Let and the "bare infinitive": an exploratory exercise in traditional (notional) grammar.

ABSTRACT

The present-day English verb let is descended from one of a set of causative verbs in Old English. In part, its history and present-day status are shared with some of these verbs and others that develop as causatives in the course of time; and the general outlines of this history and, in particular, the resultant system will be my concern in [section]2 of this paper. But let shows some idiosyncratic developments compared with these others; and it is these I shall take up in [section]3. These developments depend in part on the results of the history outlined in [section]2, but they also involve an idiomatisation that throws some interesting light on the interaction of syntax and lexicon in syntactic change. All of this is of some interest in our assessment of the goals of grammar.

1. Introduction

Present-day English let belongs to that small group of causative verbs that take what I shall refer to as the "bare infinitive" as a complement, as illustrated in (1):

1) a. He let the butler leave.

b. He made the butler leave.

c. He had the butler leave.

With them we can contrast in this respect the causatives of (2), where we find rather the periphrastic infinitive as complement: (1)

2) a. He allowed the butler to leave.

b. He caused the butler to leave.

c. He got the butler to leave.

As is familiar, most complement-taking verbs take the periphrastic infinitive. But the capacity to take the bare infinitive infinitives shown in (1) is shared with the "core" verbs of physical perception, as illustrated in (3):

3) a. He saw the butler leave.

b. He heard the butler leave.

And, as is again very familiar, this is also a property of "core" modals:

4) He may leave.

The focus of one of our concerns here will thus specifically be on the status of the bare infinitive and, rather briefly, the history that led to this.

The group in (1) are heterogeneous in some other respects. Thus, while let, like perceptives (Fischer 1991: [section] 1.5), shows straightforward bare passive infinitive complements, as in (5a), make is much more reluctant to accept such complements, while with have they are "truncated":

5) a. He let the butler be replaced.

b. He made the butler ?*be replaced/be examined.

c. He had the butler (*be) replaced.

Passives are better after make if they can be construed as involving some volition on the part of the subject of the infinitive, as with the second possibility in (5b).

The causatives with periphrastic infinitives take a straightforward periphrastic passive complement, except for get, where the usual passive is again truncated (and to-less):

6) a. He allowed the butler to be replaced.

b. He caused the butler to be replaced.

c. He got the butler *((to) be) replaced.

This pairs the indirect causatives have and get.

All of the causatives of (2), when passivised, take a regular periphrastic infinitive. But, though make, when passivised itself, takes a periphrastic infinitive, the other two do not passivise, whichever infinitive is involved:

7) a. *The butler was let (to) leave.

b. The butler was made *(to) leave.

c. *The butler was had (to) leave.

The notation in (7b) indicates that to cannot be omitted.

Make patterns with the perception verbs in this last respect, as with the subordinate active infinitive of (8a), but the perceptives can also take in certain circumstances a "truncated" passive infinitive, like have in (5c) and get in (6c), as is illustrated in (8b), but in this case there is also a full one, but periphrastic, as in (8c):

8) a. The butler was seen/heard *(to) leave.

b. The butler was seen (*?be) dismissed/humiliated.

c. The butler was seen to be dragged away.

But the "truncated" passive in (8b) is much more restricted than with the causatives--though the bare passive infinitive seems to be even more marginal, as marked in (8b).

Whereas all of the causatives straightforwardly take agentive complements, have is less happy than the others with non-agentive complements:

9) a. He let the vase fall over.

b. He made the vase fall over.

c. * He had the vase fall over.

The infinitives with the animates in (10a-b) are also normally interpreted as non-agentive:

10) a. He let the butler fall to his death.

b. He made the butler fall to his death.

c. *He had the butler fall to his death.

If forced to provide an interpretation for (10c), we seek an agentive reading; so that we must resort to, for instance, a reading whereby "he" is the director of a film arranging some scene.

Let shows some more severe idiosyncrasies. Notably, when it takes go as a complement, we find an alternative construction, with a distinct interpretation:

11) a. He let the butler go.

b. He let go (of) the butler.

The instance of the general construction in (11a) involves a regular interpretation of go as a directional verb, so that we can complete it as in (12a):

12) a. He let the butler go to Paris.

b. *He let go the butler to Paris.

The let go sequence of (11b) has an idiosyncratic interpretation. Some light is cast on the development of this structure and its interpretation by the general history of these causative verbs, however--to which I now turn.

In what follows I do not endeavour to provide an account of all the diversity we have encountered in this section; that would involve a much more extended exercise. Here I concentrate (in [section]2) on the (particularly Present-day) status of the periphrastic vs. bare infinitive distinction and (in [section]3) its relation to the development of let + infinitive. I argue that there is a semantic and functional basis for the distribution of the bare infinitive, over-tiding the default status of the periphrastic infinitive, and that its history illuminates the development, in particular, of the idiom of (11b). This development in turn throws some light on some of the other idiosyncrasies of let.

2. The development and status of the bare infinitive

Old English, similarly to Present-day, had two infinitive constructions: one is marked by a suffix, usually in West Saxon realised as -an; the other is periphrastic, involving to + a verb form bearing another suffix, -anne/-enne. These verb forms are historically inflected forms of a verbal noun. The latter construction is often misleadingly called the "inflected infinitive"; here I shall distinguish the (purely) suffixal (ultimately bare) vs. the periphrastic infinitive. These are respectively illustrated in (13):

13) a. We wilniad mid urum hlaforde claenlice sweltan, swidor donne

'we desire with our lord purely to-die rather than

unclaenlice mid eow lybban

impurely with you to-live'

b. da wilnade ic [??]ara monna onsyne to seonne

'then desired I those men's faces to see'

(Mitchell 1985: [section] 955).

