This paper contrasts finite and non-finite complement constructions containing the matrix verb promise. Using data from the British National Corpus, I show that when no explicit mention is made of the promissee the non-finite form of complement is overwhelmingly preferred to its finite counterparts. The exact opposite is the case when the promissee is mentioned between the matrix verb and the complement clause. In addition, the promiser in the x promise y to infinitive construction is almost always pronominal. I suggest that these two facts, the dispreference for the to infinitive form of complement when the promissee is mentioned and the pronominal encoding of the promiser in such cases, are both related to the very rarity of this form of construction in English. Data is adduced showing that another rare construction, the so-called possessive - ing construction, also occurs with a disproportionate number of pronominal subjects. It is suggested that the preference for pronominal subjects in these constructions may be related to a wish to reduce the overall processing complexity of the predications in question.

There are some matrix verbs in English that occur in two or more constructions which are almost, if not completely, synonymous. [1] One example of such a verb is help, which may be followed by both the bare and the to form of the infinitive. In many instances, the difference in meaning between pairs of sentences containing help (bare) infinitive and help to infinitive (see, for example, Mair 1995) is very subtle and difficult to pin down. The focus of the present article is another such case of near-synonymity. In this case one of the constructions in question is much rarer in occurrence than its counterparts. The constructions to be explored all contain the matrix verb promise followed by either a finite or a non-finite complement clause. The finite clause may or may not contain the complementiser that. In addition there may or may not be explicit mention of the promissee. There are thus six constructions in all. These are listed in (1) to (6).

There are few matrix verbs in English that have attracted so much attention from scholars as promise. Note, for example, that some of the articles from which examples (9) – (14) are taken have been very influential indeed in the development of various paradigms in the last quarter of the last century. Linguists have shown a particular interest in promise constructions with a to infinitive complement, as in (7) and (8), taken from the British National Corpus (hereafter BNC). [2]

The popularity of the promise to construction(s) is easy to understand, as the construction in (4) and (8) is unique in English, being the only same-subject (equi-subject) construction containing a to infinitive complement to allow the explicit mention of the addressee between the matrix verb and the complement clause. The discussion below opens in section 2 with some examples of how the construction in (4) has been exemplified by a handful of grammarians of various theoretical persuasions, and then goes on to describe how it is actually used in present-day British English as evidenced by the material in the BNC. The promise to constructions are then compared to the related same-subject finite complement constructions illustrated in (2), (3), (5) and (6). Section 3 contains a short presentation of the relative incidence of various types of subject in the three constructions with explicit mention of the promissee. Section 4 looks at the promise constructions in spoken and written English and section 5 at the form and discourse status of the promissee in the X promise Y constructions. Section 6 contains a discussion of the facts presented in the previous four sections and some data from the BNC about another rare construction, the so-called possessive -i ng construction. Finally, section 7 contains a short summary and some suggestions for further research.

The X promise Y to construction has received a lot of attention in the literature on complementation. As for its exemplification, it turns out that this is most commonly done in the form of sentences in which both X and Y, both the promiser and the promissee, are cited in the form of full nominals. Examples (9) – (14) may be taken as representative in this respect.

Examples (9) – (14) are all discussed at some length in the articles from which they are taken. Moreover, as was mentioned above, some of these articles have been very influential indeed in the development of linguistics over the last twenty five years. All six grammarians from whose work examples (9) – (14) are taken encode the promiser, and five of six the promissee, in the form of a proper noun. The sixth, Bach, encodes the promissee as a definite NP. We will explore below to what extent these examples are similar to the products of users of the language as these are recorded in the BNC.

Like most verbs, promise is polysemous. It is most commonly used to encode a commitment on the part of the subject to carry out some course of action. This sense is exemplified in (15) and (17). In another sense, which we find exemplified in (16), promise to x means something like ‘looks like x-ing’. In a third sense, exemplified by (18), promise means ‘assure’.

