticipated event, state, etc.)’ (sense 9.a of the verb in the OED).
The conclusion with respect to the NP from -ing pattern is therefore that it may involve either object control or Subject to Object Raising, and that each syntactic analysis may be associated with a specific sense of the verb. The associations of the two syntactic analyses of the from -ing pattern to two different specific senses of the verb has not been noted in the literature before, as far as the present authors know, and it helps to give a grounding to the present analysis.
Proceeding to the bare NP -ing complement pattern with prevent, the present authors propose that an ACC Ing analysis is applicable to the pattern. This proposal is motivated by the findings given in Table 2, where an astonishingly high number of the special NP was found with bare -ing complements, and by the ill-formedness of the passivized versions of such sentences with special NPs, as illustrated in (10a–b). The sentences in (7b) and (8b) are thus analyzed as involving ACC Ing, not Subject to Object Raising.
The conclusion that the bare NP -ing pattern with prevent can involve ACC Ing complements and does not involve Subject to Object Raising brings up the question of whether the bare NP -ing pattern can involve object control. That is, the question is whether in a sentence without any special NP an ACC Ing analysis is likewise applicable to bare NP -ing complements, with the NP remaining in the lower clause, or whether an object control analysis should be countenanced for such sentences, with the NP a constituent of the higher sentence. To illustrate, consider the example in (14).
(14) […] we are determined to do all we can to prevent such weapons reaching our streets. (2019-12-28 GB)
To approach the question from the point of view of constituent structure, we may note that pseudocleft sentences such as those in (15a–b), from the NOW Corpus, are well formed.
(15)
-
a. What health officials want to prevent is people believing the common adage of “I've had the flu. It wasn't that bad.” (2019-10-13 US)
-
b. What we are trying to prevent is students putting out their hands when they fall because that's going to cause an injury. (2013-07-22 US)
-
On the analogy of the sentences in (15a–b), a pseudocleft variant of sentence (14), abridged in non-essential ways, also seems possible, as in (16).
(16) What we are determined to prevent is such weapons reaching our streets.
On the reasonable assumption that substrings in the focus position of a pseudocleft sentence are constituents (see Higgins Reference Higgins1973: 12; Duffley Reference Duffley2000: 227, Reference Duffley2006: 36–37; Duffley and Fisher Reference Duffley and Fisher2021: 85), the well-formedness of sentence (16) suggests that the NP of the bare NP -ing complement, even when not a special NP (such weapons in the case of (14)), can indeed be a constituent of the lower clause.
It is also pertinent to take note of a comment in the OED on what has here been called the bare NP -ing pattern:
With gerund but without from. The construction prevent me (you, etc.) going appears to be short for prevent me (you, etc.) from going, perhaps influenced by prevent my (your, etc.) going (see sense 9b) from which it is, in any case, indistinguishable (in the absence of a written apostrophe) when a plural noun precedes the gerund. This construction has sometimes been criticized as incorrect. (OED, comments under sense 14 of prevent)
The suggestion that the bare NP -ing complement may have been influenced by what may be called the POSS Ing pattern (see Ross Reference Ross, Aarts, Denison, Keizer and Popova2004), as in prevent my going, and that it is often hard to distinguish it from the POSS Ing pattern provides some background or perhaps even indirect support for the idea that even in the absence of a special NP, the bare NP -ing pattern in a sentence such as (14) might not involve object control.
Another way to approach the question is to consider the syntactic behavior of the NP in question. The most obvious rule to consider is Passivization. Kaunisto and Rudanko (Reference Kaunisto and Rudanko2019: 122–4) provide some evidence of recent usage based on the Hansard Corpus (2014) that is relevant here. The sentence in (17) is one of their examples:
(17) Matthew Kelly, county councillor of county Clare, was prevented addressing his constituents, […] (Hansard, House of Commons, May 10, 1901)
Sentence (17) attests that the pattern has existed in fairly recent English, and Kaunisto and Rudanko (Reference Kaunisto and Rudanko2019: 123) show how it was relatively frequent in the nineteenth century, with as many as 66 tokens in the 1880s. However, they also show that in the twentieth century its frequency has been very low, to the point that no tokens at all were found in the Hansard Corpus in the period from the 1960s to the 1990s. This represents a dramatic decline, with the pattern going into desuetude, and it makes it easy to understand why Aarts (Reference Aarts2012), while also noting that the construction was found in earlier English, starred a sentence of the type *The sailor was prevented drowning the cat, alongside of the well-formed The sailor was prevented from drowning the cat. On the basis of data from those four decades, it would be reasonable to put forward the claim that the bare NP -ing pattern should be limited to the ACC Ing construction and that an object control analysis should be excluded for the pattern. That said, it still needs to be added that the British National Corpus (BNC, 2001) brings to light one token suggesting object control: If one person is prevented getting AIDS by the officer […] (KRL 4350, 1985). The token shows that object control cannot be totally excluded for the bare NP -ing pattern even during those decades, even if there was a tendency at that time to view the construction as being of the ACC Ing type.