The factors determining choice of construction when the complements involve the same verb, as in (13), are unclear. But there are some general patterns in the use of the infinitives. For instance, passive infinitive complements are suffixal (according to Traugott 1992: 243-244); and the periphrastic infinitive, in accord with its historical source, is associated with infinitives that are adjuncts of purpose (Kisbye 1971: A1-13). I am not concerned here, however, with the adjunctival periphrastic infinitive as such, or with the development of the for (...) to construction.

2.1. The basic distribution of the bare infinitive

What is relevant here, though, is that in Old English there is a set of causatives that do not take a periphrastic infinitive complement (Mitchell 1985: [section]955, referring to Callaway 1913: 31, 93). And Fischer (1992: [section]4.6.2.3) discusses causatives and perception verbs as taking the bare infinitive in Middle English. Causatives and verbs of perception in Old English are also the classes most specifically associated by Traugott (1992: 247-248), basing herself on Fischer's (1990: 218-309) analysis of "accusative-and-infinitive" constructions, with "subject-to-object raising". However, Fischer et al. (2000: [section]7.2.2), for example, reject a "raising" ("ECM", in their terms) analysis. I see no reason, however, to adopt their proposal. (2) It is perhaps not coincidental that we can associate with the grouping of causatives and perceptives both co-inheritance of the "subject-to-object raising" construction (however this is analysed) with a bare infinitive and their joint retention in later English of the bare infinitive complement and their sharing of the interpretation it attracts. Let us look at this later history.

Fischer (1990: [section] 4.6.2.3) cites the work of Kaartinen and Mustanoja (1958) on Late Middle English prose and of Quirk and Svartvik (1970) on Chaucer as revealing two "parameters" that are important in the choice between bare and periphrastic infinitive. One involves the "'intimacy' of the relationship between the matrix verb and the infinitive" (Fischer 1990: 317); the other "the physical distance between the matrix verb and the infinitive". I am not concerned here, except incidentally, with the latter. Fischer associates greater "intimacy", in the first instance, with the complemented verb being "more grammaticalised" or "emptier ... of referential meaning". Thus, the "core" modals take the bare infinitive, as do the causative uses of "verbs such as haten, bidden, let, gar, do and maken" (Fischer 1990: 318). On the other hand, only the periphrastic infinitive is found "where such grammaticalisation is out of the question", so that "only the (for) to-infinitive is found after nouns and adjectives" (Fischer 1990: 318). However, as Fischer observes (1992: 320), perception verbs "clearly retain their full semantic content but nevertheless normally retain the bare infinitive". And she cites examples such as the Chaucerian (14):

14) For sorwe of which she felt hire herte blede ...

She suggests we interpret "intimacy" here as involving "as it were 'intimacy' of tense, the simultaneity of the actions expressed by matrix verb and infinitive" (Fischer 1992: 320). Fischer notes that when "simultaneity" or "direct perception" is not involved, a finite complement appears, as in the Chaucerian examples of (15) (which retain her italicisations):

15) a. Up to the tree he caste his eyen two,/And saugh that Damyan his wyf had dressed/ In swich manere it may nat been expressed

(CT IV: 2360-2362).

b. The sothe is this, the cut fil to the Knyght,/ And telle he moste his tale, as was resoun,/ ... And when this goode man saugh that it was so ...

(CT I: 845-850).

And she goes on (1992: 320-2; see too 1995, 1996) to argue for a wider role in Middle English for such a temporal distinction as underlying choice between the two infinitives.

It may be that we can give some theoretical content to the application of this "intimacy" metaphor to both periphrases and temporal domains. Indeed, it may be possible to formulate a relationship between these "intimacies" and physical proximity, and so be able to explicitly generalise over all these factors favouring choice of the bare rather than the periphrastic infinitive. But it seems to me that causatives and perceptives are already linked by some version of the temporal, or directness, constraint invoked by Fischer; and that "grammaticalisation" may not be important in the former (causative) case, either.

Views of what constitutes "grammaticalisation" vary, of course. But, in the course of a survey of causatives, Palmer, for instance, claims that "English does not ... appear to have a grammaticalized causative construction" (1994: 217). And, indeed, the only apparent potential evidence of causative "grammaticalisation" of some sort in Present-day English comes from the selection of the bare infinitive by some causatives (as illustrated in (1)). This property is shared with perception verbs, which "clearly retain their full semantic content" (Fischer 1990: 320), however. Certainly, various causatives, such as do, can be said to have been "grammaticalised" (in some sense) in the history of English, but as non-causatives. However, something else seems to be involved in the selection of the bare infinitive by some causatives. And it is, I am suggesting, related to an elaboration of the constraint on perception verbs when they take a bare infinitive.

We can illustrate the force of "directness" in the perceptive + bare infinitive construction in terms of the contrasting acceptability (in normal circumstances) of the pair of sentences in (16):

16) a. *He saw him leave on Wednesday on Tuesday.

b. He told him to leave on Wednesday on Tuesday.

(16a) is incompatible with a requirement that perceptive + bare infinitive involves direct, instantaneous perception. In (16b), moreover, the infinitive is irrealis, non-factual, unlike that in (16a), where "directness" also involves factuality. Perceptive + bare infinitive involves the factuality of two punctual events, one of them representing a perception of the event represented by the other.

Non-punctuality of the infinitive denotatum is marked, as in (17a), with a non-infinitive construction:

17) a. He saw him leaving.

b. He could see him leave/leaving.

Non-punctuality of the perceptual verb with either the infinitive or the -ing-form is marked "modally" as in (17b). I am going to suggest that punctual factuality is part of what links causatives and perceptives.

"Directness" and, perhaps, simultaneity are also part of our normal interpretation of most of the causatives of (1) and (2). So that our giving an interpretation to (18) involves us in a reading whereby the referent of the subject has directly, by word or deed, affected the "object" referent at some point on Tuesday, such that this leads, directly and factually, to the Wednesday leave-taking:

18) a. He made him leave on Wednesday on Tuesday.

b. He caused him to leave on Wednesday on Tuesday.