The present article is concerned solely with contexts in which promise functions as a communication verb (i.e. a matrix verb used to encode an act of communication). Thus utterances like (16) will be ignored. Nor will utterances such as (18), in which the subject of the complement clause differs from the subject of the matrix verb, be dealt with. The discussion will thus be restricted to constructions like (15) and (17). These may be described as same-subject communication constructions, i.e. constructions encoding an act of communication in which the subject of the complement clause is identical in reference to the subject of the matrix verb. (Included are utterances in which the subject communicates with him- or herself. In these cases the promissee is realised as a reflexive pronoun.) In non-finite clauses, such as (19) the promiser is only mentioned once; in finite clauses, such as (20) and (21), he or she is mentioned twice.

These three constructions may be said to be near-synonymous in the sense that they all encode the expression of a commitment on the part of the subject to carry out a stipulated course of action at some time in the future. In (19) – (21), for example, the promisers commit themselves to being (or not being) in a particular location at or by a particular time. This is not to deny that there may be subtle differences of meaning between the constructions, and we will uncover some pragmatic differences in the course of this article. Nevertheless, it does certainly seem to be the case that speakers may choose any one of these three partially synonymous constructions to convey much the same point to their addressees.

Taken together, the same-subject communication constructions account for just over a quarter of all utterances containing the verb promise in the BNC. [3] The promise to construction is by far the most common of them, as may be seen in figure 1, which depicts the incidence of the three forms without the explicit immediately post-verbal mention of the promissee.

As may be seen in figure 1, there are over 1,100 tokens of promise to in the corpus as opposed to fewer than 200 of the two finite constructions. The predominance of the infinitive constructions is predicted by Hawkins’ principle Minimize Domains, which states:

The human processor prefers to minimize the connected sequences of linguistic forms and their conventionally associated syntactic and semantic properties in which relations of combinations and/or dependency are processed” (Hawkins 2004:31)

Indeed Hawkins himself states of this construction: “If no other constituent intervenes [between promise and to], the infinitival should be highly preferred” (Hawkins 2004:157). The picture changes completely, however, when we look at constructions containing explicit mention of the promissee, as in (22) – (24).

Details of the number of tokens in the BNC of the three same-subject promise constructions with explicit mention of the promissee are contained in figure 2.

As is shown by figure 2, the construction containing the infinitive is by far the least popular of the three when the promissee is mentioned between the matrix verb and the complement clause. Rohdenburg (1999: 105) notes a similar difference in distribution in a corpus of texts from the American News Magazine Time and states that what he calls “the dramatic effect produced by personal objects … may come as a bit of a surprise”. There is no doubt that Rohdenburg is correct in describing the effect as dramatic. The change in the relative distribution of the three complement types in the BNC is statistically significant at the level of p=0.0000. [4] There are two reasons why it may come as “a bit of a surprise”. The first is the fact that the infinitive construction without mention of the promissee is so common. The second reason why it may come as a surprise is that grammarians are so used to encountering the construction in (22) in the literature that they may have taken it for granted that it is common among the products of the language. [5]

Why should the finite complement form be so much more favoured than the non-finite one in utterances containing an explicit promissee? We will return to this question in section 6. At this stage some general considerations may suffice. The insertion of the promissee between the matrix verb and the complement clause does not lead to greater complexity in the complement clause per se. However, there is no doubt that the simple promise to construction in (1) is less complex than the finite ones in (2) and (3) according to Hawkins’ principle of minimising domains (see, for example, Hawkins 2004: 157). Insertion of the promissee does not alter the relative complexity of the three constructions. It merely adds one more element to each of them. In so far as a hearer must access fewer morphemes in order to discern the syntactic structure, the infinitive construction is still less complex structurally than the two finite ones. Thus the principle of minimizing domains cannot explain the preference for the finite construction when the promissee is explicitly mentioned. [6] But there is a second idea to be pursued here: Rohdenburg (1995), having addressed this sort of issue, formulated a principle called the Complexity Principle. It goes as follows:

The less directly the dependent clause is linked to its superordinate clause, or the more complex the dependent clause turns out to be, the greater is the need to make its sentential status more explicit. (Rohdenburg 1995: 368)

Since, although the insertion of the promissee does not lead to greater complexity in the complement clause, it does lessen the degree of directness in the link between the superordinate and subordinate clauses, Rohdenburg’s principle predicts, correctly, that a finite form would be favoured in contexts like this. It should be noted here that Rohdenburg’s Complexity Principle is formulated with respect to varying degrees of complexity in strings to the right of the matrix verb in the canonical order of utterance. In particular, it focuses on the complexity in the subordinate clause and in the complementation link between this and the matrix clause. The principle does not address directly the question of whether degrees of complexity in constituents to the left of the matrix verb may influence the choice of complement form. In the next section we will examine the question of whether such complexity, as evidenced by the choice of full nominal as opposed to pronominal subjects, may also influence this choice.

