I DON’T WANT CARMELO READING THIS TWICE: NONFINITE SYNTACTIC ALTERNATION GOVERNED BY WANT IN CONTEMPORARY ENGLISH

The verb want selects at least two types of complementation patterns when it is followed by a nominal constituent, as illustrated in want Carmelo to read this and want Carmelo reading this. In the light of data retrieved from several corpora of Present-day English, this paper explores the syntactic, dialectal, textual and semantic characteristics of both structures in the very recent history of the language. As regards the syntax of the constructions, an analysis based on the notion of extended transitivity is here suggested in an attempt to integrate the syntactic features of the patterns within a single syntactic schema.


INTRODUCTION
This paper focuses on syntactic variation between two structures of complementation of the verb want, illustrated in (1) and ( 2): (1) I don't want Carmelo to read this paper again.[want NP to-V] (2) I don't want Carmelo reading this paper again.[want NP V-ing]2 1.The research reported has been funded by the Spanish Ministry of Education and Science, grant number HUM2005-02351/FILO, which is hereby gratefully acknowledged.This investigation is couched in a larger project on the degree of variation experienced by the English language in its recent history as far as the syntactic complexity of clausal constituents is concerned.
In both patterns, want is followed by a nominal or pronominal constituent (Carmelo) and either a to-infinitive or an ing clause of which the nominal element is capable of functioning as subject, at least on semantic grounds.That there must hold an external argumentative syntactic dependency between the nonfinite clause and the noun phrase occurring after want implies that examples like (3) will not be included in the [want NP to-V] pattern since chance is the governing head of the infinitive clause to observe him with Magda within the noun phrase a chance to observe him with Magda, and not its subject.
(3) (...) because she wanted a chance to observe him with Magda (LOB K10) As regards the [want NP V-ing] construction, the condition on the type of syntactic dependency between the nominal element and the ing clause leads to the rejection of the following examples from the pattern.These examples illustrate, respectively, an ing clause which is not syntactically linked either to want or to the nominal constituent following want (example ( 4)), a nominal ing form (examples ( 5) and ( 6)), an ing form premodifying a noun (example (7)) and an ing form postmodifying the nominal constituent acting as the object of want (in (8)). 3  (4) and he straightway sent message to the king <,,> saying that he accepted the whole <,> principle of string <,> and indeed <,> wanted some <,> giving his reasons for so doing <,,> (DCPSE:DL-I02/LLC:S-11-03 #0121:2:A) (5) Call this person if you want this booking immediately <,,> (DCPSE:DI-B59/ICE-GB:S1A-074 #0339:6:A) (6) they want long-range fighting here (DCPSE:DL-F03/LLC:S-10-03 #0292:1:A) (7) she she wants the sleeping bag (DCPSE:DL-B25/LLC:S-02-13 #0664:1:A) (8) whenever they wanted the sound of a body falling to the ground with a thump <,> we used to drop <,> Marie (DCPSE:DL-I02/LLC:S-11-03 #0310:5:A) In this paper I will, on the one hand, review the literature on nonfinite complementation depending on verbs such as want in an attempt to find evidence for the existence of factors which account for variation between the [want NP V-ing] and the [want NP to-V] constructions, and, on the other hand, on the basis of personal corpus-driven research, I will suggest an analysis of patterns which takes into account both the structural factors and the findings offered by my data.To that end, I will refer to syntactic, structural, semantic, dialectal and text-type variables, both synchronic and diachronic, which may play a role in the distribution of the two patterns.As pointed out by Mair (2003: 342), any attempt to account for complement choice after verbs (...) on structural and semantic grounds alone will remain incomplete (...).What is needed (...) is a variationist account (...) integrating synchronic regional and stylistic variation, on the one hand, and ongoing diachronic change, on the other.[Mair's italics].