However, it must be admitted that, apart from strict application of the simultaneity expectation with causatives thus being frangible, as in (18), the "object" referent in (1/2c) is not affected directly. So, it is apparently the linked factuality of the two punctual events that crucially relates causative and perceptual verbs. Of course, the punctuality of the causative can be over-ruled by the progressive construction, or it may be made habitual by choice of accompanying temporal adverbial (every morning etc.), but the unmarked interpretation is simply punctual.

This relationship is associated with both causatives and perceptives. But, in addition, perceptual verbs require strict simultaneity. It is this, I conjecture, that underlies their greater insistence on selection of the bare infinitive. The less constrained (in this respect) causatives are divided: only some of them take the bare infinitive, as illustrated by (2) vs. (1). Punctual actuality of two events seems to be a general feature of use of the bare infinitive, however. Crucial, however, is the punctual character of the infinitive, in so far as the perceptives themselves can take at least a "modal progressive", as in (17b), and the causatives can take an ordinary progressive:

19) a. He is making the butler leave.

b. He is letting the butler leave.

c. He is having the butler leave.

This suggests that we formulate the bare infinitive condition (BIC), provisionally, simply as it affects perceptives:

Bare Infinitive Condition

The bare infinitive represents a punctual factual event co-temporal with the governing verb.

This is a semantic condition. (3)

However, that causatives are less strict in their observance of BIC than perceptives, was illustrated by (18). And they accordingly also allow independently habitual infinitive complements, unlike the perceptives:

20) a. He made the butler serve drinks every day.

b. He saw the butler serve drinks every day.

In (20a) either only the complement or the whole sentence comes within the scope of every day, whereas in (20b) the scope of the adjunct can only be the whole sentence; the complement of the perceptive in (b) cannot be independently habitual. Thus causatives obey only the weak bare infinitive condition (WBIC):

Weak Bare Infinitive Condition

The bare infinitive can represent a punctual factual event co-temporal with the governing verb.

This may account for their varying between bare-infinitive and peripheral-infinitive complementation, as illustrated by (1)-(2) (and by the behaviour of help observed in n. (1).

The BIC is called into question, obviously, by the modals, exemplified in (4), which typically introduce irrealis complements. But we also have to account for the absence of the bare infinitive in (7b) and (8), as well as its absence in, say, (21), where we seem to have linked factual punctualities:

21) a. He happened to leave on Wednesday.

b. He managed/contrived to leave on Wednesday.

c. He remembered to leave on Wednesday.

In the case of (7b), (8) and (21) it seems that a syntactic constraint intervenes, given informally as the bare infinitive complement filter (BIF):

Bare Infinitive Filter

A lexical verb must be separated from its complement bare infinitive by another complement.

This assumes that the unmarked, default infinitive in Present-day English is the periphrastic. All of (7b/8a-b/21) involve a sequence of verb and infinitive: only the periphrastic infinitive is allowed in such circumstances. In (1b) and (3) the "object" intervenes between the verb and the bare infinitive, legitimating the occurrence of the latter. And in (22) only an (optional) adjunct intervenes, which is not sufficient to ensure the occurrence of a bare infinitive:

22) a. The butler was made by us *(to) leave.

b. The butler was seen by us *(to) leave.

c. The butler was heard by us *(to) leave.

BIC would then be to some extent a perhaps curious inversion of the role of the other factor mentioned by Fischer (1990: 317) as favouring the periphrastic infinitive in Late Middle English: physical distance between the verb and its infinitive complement. By BIF it is the selection of the bare infinitive that depends on separation--though, admittedly, strictly local.

However, there are perhaps semantic considerations at work with (21) too. For, unlike with the perceptives and make, (20a), at least, scarcely involves two discrete "events"; the verbs in some sense represent the "same event". Perhaps this is what blocks the bare infinitive here. And this suggests a way of approaching the rest of (21). For they too represent events that are in one way less discrete than the perceptives and causatives, by reason of the semantic distinction that, I suggest, underlies their failure to conform to BIF. The element satisfying the valency of the infinitive that would appear as its finite subject is not distinct from the cogniser or agent of the main verb. The cogniser/agent is not discrete from the event represented by the infinitive. If this particular discreteness is part of the (semantic) requirements for the bare infinitive, we might be able to reduce the grammatical restriction to the requirement that passive lexical verbs cannot take a bare infinitive; its absence with (21) follows from the non-discreteness (from the cogniser/agent) of the "subject" element in the event represented by the infinitive. However, I shall retain BIF in something like its more general form. I see no reason to exclude a situation wherein absence of the bare infinitive is the result of two different kinds of violation.

Modals are at least not susceptible to BIF as it stands, since they are auxiliary, not lexical verbs. But they do not conform to BIC. They seem to have a positive requirement, despite this, for complementation by a bare infinitive, one which involves simply their categoriality, as is given in the following formulation of the infinitive filter (PIF):

Periphrastic Infinitive Filter

The infinitive complementing non-lexical verbs is bare.

Nevertheless, despite PIF, some modals exceptionally take the periphrastic infinitive, and are thus "non-core". Ought (and used) is "non-core"; and it seems to be pulled in different directions, towards the "core" (thus He oughn't leave) or towards conformity to non-modals (thus He didn't ought to leave). More interesting are modal be and have, which insist on the periphrastic infinitive (He is/has to leave). This suggests that PIF applies not to all non-lexical verbs, but only to those which are universally necessarily finite. Be and have are non-lexical verbs that (even in such uses--and not simply when, in the case of have being causative, for instance) may be non-finite--though not when complemented by the infinitive, certainly; their "necessary finiteness" is local to this use. Have and be also diverge in showing a (over-)full person-number-tense morphological paradigm when they are finite. This is another mark of their "non-core" status.

An appropriate modification to PIF is introduced as part (a) of a composite infinitive filter (IF), which identifies "core" modals in terms of "universal finiteness":

Infinitive Filter

a) The infinitive complementing verbs that are universally finite is bare.

b) The infinitive uniquely complementing verbs is periphrastic.