Figures 3 – 5 contain totals for pronominal and full nominal promisers and promissees in all three constructions with explicit mention of the promissee. As may be seen in the figures, the promiser is overwhelmingly pronominal in all three. Moreover, it is only the construction containing the complementiser that that is represented by more than a single example in which both promiser and promissee are full nominals. (25) and (26) are examples of this construction, (27) is the only instance of two full nominals with X promise Y to.

Note that in (27) the promiser, the footballer in question, has just been mentioned in the newspaper headline. Indeed, it is mainly in this sort of context that one repeats a proper noun as subject of an immediately subsequent sentence. In most other contexts one would employ a pronoun. (Other exceptions are some forms of legal document.) There would seem to be a very strong tendency for language users to avoid employing the X promise Y to construction when X cannot be instantiated pronominally. The same point may be made with respect to the X promise Y Ø construction. The X promise Y that construction would therefore appear to be overwhelmingly favoured when the promiser is a full nominal. In this connection one may note that The Cambridge Grammar of the English Language states that “[Liz promised me to phone at six] is rather marginal: many speakers find this unacceptable and it would be much more usual to use a finite complement here” (Huddleston & Pullum 2002: Ch. 14, 1229-1230). [7] One point the authors of the Cambridge Grammar do not address is whether this feeling of unacceptability is influenced by the encoding of the promiser by a proper noun rather than a pronoun or by the mere presence of a promissee.

The full nominal subject in (27) shares with the pronominal subjects of the X promise Y to construction the property of being given in the immediate discourse context. However, the fact that promisers in some of the other promise constructions are more often full nominals does not, of course, mean that these necessarily encode participants new to the universe of discourse. We must at least consider the possibility that the constraint with respect to new as opposed to given subjects which we have noted in the case of the X promise Y to and X promise Y Ø constructions (i.e. a preference for pronominal subjects) may be a property of the verb promise rather than the constructions themselves. Examples (28) – (31), containing tokens of the simple X promise to construction and the X promise Y that construction, show that this is not the case.

In (28) and (29), both of which instantiate the simple X promise to construction, the promisers are completely new to the universe of discourse. No previous mention has been made of most manufacturers in (28) nor of Major Anson in (29). Note the specifying post-modifier of Anson Airways in the latter. A similar specifying function is served by the apposed noun phrase the branch manager in (30). As for (31), which is an expanded version of (26), the promiser Pete Waterman is not mentioned in the previous 100 sentences. [8] Table 1 contains figures for given and new subjects in all six promise constructions. Practical reasons (the impracticality of trawling through almost 1,500 texts) dictate that given is to be understood in the rather narrow sense of having been mentioned in the immediate discourse context.

The fact that there are as many as 192 instances of tokens with new subjects in the data in table 1 indicates that the restriction to given subjects is not a property of the matrix verb promise. Nor is it a property of the X promise to construction. It does, however, appear to be a property of the X promise Y to and X promise Y Ø constructions.

At this point one ought to raise the question of whether the avoidance of the X promise Y to construction when X is a full nominal may be influenced by its occurring primarily in spoken genres. This question will be addressed in the next section.

Research on full nominal versus pronominal subjects has shown that there is a marked difference between written and spoken genres with respect to the incidence of the two types. (See, for example, Hindle 1981; Givón 1995: 51; Francis, Gregory & Michaelis 1999.) Spoken genres contain more pronominal subjects. This tendency is also reflected in the BNC. One must therefore ask whether the very high number of such subjects in the case of the X promise Y to construction may not be due to its occurrence in spoken contexts. Figure 6 contains percentages for all same-subject communication promise constructions in all written and genres on the one hand and spoken dialogue on the other.