Section 2 deals with the syntax of the constructions and pays special attention to the syntactic status of the noun phrase occurring between want and the nonfinite constituent in the constructions under analysis.Section 3 investigates the semantic, dialectal, textual and historical characteristics of the two patterns and offers the results of personal corpus work.Finally, section 4 puts forward some concluding remarks.

THE SYNTAX OF THE [WANT NP NONFINITE-CLAUSE] CONSTRUCTION
The literature has paid extensive attention to the [want NP to-V] construction.By contrast, the [want NP V-ing] construction, which is statistically infrequent, has been devoted mere in-passing comments or very brief analyses.In this section I will focus, first, on the status of the intermediate noun phrases occurring in the [want NP to-V] patterns (section 2.1) and, second, on the syntax of [want NP Ving] (section 2.2).In section 2.3 I suggest a possible analysis of the construction [want NP nonfinite-clause].

THE SYNTAX OF THE INFINITIVE CONSTRUCTION
The twofold status of the intermediate noun phrase in the [want NP to-V] construction as either the subject of the infinitive clause or the object of want has been discussed extensively in the literature.Such a fuzzy nature of the noun phrase justifies the label of (1) as a 'raising' construction, that is, as an illustration either of the subject of a subordinate clause which becomes the object of the main clause (subject-to-object raising) or of the object of a main clause which ends up by fulfilling the function of subject of a nonfinite subordinate clause (object-tosubject raising).
The issue of the status of the nominal constituent as either subject or object has received explanation in Langacker's Cognitive Grammar but assuming that, for example, in (1), Carmelo is both the landmark of want and the trajector of read (Langacker 2000: 347), as shown in Figure 1, from Langacker (2000: 344, his  Langacker maintains that the conceptual basis of ( 9) and (10) below is unique, both structures being connected by his concept of 'transparency'.In his words, "any element which could occur as the subject of a complement clause can also function as the raising verb's object in the main clause" (Langacker 2000: 319).
(9) I expect Don to leave.
(10) I expect that Don will leave.
From a different perspective, Matthews (1981: 185-186) emphasises the twofold analysis of the nominal constituents in sentences such as those with want and maintains that these are examples of 'catenative fused constructions'.In his words, "a single element [author: the noun phrase] is a complement of both a controlling and a dependent predicator" (185); "'catenative' would (...) refer to the fusion of a dependent verbal construction with that of either the transitive or the intransitive" (186), as shown in (11): The adaptation of (11) to the infinitive construction with an intervening noun phrase could be (12) below, in which the dependent verbal construction would be you to do it and the main transitive one would be I want you: In section 2.1.1I will offer arguments in favour of the analysis of the (fullyfledged or pronominal) noun phrase as the object of matrix want.Section 2.1.2is devoted to the alternative option, namely the analysis of the noun phrase as the subject of the subordinate infinitive clause.

Generation of the intermediate nominal as a matrix object
The analysis of the intermediate nominal constituent as an object of want has been given support by the following facts: (i) The nominal constituent has oblique case: 4   (13) I don't want him to read this paper again.
(14) I don't want him/his/he reading this paper again.
As pointed out by Quirk et al. (1985Quirk et al. ( : 1186)), "a subject pronoun in the objective case can often be replaced, in formal style, by a possessive pronoun" in the ing pattern.Non-genitive case-marking of the ing construction is a signal of informal style (Huddleston et al. 2002(Huddleston et al. : 1190)).