I have (slightly) recast BIF as b of IF, where part a takes precedence. This formulation assumes of course that passive by-phrases, as in The butler was seen by me to leave the house (at five), are adjuncts. In the light of the discussion of (21), I interpret the constructions in these sentences as semantically excluded from taking bare infinitives, as well as by violating BIF/IFb, which also excludes the bare infinitive after the passives in (7b/8a-b).

Let us sum up this subsection, before testing its results against other infinitive constructions. BIC, which requires the infinitive to be bare, is clearly semantically based. There is also an iconic element. The lack of separation by to can be seen as a realisational reflection of the lack of separation in time of the events represented.

IFa, however, represents a "grammaticalisation", with marginal lexical exceptions like ought (and perhaps used). "Core" modals (or, more precisely, items that are universally necessarily finite) do not conform to BIC, and we cannot associate the exceptional behaviour of modals only with other semantic properties to do with, say, non-factuality, since other irrealis "object-to-subject raising" verbs take periphrastic infinitives, as illustrated in (23a):

23) a. He seems to have left.

b. He may/must/will/might/should/would have left.

Compare the range of non-factualities associated with the modals of (23b). Similarly, it appears to be the absence specified by IFb of another (preceding) complement that is crucial in limiting the distribution of bare infinitives after verbs that are semantically suitable to take these. This is a syntactic restriction.

But IFa clearly has a functional motivation. The absence here of infinitival to (otherwise mandatorily present unless the absence is sanctioned by (W)BIC) signals the presence, as governing verb, of an element that is universally necessarily finite, that has no non-finite distribution, that is a "core" modal.

IFb, as formulated, seems to lack such a functional motivation. But consider the consequence of the presence of another complement that leads to violation of IFb, and thus to presence of a bare infinitive if BIC is satisfied. BIC is frustrated if another complement than the infinitive is absent. Now, this other complement is again an element shared by the main and the subordinate clauses, but in this case it intervenes between main verb and infinitive. In (20b), for instance, the butler is part of the semantic valency of the infinitive verb, but it is also the "object" of the main verb, as indicated e.g. by the pronominal morphology (He saw him/*he serve drinks every day). In the case of the corresponding passive the shared element is separated physically from the infinitive, as shown in (8a), as it is in (21):

8) a. The butler was seen/heard *(to) leave.

It is only in the active that the main verb, shared element and infinitive can form a continuous sequence. The presence in (20b) etc. of the shared element between the main verb and the infinitive enhances the "intimacy" conveyed by the absence of to. It is this presence and the absence of to that together mark iconically the co-temporality of reference of the two verbs. The passive structure, despite its own non-discreteness (shared "subjects"), destroys this more radical possibility of iconicity. And the default to appears despite BIC being satisfied. It is thus not non-discreteness itself that is important for the presence of the bare infinitive, but non-discreteness of an element at the clause boundary that permits iconicity. Even in the case of IFb the syntactic behaviour observed has a functional, indeed iconic, basis: in this case the failure to achieve iconicity.

2.2. The basic distribution of the periphrastic infinitive

Other non-finite complements of verbs are associated with the default:

Infinitive Default

The infinitive is periphrastic.

The default applies to those infinitival complements that do not fall under the co-temporal punctual factuality generalisation (W)BIC or IF. These are constructions involving main verbs that are not necessarily finite and which take infinitive constructions that are associated with distinct interpretations, particularly as concerns the complements, from those associated with (W)BIC, as illustrated by the exclusion of the examples of (21) (though they also conform to IFb, of course). This emerges from a consideration of the various non-finite complement types familiar from descriptive grammars of English, and I shall therefore not dwell here on the character of these interpretations, except to try to establish that they do not conform to (W)BIC.

Thus, with many verbs, involving either "raising" or "control", the periphrastic infinitive involves an irrealis, not factual complement:

24) a. He expected (the butler) to leave.

b. He urged the butler to leave.

c. He intended (the butler) to leave.

These main verbs also take finite complements where irrealis is signalled by the presence of a modal (would/should). With others the periphrastic infinitive complement is habitual, or represents a capacity:

25) a. He liked (the butler) to swim. b. He taught me to swim.

In so far as these take finite complements, these are not marked as irrealis (He liked it that the butler swam).

With the -ing-complement we can often provide a progressive interpretation, particularly when the verb allows a choice of non-finite complement. The -ing-complement may be habitual or not, as in (26), whereas (25a) is only habitual: (4)

26) a. He liked (the butler) swimming in those days.

b. He liked (the butler) swimming for once.

Enjoy insists on the progressive, habitual or not:

27) a. * He enjoyed (the butler) to swim.

b. He enjoyed (the butler) swimming (for once/in those days).

We return in a moment to other functions of the -ing-form.

Inceptives (as in (28)), take all of the non-finites surveyed so far, as does the anti-terminative (continuative) of (29):

28) a. He started to rise (but couldn't).

b. He started to get up early.

c. He started walking towards the door.

d. He started rising early.

29) a. He continued to rise (for a moment).

b. He continued to get up early.

c. He continued rising (for a moment).

d. He continued getting up early.

On typical interpretations, (28a) is irrealis, (b) is habitual, (c) is progressive and (d) is progressive habitual. Interpretations of (29) differ from those of the corresponding sentences in (28) primarily in that the anti-terminative itself is, as the name suggests, inherently non-punctual.

The presence of -ing may also be associated with the (sometimes overtly expressed) presence of a source or "away from" (negatively oriented) relation, as with terminatives: (5)

30) a. He prevented/stopped me (from) swimming.

b. He stopped/ceased (*/?from) swimming.

With remember/forget/regret, the -ing-form in (30a) has past time reference, relative to the main verb, and this may be marked overtly, as in (30b):

30) a. He remembered (about) the butler swimming.

b. He remembered (about) the butler having swum.