An examination of all tokens of the promise constructions shows that five of the six constructions occur in the spoken dialogue part of the BNC. Details are given in figure 7.

As may be seen in figure 7, the only one of the six constructions not to occur in the spoken dialogue part of the BNC is X promise Y to. In other words, one of the two constructions with the highest incidence of pronominal subjects is the only one not to appear at all in dialogue. It differs markedly in this respect from the other one, the X promise Y Ø construction. There is thus no doubt that the high incidence of pronominal subjects in the X promise Y to construction cannot be ascribed to the medium of usage.

As regards the other five constructions they do occur in spoken dialogue, with pronominal subjects being in the majority. Indeed only two of them are represented in the spoken dialogue part of the BNC with full nominal subjects. These are X promise to and X promise that, exemplified by (32) and (33).

There are only six tokens in all that resemble (32) and (33) in having a full nominal subject. These represent 11% of all the tokens of promise in the six constructions in the spoken dialogue section. The corresponding percentage for full nominal subjects in the written part of the BNC is 38%. This is in line with what one would expect from the discussions of the interaction of medium and subject form in the literature mentioned at the start of this section.

One item remains to be investigated before we proceed to the discussion of the corpus data. This is the promissee. In the next section we will look at the form and discourse status of the promissee in the X promise Y constructions.

Table 2 contains figures for pronominal and given and new full nominal promissees in the three promise Y constructions. As was the case with table 1, practical reasons dictate that given is to be understood as mentioned in the immediate discourse context.

While the promiser in the X promise Y to construction is invariably given in the universe of discourse, the same is not true of the promissee. As may be seen in table 2, nine of the fifteen promissees are full nominals. Moreover, a majority of these do not encode given participants. (34) – (36) are typical examples of the construction with given promissees.

In (34), an expanded version of (30), the promiser and promissee are the two, maximally salient, discourse participants. In (35) the uncle in question has been mentioned by the promiser in her most recent contribution to the conversation, and in (36) Rosie-Posie or Rosette is present in the background, where she can be heard talking in the kitchen. In addition to examples such as these, the promissee is, of course, also given in the utterances in which he or she is identical to the (given) promiser. One such instance is cited as (37).

Two tokens containing promissees new (or at least relatively new) to the discourse, are cited as (38) and (39).

In both (38) and (39) the promiser is very salient in the discourse. In (38) Athelstan is the subject of three of the previous four predicates. In (39) the promiser, in the form I, is the subject of 8 of the previous 9 predicates. In both these utterances, however, the promissees, Benedicta in (38) and Peter in (39) are not given in the immediate context.

Both the X promise Y that and the X promise Y Ø constructions also favour given promissees. [9] In table 1 we saw that not one of the 100 tokens of the X promise Y Ø construction in the present study contains a promiser that is new to the immediate discourse context. In the case of the promissee, on the other hand, almost a third of the tokens are new. (40) – (42) are typical examples.

(40) occurs some twenty sentences into the sixth chapter of a novel: William Tanner is not mentioned in these twenty sentences. And in (41) and (42) the promissees are not mentioned in the previous 100 sentences. In all three examples the promiser is encoded by a pronoun, as is the case in 99 of 100 tokens of this construction.

To sum up, the form and discourse status – as opposed to the mere existence – of the promissee appears to exert little influence on the choice between a finite and non-finite complement form, which is in marked contrast to the form and status of the promiser in these same constructions.

Let us now take stock of the facts. We have seen that:

How are we to account for these facts? Rohdenburg’s Complexity Principle was cited in section 2. With respect to promise Rohdenburg speculates that

the replacement of [the infinitive complement type] by [the that complement type] may be attributed to a desire to avoid the processing complexity associated with an increasingly unusual choice of controller (Rohdenburg 1999: 105)