(ii) That the intermediate nominal constituent can be a reflexive proform controlled by the main subject in examples similar to (1) would constitute an argument in favour of its analysis as the object of want (the same argument will be used to support the subject analysis below). 5  (15) Zelda believes herself to be virtous.[from Langacker 2000: 319] (iii) Given that no elements can be interpolated between a verb and its object unless saliency is intended, the fact that Carmelo in (1) has to be placed immediately after want would underline its status as an object: (16) *I want very strongly Carmelo to read this paper again.[versus, for example, I believe very strongly that Carmelo will never read this paper again; adapted from Postal (1974: 134)] Huddleston et al. (2002Huddleston et al. ( : 1180) ) mention that the distribution of specifically adjuncts in the constructions under analysis constitutes a further argument in favour of the status of the nominal constituent as the object of want.In their words, "[i]n general, adjuncts cannot occur between a verb and an NP object, but they are permitted between a verb and a clausal complement".Example (17) shows that adjuncts are not allowed between verbs and non-clausal objects, whereas ( 18) and ( 19) demonstrate that the adverbials all along and at once are acceptable before, respectively, an object that-clause and an object nonfinite clause.
Since (17) is ungrammatical, Huddleston et al. conclude that an improvement to be postponed is not a clausal object and thus an improvement alone is eligible as the object of want in (20): (20) *We want all along an improvement to be postponed.
That additional material can be inserted between Carmelo and the nonfinite verb, as in our corpus example in (21) or in the made-up one in (22), leads Gonzálvez García (1999) to the rejection of a strong dependency link between Carmelo and read in, for example, (1).In his words, "the insertion of [...]  In this respect, Huddleston et al. (2002Huddleston et al. ( : 1179Huddleston et al. ( -1180)), who mention the resistance of want about adopting passive morphology 6 (see example (25) below, the passive version of (24), versus (26), which is perfectly grammatical), disregard passivisation as a necessary proof for objecthood and maintain that Carmelo is the object of want in (1) above.In their words, "there is no other relevant difference between want and expect, and given that passivisation doesn't provide a necessary condition for objects, we shall not wish to assign different structures" (1179).
(26) The performance was expected to begin at six.Huddleston et al. (2002Huddleston et al. ( : 1180) ) thus claim that the ungrammaticality of ( 25) is due to a property of the verb want, not to the structure of the active clause (in (24)).b.I want a new computer to be placed in my office.
6. Our corpora contain few examples with passive want, in which the interpretation of want is associated either with job recruiting, as in (i) or (ii), or with police calls, as in (iii): (i) Speakers from England were not wanted any longer.(Brown G66).
(ii) When a specialist was wanted anywhere the application came to me.(LOB G23).
In fact, Bolinger (1974: 70-71;reported by Gonzálvez García 1999) claims that the construction in (27b) is possible if and only if (27a) is acceptable.In his words, "the conceptual verb is forced to be compatible with two complements at the same time.So a sentence like 'Do you admit the facts to be true?' is normal because 'Do you admit the facts?' and 'Do you admit the truth of the facts?' harmonize" (70-71).However, the application of such a hypothesis to, for example, (1) is not in keeping with the unacceptability of ( 28): (28) *?I want Carmelo. 7  (vi) Examples such as (29) below, regarded as "unlikely" by Matthews (1981: 182) 8  or "rare" by Mair (1990: 120), in which only the infinitive and its de-pendents are fronted to sentence-initial position by the operation of topicali-sation, could reinforce the non-monoclausal analysis of John and to help me: (29) To help me I want John.
7. The status of the nominal constituent as either an argument or a non-argument of the matrix verb has led Huddleston et al. (2002Huddleston et al. ( : 1201Huddleston et al. ( -1202) ) to distinguish two types of [V NP to-V] constructions, namely those with ordinary objects, such as (i), in which Liz is an argument of persuade, and those with raised objects, as in (ii), where Liz cannot be said to be an argument of intend.In Huddleston et al.'s words, in (ii) "[w]ith intend [...] we have three complements but only two arguments: Liz is a raised object" (1201).
(i) Pat persuaded Liz to interview both candidates.
(ii) Pat intended Liz to interview both candidates.