The construction may also in some cases be marked with a preposition, but in this case associated with the whole complex following the main verb. Let's look now at the other perception/cognition verbs.

Stative cognitives introduce a predication that is at least provisionally true for the "cogniser"; and this is marked by a periphrastic infinitive that is preferably stative itself, particularly a be construction, including progressives:

32) a. He believed the butler to know the answer.

b. He believed the butler to be sincere.

c. He believed the butler to be lying.

Habitual infinitives are marginal, and a finite complement is preferred:

33) a. ?*He believed the butler to go sailing at weekends. b. He believed that the butler went sailing at weekends.

But they are much better in the passive, as in (34d):

34) a. The butler was believed to know the answer.

b. The butler was believed to be sincere.

c. The butler was believed to be lying.

d. The butler was believed to go sailing at weekends.

e. It was believed that the butler went sailing at weekends.

Simple punctual interpretation of the infinitive construction are excluded. Know conveys that the speaker also holds the subordinate predication to be true, so factual:

35) a. He knew the butler to suspect the answer.

b. He knew the butler to be sincere.

c. He knew the butler to be lying.

d. ?*He knew the butler to go sailing at weekends.

e. He knew that the butler suspected the answer/was sincere/lying/went sailing at weekends.

Again the passive of the habitual of (35d) is felt to be better--cf. (36d):

36) a. The butler was known to suspect the answer.

b. The butler was known to be sincere.

c. The butler was known to be lying.

d. The butler was known to go sailing at weekends.

e. It was known that the butler went sailing at weekends.

But again punctuality is excluded. As illustrated, all these verbs also take finite complements, not marked as irrealis.

One, at least, of the (physical) perception verbs share this pattern when used as a cognitive, except that it does not easily allow infinitival cognitives, as in (37a):

37) a. *He felt the butler to suspect the answer.

b. He felt the butler to be sincere.

c. He felt the butler to be lying.

d. ?*He felt the butler to interrupt too often.

e. He felt that the butler suspected the answer/was sincere/lying/interrupt too often.

So too in the passive:

38) a. *The butler was felt to suspect the answer.

b. The butler was felt to be insincere.

c. The butler was felt to be lying.

d. The butler was felt to interrupt too often.

e. It was felt that the butler suspected the answer/was sincere/lying/interrupted too often.

One difference from the other cognitives is that in these cases the subordinate predication must introduce a suitable source of emotion; hence the change in the choice of the subordinate verb in (d) compared with the (d) of previous such sets.

Non-stative cognitives are associated with a subordinate predication that comes to be true for the cogniser; and they typically do not take non-finite complements:

39) a. * He realised the butler to suspect the answer/be sincere/lying/to go sailing.

b. He realised that the butler suspected the answer/was sincere/lying/went sailing.

There are also some verbs involving the gaining of physical perception (usually necessarily of a location, in the case of the verbs in (40a)) that can also be used as non-stative cognitives, as in (41):

40) a. He discovered/found the butler's hiding place.

b. He perceived the butler's hiding place.

41) a. He discovered that the butler was sincere.

b. He found (out) that the butler was sincere.

c. He perceived that the butler was sincere.

But, unlike realise, they can also take infinitives:

42) a. He discovered the butler to be sincere.

b. He found the butler to be sincere.

c. He perceived the butler to be sincere.

But again punctuality is excluded. In their use as physical perception verbs, find and perceive take only the -ing-form, not the bare infinitive (unlike see etc.), or a periphrastic one:

43) a. * He found/perceived the butler (to) leave early. b. He found/perceived the butler leaving early.

Whereas discover takes only nominals, as in (40a). All of these non-bare-infinitive complements fail to fulfil the requirements for (W)BIC. These requirements, together with the grammatical constraints of IF, regulate the occurrence of the bare infinitive, and over-rule the default option, the periphrastic.

3. Let

Let conforms to WBIC, and so can (as a causative) take a bare infinitive, and indeed does so. This property, which is typical of it throughout its history, underlies some of the other, more idiosyncratic developments it undergoes.

One development we took a note of in [section] 1 was the emergence of an idiomatic formation based on let + infinitive, illustrated in (11b), repeated here, together with the regular causative (11a), for ease of reference:

11) a. He let the butler go.

b. He let go (of) the butler.

The infinitives which participate in this idiom are very few and the combinations are mostly obsolescent or obsolete, but they are semantically homogeneous. Thus we have let slip, let drop, let pass, let run, let fly, let be, though in some cases the let and the infinitive need not be adjacent. We also find some let-formations with no overt verb of motion, such as let away, let down, and also with adjectives, such as let alone, let loose, again, in both bases, with the possibility of separation. Some of these are already found in Middle English (Fischer 1990: 237). (6)

All of these idioms, as well as being motion(-verb)-based, have the (sometimes metaphorical) sense of 'release from a grip'. And this is why the directional complement, including the source, in (12b), unlike in (12a), is unacceptable--"direction from" (source) is incorporated in the meaning of the idiom:

12) a. He let the butler go away/to Paris/out of the room.

b. * He let go the butler away/to Paris/out of the room.

These are full lexical idioms: each combination of let + infinitive is non-compositional; but they all share this particular additional component of meaning beyond the "normal senses" of the component parts. They are all manifestations of the "let + infinitive idiom". They are thus not totally idiosyncratic in interpretation. But they have progressed, nevertheless, in terms of idiomatisation, beyond the "syntactic specialisation" exhibited by other residues of "subject-verb inversion" such as the (again directional, or locational) constructions illustrated by Here comes Marybeth and Up drove Uncle Jack.