In other words, it may require more effort to process an instantiation of an uncommon construction type than a common one. Talmy Givón makes a very similar point with respect to experience types. He writes: “Frequent experience types are accorded more efficient, automated processing. Infrequent experience types are processed less efficiently” (Givón 1995: 63). That means that situations which are only rarely encountered in what Givón refers to as “external reality” require more effort to be communicated in language both from the point of view of the speaker and the hearer than do situations which we encounter on a regular basis. It is the contention of the present article that the same point may well apply to construction types. The controller of the complement predicate in the X promise Y to construction is maximally “unusual”, to borrow Rohdenburg’s expression, in that this is the only construction in English to contain an element in the syntactic position of Y in which it is X and not Y that fulfils the role of subject of the complement predicate. Its unusualness is cemented by the fact that there are hundreds of constructions of the form X matrix verb Y to where it is Y rather than X that functions as the subject of the complement clause predicate. As Hawkins puts it:

since verbs like promise are in a distinct minority compared with persuade-verbs, the finite complement could also be motivated by a desire to make clear that the subject of the subordinate verb is not the matrix object, as it usually is” (Hawkins 2004: 158)

If Rohdenburg’s (and Hawkins’) assumption that finite clauses are preferred to non-finite ones in order to avoid processing complexity is correct, then the avoidance of such complexity may also be the reason why one restricts the utilisation of an infrequent construction type to contexts in which the primary argument (the subject) is particularly salient in the discourse context. The speaker ensures thereby that a maximum degree of processing capacity is available to process the construction itself.

Of a total of 18 tokens of the X promise Y to construction in the BNC, 15 contain a pronominal promiser. Of the three exceptions one has already been cited as example (27). In its case we saw that the repetition of the subject in proper noun form in consecutive sentences was conditioned by the transition from a newspaper headline to the first line of the journalist’s article. The other tokens in which the subjects are full nominals are (43) and (44).

According to the BNC, the text from which (43) is taken is called “Warriors of Christendom”, written by John Matthews and Bob Stewart, and published in 1988. The actual author of the text in question, however, is Robert Southey. It is his 1808 translation of a Spanish text from 1637. Example (43) was therefore excluded from consideration in this study, as was (44) which is taken from Great Expectations and (45) which is a quotation from an eighteenth century gardening treatise.

This leaves 15 tokens of X promise Y to, in 14 of which X is realised by a pronoun. In the sixteenth token, example (27), it is realised by a proper noun. How should we account for these facts in a usage-based, bottom-up approach to grammatical classification? Recall that this construction is unique in English, so we cannot avail ourselves of analogy with other constructions containing two animate participants. Faced with such a rare construction, the only one of its type, there are, in theory, at least two ways, (a) and (b), of trying to account for the fact that its subjects are overwhelmingly pronominal.

The discourse reasons referred to in both (a) and (b) are not difficult to describe. With respect to the reasons mentioned in (a), there is no doubt that in all 15 tokens of X promise Y to, X is not only given in the universe of discourse, but is either first or second person or, if third, is extremely salient in the discourse. [10] This explains why it is normally realised by a pronoun. With respect to the reasons mentioned in (b), the only one encountered in the present study was the decision by a copy editor to state a topic in a newspaper headline and to restate this same topic in the first line of the body of the text.

Given a choice between these two approaches, the second one may appear more attractive, in the light of the genre restrictions that seem to apply to the realisation of the subject as a full nominal. However, one may question whether there is any need to choose between the two, whether the linguistic facts might not be better captured by a statement, such as (c), that makes reference to the current discourse status of the subject.

(c) would appear to account satisfactorily for all 15 tokens of the construction.

The X promise Y to construction is not unique in English in its tendency to occur with pronominal subjects. This tendency also applies, for instance, to the possessive - ing construction, illustrated in (46). The possessive -i ng construction also shares with the X promise Y to construction the property of having a near-synonymous counterpart, the accusative - ing construction, illustrated in (47).

These two constructions, hereafter referred to as poss. - ing and acc. - ing, often appear to be almost, if not completely, synonymous. This is certainly true of examples (46) and (47), both of which are the product of the same author. In these two examples the replacement of the accusative form of the complement clause subject by the possessive form, or vice versa, would appear to result in an equally idiomatic utterance. According to Biber et al. (1999: 750) in cases like this “the possessive option focuses attention on the action described in the ing-clause. In contrast, use of the objective form emphasises the person doing the action.” On the other hand one could argue that the more nominal structure of (46) may incline one to interpret the act of saying as more of a reified process than is the case in (47), where saying retains more of the ‘ongoing’ sense normally associated with the verbal - ing form. However, there is no doubt that in the case of both interpretations we are dealing with very fine meaning distinctions.