Providing that examples such as (28) in the main text are hardly acceptable, it goes without saying that the want construction fits in the pattern in (ii), that is, the one in which the nominal constituent following want is a raised object.The arguments in (iii) to (vi), offered by Huddleston et al., demonstrate  b.Liz {intended / wanted} to leave.8. Matthews (1981: 182) maintains that examples such as (29) are "unlikely" not because the infinitive plus its object (to help me) and the nominal constituent following want do not belong to the same clause but because the infinitive (to help) and its object (me) are "two constituents and not one" (184).Matthews' analysis of (29) would be:

Generation of the intermediate nominal as a subordinate object
The existence of arguments in favour of the (base-)generation of the nominal constituent as the subject of the nonfinite clause has led scholars such as Chomsky (1965: 22-23) to maintain that Carmelo in (1) or ( 2) is both the subject of read and the object of want and to sustain that a (subject-to-object) raising 'transformation' operates in these constructions.Among the arguments which give support to the subject status of the noun phrase (at least, in an initial stage of the derivation) are the following: (i) The nominal constituent can control a reflexive proform in the nonfinite clauses, which seems to indicate that the noun phrase and the nonfinite predicate constitute a monoclausal constituent: (30) Liz wants John i to shave himself i before the ball. 9(ii) Passivisation in sentences containing idiom chunks can prove that the noun phrase originates within the nonfinite clause.
(31) We all want tabs to be kept on all the radicals.
Tabs, the object of the idiom keep tabs on in the active, has to be regarded as part of the nonfinite clause and fulfils the function of subject of the passive version.Quirk et al. (1985Quirk et al. ( : 1193) ) observe that passivisation cannot be applied to, among others, the want examples, as shown in (32) below -I pointed out above that the morphological passivisation of want is practically impossible.As a consequence, Mair (1990: 98) concludes that this to happen in (32) is not divisible and, accordingly, this has to be analysed as the subject of to happen: (32) a.He can't want this to happen.[Mair's example (17a)].
b.*This can't be wanted to happen.[Mair's example (17b)].9. Langacker (2000: 319) provides the example in (15) in the main text, repeated below for convenience, as evidential, on the one hand, of the monoclausal status of the raising construction, with which I agree, and, on the other, of the status of the intermediate nominal constituent as subject of the nonfinite clause.To my knowledge, the example is not felicitous since it illustrates precisely that Zelda and herself are constituents of one 'umbrella' clause, which implies that herself has to be analysed as the object of believes.By contrast, my example (30) in the main text shows that the nominal constituent under research is the subject of the nonfinite clause.
However, Mair recognises that the active-passive relation is often skewed in some contexts and favours the use of semantic facts as criteria for the distinction between subject and object noun phrases controlled by or depending on, for example, want -as he points out, ( 33) is ungrammatical and this cannot lend support to the claim that to ask a question is not a constituent: (33) *To ask a question was wanted.In Mair's words, there are matrix verbs such as want "which either do not passivise at all or only do so exceptionally, in very narrowly defined contexts" (113).
(iii) Examples similar to the [want NP to-V] and the [want NP V-ing] constructions, such as ( 34) and ( 35) below, in which it can only be described as the dummy subject of the subordinate nonfinite clause, show that the subject status of the intermediate noun phrase is in the driving seat: (34) I expect it to rain this afternoon.[adapted from Langacker (2000: 324)].
(35) I heard it raining.[from Langacker (2000: 322)]. 10  In this vein, Mair (1990: 119) provides an example with want in which the intermediate constituent is expletive there.According to Mair, this example clearly demonstrates the analysis of the nominal constituent as the raised subject of the infinitive clause. 11(36) I don't want there to be any trouble.
(iv) From a semantic point of view, want is, in most of the examples, a monotransitive verb which thus subcategorises for an object, which would favour the analysis of the material following want as a monoclausal constituent fulfilling the function of object of want.In this respect, Mair (1990: 119) points out that want can (rarely) be complex-transitive, as in I want my coffee hot, which thus would put into question the uniqueness of the analysis of the elements following it -Mair recognises that he has found no examples of either ditransitive (similar 10.Andersson (1985) regards (35) as a participial construction, whereas the pattern with want is considered monotransitive ('B-verbs' in his terminology).