The "object" in (11b) is "negatively oriented" with respect to the action of the verbal idiom: it is released from, out of some grasp, concrete or abstract. The marking with of in (11b) appears to be a late-ish development. Consider e.g. Thackeray's (The English humourists, chapter on Steele) Hill let go of his prey sulkily. It is not found consistently with the other instances of the let + infinitive idiom. But elsewhere in English we find such "negative orientation" marked by of where the source is again not given overt expression:

44) He disposed of the garbage.

We also find of in other circumstances where (for whatever reason) overt expression of a "negative orientation" is (otherwise) lacking, as in (45a):

45) a. He robbed/deprived the butler of the uniform.

b. He stole the uniform from the butler.

Compare with (45a) the overt expression of the source in (45b). (7)

This idiomatic development with let was favoured by the course of its regular development as a causative. And, conversely, we can associate the restriction on (passivisation of) the regular construction illustrated by (7b) vs. (7a), again repeated, with the development of the idiom:

7) a. *The butler was let (to) leave.

b. The butler was made *(to) leave.

Let us now look a little more closely at this history.

Fischer (1990: ch.4, [section]2.3.1) traces a split in the Old English ancestor of let in the Middle English period. According to her account, the causative differentiates itself syntactically from other let verbs meaning 'leave (off)' or 'lend, rent', as well as 'hinder' (from OE lettan rather than loetan). Fischer associates the causative with various infinitive constructions. The basic ones are those given in (46), where the functional labels refer to the role which the arguments concerned would have in finite sentences containing the verb whose infinitive form they precede--in all cases the argument is marked as accusative (where this is distinctive):

46) a. let + "subject" + infinitive

b. let + "object" + infinitive

(These correspond to what Fischer calls respectively "the 'subject'-construction" and "the 'object'-construction" (1990: 239).) Fischer (1990: 239-40) illustrates these with the respective sentences from the Confessio Amantis and the Brut in (47), among other examples:

47) a. I pray you let me noght mistime Mi schrifte (CA 1:220-221).

'I pray you let me not mistime my confession'

b. they took AEstrild and Abren and let them bind ... (Br: 1244).

'they took AEstrild and Abren and let them bind'

Either "subject" or "object" may be omitted (giving what Fischer (1990: 239) calls "the 'pure' infinitive construction"):

48) a. This Leonin lets ever aspie, And waiteth after grete beyete (CA 8: 1432-1434).

'this Leonin lets ever spy, and waits after great profit'

b. pe king lette blawen; and bonnie his ferden (Br: 4016).

'the king let blow; and assemble his army'

Type (46a) includes "subject" (Fischer 1990: 240) arguments that are the "subjects" of passive infinitives (in what Fischer (1990: 239) calls "the passive infinitive subject-construction"):

49) If that my litel Sone deie, Let him be beried in my grave (CA 3: 292-294).

'if that my little son die, let him be buried in my grave'

Much of Fischer (1990: [section]2.3) is concerned with various developments in Middle English that, to simplify, saw the causative "object" construction of (4b) eventually "replaced" by such passive "subject" variants of (46a) as are exemplified by (49). The possibility for the constructions of (46) to appear with no "subject" or "object", illustrated by (48), after a long Middle English persistence, has also disappeared from Present-day English (Fischer 1990: [section]2.3.1.3). However, another variant of the basic constructions of (46) is more immediately relevant to our present concerns.

Fischer observes (1990: [subsection]2.3.1.1-2) that both the "subject" of (46a) and the "object" of (46b) may appear in post-infinitive position, if the argument concerned is "a so-called heavy constituent, or when it is a clause" (1990: 241). This is illustrated, for the (46a) construction, by (50):

50) Heo letten to-gliden; gares swipe scarpe. Heo qualden pa Frensce (Br: 877-878).

'they let go darts very sharp; they killed the French'

Fischer notes that the order of (50) is relatively frequent in the Brut, and she goes on to observe further that "Let is also the only verb that allows a post-infinitival subject when this subject is neither a heavy constituent nor a clause" (1990: 250). This is illustrated by (51a):

51) a. Cnihtes fuseo me mid; leteo slaepen pene king (Br: 368).

'knights come me with, let sleep the king'

b. Cniptes comep mid me let pane king slepe (Br: 368).

'knights come with me, let the king sleep'

Compare the pre-infinitival "subject" of the Otto version in (51b).

According to Fischer (1990: 238), "the consistent use of the bare infinitive further marks the near auxiliary status of let in Middle English". The term "near auxiliary" involves some equivocation, but this development certainly may reflect "grammaticalisation". Analogous perhaps is the modern construction in (52) involving go (and, marginally, come):

52) a. Go eat/dance/party/demonstrate/kill yourself.

b. Let's go eat/dance/party/demonstrate/kill ourselves.

c. We want to go eat/dance/party/demonstrate/kill ourselves.

But this latter construction is much more obviously "grammaticalised", in so far as the first element is limited to the "bare form" of this go verb, the form that in (52b-c) is itself found complementing a periphrasis. We find this form as the imperative and after the hortative let's or the infinitival to, as in (52). (8) Presumably the "bare form" of go in (52b) is the bare infinitive, as it is historically. We can also find the "bare form" as the bare infinitive following perceptives and causatives (53), as well as following modals (54):

53) I saw/made/let them go eat.

54) We'll go eat.

But overtly inflected forms of go are awkward in this construction, as are even those apparently "bare" forms that are in tense or person-number contrast with overtly inflected forms:

55) a. * They went eat.

b. ?* They go eat (regularly).

If the verb form following go is also a bare infinitive, then (52-53), even if they are interpreted as satisfying BIC in order to account for its (non-periphrastic) form, apparently violate IFb: "The infinitive uniquely complementing verbs is periphrastic". The supplementary generalisation seems to be that, with this "grammaticalised" go, the bare infinitive can follow its "bare form" (on the restrictive understanding that excludes go in (55b) as such).

It may be that go then is a new auxiliary, which has a distinctive distribution, one which, as with have and be, includes both finite and non-finite occurrence.