Of the two constructions, acc. - ing is much more common than poss. - ing. Biber et al. (1999: 750) note that “the objective case is used for the overwhelming majority (over 90%) of noun phrases occurring in the pattern verb + NP + ing -clause”. In a similar vein, Mair (2002: 112) notes that the “archaic type of gerund, with the notional subject in the genitive, is attested very rarely in present-day English”. Although both Biber et al. and Mair are undoubtedly correct in stating that the poss. -i ng construction is the less common of the two, both constructions are represented in the BNC.

To elaborate further on similarities with the promise construction; an investigation was carried out of constructions containing 15 matrix verbs that occur with both types of complement. The verbs in question are anticipate, avoid, fancy, forbid, hate, involve, justify, mind, postpone, preclude, prevent, recall , resent , risk and understand. Random samples of 1,000 tokens containing each of these verbs were selected and the relevant tokens were extracted from these samples (in the case of matrix verbs represented by fewer than 1,000 tokens in the BNC, all tokens were considered). The accusative form of the complement clause subject is the more common of the two in the case of all verbs except postpone, of which were found only two tokens, one of each sort. Of a total of 433 tokens of the constructions in question, 372 are accusative and 61 possessive. All 433 tokens were analysed with respect to the form of the subjects of both the matrix verb and the complement clause predicates. The subjects were divided into two categories. The first category, labelled “Pr”, comprises pronouns of all sorts, as well as the understood subjects of imperatives. The second, labelled “FN/C” comprises full nominals and, in the case of matrix verb subjects, clauses. Table 3 and Figure 8 contain details of the matrix verb and complement clause subjects of all tokens.

We can see in table 3 that 54 possessive complement clause subjects (89%) are pronominal. The corresponding percentage for accusative complement clause subjects is 35%. This difference between the two types of subject is statistically significant at the level of p=0.001. 42 of 61 tokens of‘poss. -i ng contain pronominal matrix verb subjects, compared to 197 of 372 tokens of acc. - ing. The difference between the two is statistically significant at the level of p=0.05. [11] The fact that the possessive -ing construction favours a pronominal complement clause subject has frequently been remarked upon in the literature. For instance, Quirk et al. (1985: 1063) state: “In general, the genitive is preferred if the item is a pronoun, the noun phrase has personal reference, and the style is formal”. Similarly, The Cambridge Grammar of the English Language states:

…the genitive in a gerund-participial is awkward with NPs of any significant length or complexity, especially those with post-head dependents. It is most likely with personal pronouns, and after that singular NPs that refer to people and have no more than one or two words as pre-head dependents (Huddleston & Pullum 2002: 1193)

Neither of these major reference works note the tendency for the construction to occur with pronominal matrix verb subjects. Indeed, to the best of my knowledge, this tendency has not been remarked upon in any previous study of these constructions. Nevertheless, the tendency does exist. Moreover, it resembles the tendency found in the case of the promise constructions: in both cases the more/most unusual of two or more near-synonymous constructions is the more/most likely to occur with a pronominal subject.

The data adduced in this article reinforces the conclusion reached by Rohdenburg (1999) that the explicit mention of the promissee in a promise construction leads to a marked shift in favour of finite complement forms. In addition to having demonstrated how rare the X promise Y to construction actually is, I have also shown that it appears to strongly favour pronominal subjects, as does another comparatively rare construction, the poss. -i ng construction.

The question was posed in section 2 as to whether utterances like those cited in (9) – (14), such as John promised Mary to leave, are representative of how the X promise Y to construction is used in English. We have seen that they are not. This is not to say that one can totally exclude the possibility of their ever occurring in real speech. They may well do so, provided that John has already been firmly established in the current discourse context. The six authors whose examples were cited as (9) – (14) are interested in what they term “control” in sentences of the type X promise Y to. In particular, they are interested in capturing in their models the fact that X rather than Y instantiates the subject of the predicate in the complement clause. We have seen in section 3 that the form of the complement of promise is influenced by the form of the matrix verb subject. The facts about the X promise Y to construction that have been pointed out in this article may well be quite irrelevant to the theoretical discussions of control relations in articles like Chomsky (1980) and Bresnan (1982). It is, nevertheless, possible to argue that these analyses do neglect a significant fact about the construction. This is the fact that the subject X must itself be sanctioned in the universe of discourse, either by encoding the speaker or hearer, or by anaphora. As was pointed out above, the reason for this constraint may well be related to the very rare occurrence of the construction in actual usage. Being so unusual, it may be rather difficult to process. It is therefore more likely to be employed in contexts where the subject is easy to process. In contexts where the subject is more difficult to process, where it is encoded by a full nominal, we have seen that the maximally explicit that complement clause is preferred.