However, Kaunisto and Rudanko (Reference Kaunisto and Rudanko2019: 123–4) also note that the 2000s, the most recent decade available of the Hansard Corpus (2014), brought a resurgence of the pattern in the corpus, with as many as 31 tokens. An example is given in (18), from Kaunisto and Rudanko (Reference Kaunisto and Rudanko2019: 122).
(18) We would not want to be prevented obtaining information about a property that someone had failed to declare when making a claim for benefit […] (Hansard, House of Lords, Feb. 1, 2001)
Such dramatic fluctuations in the use of the pattern in very recent decades suggest that the bare NP -ing pattern may be undergoing change, with the object control pattern again becoming increasingly available to the bare NP -ing complement of prevent in very recent British English.
Confirming the trend favoring the availability of object control with the bare NP -ing pattern, Kaunisto and Rudanko (Reference Kaunisto and Rudanko2019: 124) also provided further examples from the British English component of the NOW Corpus. Their latest example is from 2017. Consulting data from the same corpus for even more recent years for this study, the present investigators used the search string ‘[vb*] prevented [v?g*]’, where ‘[vb*]’ stands for a form of the verb be, and retrieved six tokens. Three of them are not relevant, with an example of an irrelevant token given in (19), but three are relevant, with examples given in (20a–b).
(19) Scientists say that half of all premature births could be prevented using simple
tests and antibiotics. (2018-06-10 GB)
(20)
a. Milner tries to get away from Young but is prevented doing so by his opponent, who grabs hold of his shirt […] (2019-02-24 GB)
b. There were plenty of Toon fans who wanted to attend but we're prevented doing so by petty vindictiveness. (2019-01-09 GB)
The sentence in (19) does not represent object control, but those in (20a–b) do. They are also quite in line with the typical semantics of object control, with the sense of the verb being ‘to stop, keep or hinder (a person or thing) from doing something’ (OED sense 14a). The sentences in (18) and (20a–b) are also worth noting because they show that the Complexity Principle, while it would favor the variant with from, is not strong enough to block such sentences in current British English.
To sum up the discussion of the bare NP -ing complement with prevent, it is argued here that the ACC Ing complement is relevant to its analysis. Object control, on the other hand, does not seem to have had much prominence with the bare NP -ing complement in the second half of the twentieth century, but there is increasing evidence that it is available in recent British English.
4. Concluding observations
This article was designed to inquire into the syntactic properties of NP from -ing and bare NP -ing complements of the matrix verb prevent, with evidence from large corpora. It is well known that the former pattern is commonly found in both British and American English, and that the latter pattern is more or less restricted to British English, as far as fairly recent usage is concerned. In section 2 the authors examined the complementation of the verb in fairly small samples of British and American English, without specifying the context of the verb, with the original aim of shedding light on the incidence and the status of the two types in these varieties. That section provided two surprising results. Firstly, the frequency of the bare NP -ing pattern in British English turned out to be lower than expected. Secondly, the investigation also brought to light an unexpected complement of prevent, as in Workers in the large barge began to jump into the water to prevent from sinking with the ship. As far as the present authors are aware, the construction has not been noticed in earlier work on prevent, but additional illustrations are presented from both British and American English. In light of the data here, the pattern deserves to be recognized as an innovative construction, and more work on its emergence and spread in British and American English is desirable.
In section 3 the discussion moved onto comparing the more familiar NP from -ing and bare NP -ing constructions. It has sometimes been taken for granted that the same syntactic analysis should be applied to both constructions. However, the present study argues that this is not the case. As far as the NP from -ing construction is concerned, it is argued that the pattern straddles the divide between object control and Subject to Object Raising, and that both structures should be permitted for the verb. Further, it was argued that each structure can be linked to a specific sense of the verb. As for the bare NP -ing construction, it is suggested that an ACC Ing analysis should be postulated for it. Object control was also possible in the nineteenth century, but during several decades of the twentieth century it seems to have become more marginal. However, in very recent English there is clear evidence that object control is again more readily available to prevent when the verb selects the bare NP -ing complement. Such trends of change will naturally bear watching in future decades. Another obvious research task is to consider other verbs of prevention in relation to the syntactic analyses proposed in the present article.