As a finite, it is imperative; otherwise it is part of either a bare or a periphrastic infinitive construction depending on what it complements. In itself it is thus always "bare". It would then fall under an extension of IFa, which would apply to verbs that are either always finite or always "bare", including when finite. However that may be, Middle English let doesn't appear to have achieved the level of "grammaticalisation" illustrated here. Rather, we have something akin to the "syntactic specialisation" I attributed to Up drove Uncle Jack and the like. The bare passive infinitive of (5a) comes to be another such "specialisation" associated with let:

5) a. He let the butler be replaced.

But it is the early development of the former "specialisation" that is crucial to the development of the idiom.

Fischer argues further (1990: 252) concerning let that "there is no evidence for earlier Middle English that let + infinitive could be looked upon as an idiomatic unit which would account for the non-separation of matrix verb and infinitive. (9) This seems indeed a later development that becomes visible only in Malory". And her note 28 (1990: 293) points out that "in the Brut and Gower, the infinitive following let can be any verb. Only a few verbs occur more than once after let, such as gliden and blawen in the Brut, make and sende in Gower. There clearly is no fixed pattern here". However, it may be that the (relative) commonness of the "specialised" uninterrupted let + (bare) infinitive construction revealed by some earlier texts "set the stage for" the development of the idiom: juxtaposition of the two elements encouraged idiomatic reinterpretation of the sequence in cases where the infinitival verb allows the 'release from a grip' interpretation associated with such as (11b), as proposed initially in this section.

The uninterrupted let + infinitive sequence is thus appropriated by the idiom. This idiom at various times has not been limited to just one infinitival verb (such as go). The idiom is thus not entirely idiosyncratic; it begins as a sub-pattern with its own semantics. This is of course disguised and undermined by the increasing isolation of let go as a surviving manifestation of the idiom. The idiom re-establishes a sense of the non-causative verb let (sense (1b) in the entry for let, v. (1) in the OED), which we can illustrate by (56):

56) Tho was ther manye teres lete (CA 2: 3228).

'then was there many tears let'

In order to do this, the idiom reinterprets the adjacency of let + infinitive as constituting an independently meaningful unit. (10)

Likewise, to continue in this anthropomorphising vein, the idiom has apparently appropriated the corresponding passive, which again involves an uninterrupted sequence:

57) The butler was let go (of).

Let is consistently associated with the bare infinitive in Middle English (Fischer 1990: 238, 292, note 15). Of course, as a passive of the regular causative let, such a construction as (57) involving a bare infinitive would come to violate IFb: "The infinitive complementing non-finite verbs is periphrastic". However, the idiom regularises (57) by associating the butler with the active "object" of a verbal idiom in (11b), with regularly corresponding passive in (57). The idiom thus "kidnaps" the passive construction.

For the regular causative, on the other hand, a passive version of let with a periphrastic infinitive (compatible with IFb) is late to develop, and it fails to establish itself, as witness the anomaly we have already noted of (7b) vs. (7a):

7) a. * The butler was let (to) leave.

b. The butler was made * (to) leave.

There are sporadic examples of let + periphrastic infinitive through the modem period, and even, latterly, indeed then particularly, of passivised let + periphrastic infinitive (illustrated under sense 12 in the entry for let, v. (1) in the OED). And the latter would of course qualify for IFb. But causative let ultimately fails to retain a passive which allowed an infinitive complement construction of either type (bare or periphrastic).

4. Conclusion

If something like the preceding overall scenario is plausible, the history of let both reflects, along with other causatives, the development of the BIC and IF and also involves an idiomatisation that interacts with these infinitive-marking developments and with the variation in the position of the infinitive, either adjacent or not to the main verb. What I have looked at briefly in the preceding represents only a few of the specialisations, "grammaticalisations" and idiomatisations that characterise the history of let, which include idioms of the "letting blood" type, as exemplified in (56), and the development of the hortative of Let's go etc. in (52a). They serve to illustrate, however, something of the interplay between generalisation and idiomatisation in the development of syntax. Syntax is not insulated from idiomatisation.

This is part of another, even more general, conclusion that emerges here, particularly from [section] 2. This is that, to the extent, again, that the accounts offered here offer the basis for an understanding of the phenomena invoked, they illustrate the semantic and functional basis for even the most idiosyncratic and apparently arbitrary aspects of syntax. With the major syntactic generalisations, this basis is obvious: the syntax of verbs and nouns in general reflects the semantics of the prototypical members of the class: verbs are the unmarked predicators and nouns the unmarked arguments because their prototypical members represent event types and entity types, respectively. Failure to recognise this notional basis confines and condemns syntax to (at best) the sterility of simply registering apparently arbitrary generalisations, generalisations that, in the proper perspective, are not indeed arbitrary at all. In the present case, what might seem to be a set of trivial, idiosyncratic observations concerning the distribution of infinitives in Present-day English can be seen as involving phenomena which have a non-arbitrary semantic and functional basis. Assuming arbitrariness and autonomy in syntax, and the banality of the consequences, should be the last resort for the grammarian.

REFERENCES

PRIMARY SOURCES

Benson, Larry D. (ed.)

1987 The Riverside Chaucer. (3rd edition.) Boston, Mass.: Houghton Mifflin.

James, Henry 

1905 The golden bowl. (3rd edition.) London: Methuen.

Thackeray, William M.

1879 The English humourist. London: Smith, Elder & Co.

SECONDARY SOURCES

Blake, Norman (ed.)

1992 The Cambridge history of the English language, H: 1066-1476. Cambridge: Cambridge University Press.

Britton, Derek (ed.)

1996 English historical linguistics 1994: Papers from the 8th International Conference on English Historical Linguistics. Amsterdam: Benjamins.

Callaway, Morgan, Jr.

1913 The infinitive in Old English. Washington, D. C.: Carnegie Institution of Washington.

Fischer, Olga C. M.