In the case of both sets of near-synonymous constructions dealt with in this article it is the minority construction that occurs most frequently with pronominal participants. The tentative contention is that this tendency may well be related to the rarity of the construction types in question. To what extent this contention can be generalised to include other rare construction types is an open question requiring further research. Near-synonymous pairs of constructions which are asymmetric in their distribution are, of their very nature, thin on the ground in synchronic corpora. Future research could, however, take the form of laboratory experiments, or excavations of diachronic corpora. It is especially in the latter that one might expect to find pairs of near-synonymous constructions. [12]

Bach, Emmon (1979). Control in Montague Grammar. Linguistic Inquiry 10: 515-531.

Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad & Edward Finegan (1999). Longman Grammar of Spoken and Written English. London: Longman.

BNC: British National Corpus, World edition on CD-Rom (2000). Oxford: Humanities Computing Unit of Oxford University.

Bresnan, Joan (1982). Control and complementation. Linguistic Inquiry 13: 343-434.

Chierchia, Gennaro (1989). Structured meanings, thematic roles and control. In Chierchia, Gennaro, Barbara H. Partee & Raymond Turner (eds.) Properties, Types and Meanings 11. Dordrecht: the Netherlands. 131-166.

Chomsky, Noam (1980). On binding. Linguistic Inquiry 11: 1-46.

Dixon, Robert M.W. (1991). A New Approach to English Grammar on Semantic Principles. Oxford: Oxford University Press.

Farkas, Donka (1988). On Obligatory Control. Linguistics and Philosophy 11: 27-58.

Francis, Hartwell S., Michelle L. Gregory & Laura A. Michaelis (1999). Are lexical subjects deviant? Proceedings of the Chicago Linguistic Society 35: 85-97.

Givón, Talmy (1995). Functionalism and Grammar. John Benjamins Publishing Company: Amsterdam.

Hawkins, John A. (2004). Efficiency and Complexity in Grammars . Oxford: Oxford University Press.

Hindle, Donald (1981). A probabilistic grammar of noun phrases in spoken and written English. In Sankoff, David & Henrietta Cedergren (eds.) Variation Omnibus. Carbondale/Edmonton: Linguistic Research. 369-78.

Huddleston, Rodney & Geoffrey K. Pullum (2002). The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press.

Mair, Christian (1995). Changing patterns of complementation, and concomitant grammaticalisation, of the verb help in present-day English. In Aarts, Bas & Charles F. Meyer (eds.) The Verb in Contemporary English: Theory and Description . Cambridge: Cambridge University Press. 258-272.

Mair, Christian (2002). Three changing patterns of verb complementation in late Modern English: a real-time study based on matching text corpora. English Language and Linguistics 6.1: 105-31.

Matthews, John & Bob Stewart (1988). Warriors of Christendom . Poole: Firebird Books Ltd.

Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik (1985). A Comprehensive Grammar of the English Language . London: Longman.

Rohdenburg, Günter (1995). On the replacement of finite complement clauses by infinitives in English. English Studies 76(4): 367-388

Rohdenburg, Günter (1999). Clausal Complementation and Cognitive Complexity. Proceedings of the conference of the German Association of University Professors 20: 101-112.

Sag, Ivan A. & Carl Pollard (1991). An integrated theory of complement control. Language 67(1): 63-113.

Tables i, ii and iii contain details of person and number for all pronominal promisers and promissees in the three X promise Y constructions. Full nominals are classified as either given or new in the tables, “givenness” being defined narrowly in the sense of being explicitly mentioned in the immediate discourse context.