1990 Syntactic change and causation: Developments in infinitival constructions in English. (Amsterdam Studies in Generative Grammar 2.) Amsterdam: [No indication of publisher.]

1991 "The rise of the passive infinitive in English", in: Dieter Kastovsky (ed.), 141-188.

1992 "Syntax", in: Norman Blake (ed.), 207-408.

1995 "The distinction between bare and to-infinitival complements in Late Middle English", Diachronica 12: 1-30.

1996 "Verbal complementation in Early Middle English. How do the infinitives fit in?", in: Derek Britton (ed.), 247-270.

Fischer, Olga C. M.--Ans van Kemenade--Willem Koopman--Wim van der Wurff

2000 The syntax of Early English. Cambridge: Cambridge University Press.

Hogg, Richard M. (ed.)

1992 The Cambridge history of the English language. The beginnings to 1066. Cambridge: Cambridge University Press.

Kaartinen, A.--Tauno F. Mustanoja

1958 "The use of the infinitive in A Book of London English 1384-1425", Neuphilologische Mitteilungen 59:179-192.

Kastovsky, Dieter (ed.)

1991 Historical English syntax. Berlin: Mouton de Gruyter.

Kisbye, Torben 1970 An historical outline of English syntax. Part I. Aarhus: Akademisk Boghandel.

Mitchell, Bruce

1985 Old English syntax. Vol. I. Oxford: Oxford University Press.

Murray, James A. H.--Henry Bradley--William A. Craigie--Charles T. Onions

1888-1933 The Oxford English dictionary. London: Oxford University Press.

Quirk, Randolph--Jan Svartvik

1970 "Types and uses of non-finite clauses in Chaucer", English Studies 51: 393-411.

Scheurweghs, George

1959 Present-day English syntax. London: Longman.

Traugott, Elizabeth Closs

1992 "Syntax", in: Richard M. Hogg (ed.), 168-289.

Zandvoort, Reinard W.

1964 A handbook of English grammar. Groningen: J. B. Wolters.

JOHN M. ANDERSON

Methoni, Greece

(1) Help is variable in its selection of infinitive: Help the butler (to) leave etc. In itself it illustrates, in addition to the discrepancy between (1) and (2), a "hesitancy" about the causative class. This "hesitancy", compared with the perceptives of (3) below, is something we shall have to take into account in what follows. Bid is a further bare-infinitive taker, but is now obsolescent. There are other marginal optional takers of the bare infinitive, as illustrated by I have never known (found him lose his temper (see, e.g., Zandvoort 1964: [section]18). These are associated, like the perceptives, with particular temporal/aspectual restrictions. Scheurweghs (1959: [subsection] 382-387), for example, also provides a survey of uses of the bare infinitive.

(2) Fischer et al. interpret (47), discussed here in [section] 3, as involving not "raising" but "control" of an empty "subject" by the (distinct) "object" of the let verb:

(47) a. I pay you let me noght mistime Mi schrifte ... (CA 1: 220-1)

'I pay you let me not mistime my confession' ...

b. they took AEstrild and Abren and let them bind ... (Br: 1244)

'they took AEstrild and Abren and let them bind' ...

In terms of such an analysis the empty "subject" of bind in (47b) is "controlled" by an "object" which is itself never realised overtly. Surely the "subject" of bind is simply nonspecific? The historical evidence does not support the positing of such an arcane analysis.

(3) The assumption with perception verb + infinitive of contemporaneousness can be used to telling literary effect. Consider e.g. the "large words" with which Mrs. Assingham in Henry James' The golden bowl describes her "large visions" to her interlocutor, Maggie, concerning the future of Maggie's father and stepmother (James 1905: 503):

"I see the long miles of ocean and the dreadful great country, State after State--which have never seemed to me so big or so terrible. I see them at last, day by day and step by step, at the far end and I see them never come back. But never--simply ..."

I have underlined the crucial, culminating "see"; the italicisation elsewhere is in the original. Much of the book is indeed about what Fanny Assingham "sees". Henry James might well have sub-titled it "What Fanny Saw".

(4) With some verbs the choice of -ing or to + passive (including for some speakers the "truncated" passive) is a matter of overt voice (and dialectally significant), not an aspectual matter: His face needs/deserves/wants/requires cleaning/(to be) cleaned. In either case the construction is irrealis. Also irrealis are the complements of -ing-forms following fancy, suggest, for example, as well as complements of some verbs that take either construction, such as intend or plan.

(5) We can also associate "negativity" of some sort with some other verbs that take the-ing-form, like deny, dislike, put off and miss.

(6) Help too shows non-verbal (and typically directional) as well as bare-infinitival complementation, as in help off/down/along/.... This has also been characteristic of modals at various periods and in various dialects (I"ll away to the shops).

(7) The of is normally obligatory with post-infinitival pronouns: Let go *(of) it! The alternative is the ambiguous pre-infinitival position: Let it go!

Of course, of as a marker of verbal complements has other functions, as with speak/talk/think of. To the extent that these mark a variant of partitivity (they are non-bolistic), they are, however, related to sources and "negative orientation" (as reflected in, say, the behaviour of the Finnish partitive case).

(8) I have encountered too Come go eat! There is a similar construction to the present one which involves however the -ing-form: Let's/we want to go/come swimming. But the form of the complemented verb is a lot more variable, not obviously "grammaticalised": We went/came swimming etc.

(9) A number of other idioms spring from this juxtaposition: cf. e.g. Go fly a kite, Go fetch!, make believe, make do etc., as well as the let formation involved in Live and let live. See Zandvoort (1964: [section] 17).

(10) We have also noted the lack of a passive with causative have. Recall (7c): (7) c. *The butler was had (to) leave.

In this respect it patterns like the "event-recipient" have of I had an extraordinary thing happen to me (Zandvoort 1964: [section] 18; Scheurweghs 1959: [section] 384). Have in general seems to reject passivisation, except in idiomatic phrases such as A good time was had by all, You've been had.