11.In the corpora investigated I have found very few examples with expletive elements occurring after want: (i) "I do not want there to be any feeling of jealousy between the two towns in this".(LOB A43) to, for example, [ask NP to-V]) or complex-transitive ([consider this to be x]) want in the corpus. 12  (v) The nominal constituent plus the infinitive clause can be replaced with a proform, which highlights their monoclausal status and, in consequence, the analysis of the former as the subject of the nonfinite clause: (37) a.I don't want her to miss the train.
b.I don't want it.
(vi) Huddleston et al. (2002Huddleston et al. ( : 1180, footnote 6), footnote 6), who analyse the nominal constituent following want as the object of the matrix predication, mention that "[i]n the catenative construction want takes for only when the catenative complement is preceded by an adjunct".The example which they give is (38) below: (38) They had wanted all along for the performance to begin at six.
As they see it, the occurrence of for before the nominal constituent warrants its subject status only in the examples with for.
(vii) The cleaving test also sheds light on the syntactic status of the nominal constituent following want, at least when the former is preceded by for.In Huddleston et al.'s (2002Huddleston et al.'s ( : 1180) ) words, "[w]hile a for-infinitival can occur as complement in pseudo-cleft (...) constructions [,] the sequence NP + infinitival VP cannot".Example (39) shows that you to be happy can be the focus (or subject-predicative complement) of a pseudo-cleft sentence only when it is governed by for: (39) a.I want you to be happy.b. *All I want is you to be happy.[Huddleston et al.'s (2002[Huddleston et al.'s ( : 1180) ) example (29iia)].
In view of this fact, Huddleston et al. conclude that the structure with for is different from the one without the complementiser.As already pointed out, when for occurs, the ensuing nominal constituent is analysed as the subject of the subordinate verb; by contrast, in the examples without for Huddleston et al. claim that the noun phrase following want is the matrix object.
(viii) Huddleston et al. (2002Huddleston et al. ( : 1189Huddleston et al. ( -1190) maintain that (case-) marking the noun phrase in the ing construction as genitive resembles the role of for in the examples with to-infinitives since both the genitive case and for trigger the analysis of the ing constituent as a clause and, in consequence, the subject status of the noun phrase.These authors claim that "the preference for the nongenitive in informal style can be seen as regularising the clausal construction" (1190); put differently, the selection of nominative or accusative nominal segments in informal language suggests that the noun phrases are closer to the unmarked subjects of infinitives, with accusative case, or of finite verbs, which are nominative.

THE SYNTAX OF THE ING CONSTRUCTION
Table 1 summarises the already-mentioned arguments in favour of the object/subject status of the noun phrase in the [want NP to-V].I have added a column which indicates whether the arguments holding for the infinitive construction also apply to [want NP V-ing] or not.
Table 1 shows that most of the arguments which have been adduced in the literature for the analysis of the noun phrase as a matrix subject or a subordinate object in the infinitive construction also apply to the ing one.Whereas the syntactic factors in [2], [3], [4], [5], [6], [7], [9] and [11] do not merit further speculation, several comments are however in order here with respect to the remaining ones.First, as far as factor [1] is concerned, as already pointed out in the introduction to this section, [want NP V-ing] admits not only nominative and accusative but also genitive pronouns.Genitive case-marking underlines the monoclausal analysis of the noun phrase and the ing verb, that is, its subject status.Second, the cleaving test produces better results with ing clauses, as shown in (40) (versus (39b), repeated here for convenience): (40) *?All I want is Zeda being happy.
(39) b. *All I want is you to be happy.That (40) is more easily acceptable than (39b) indicates that the monoclausal status of the nominal constituent Zeda and the ing clause being happy is in the driving seat.Finally, the insertion of for before the nominal constituent is not possible in the [want NP V-ing] construction, as shown in ( 41): (41) *I don't want for Carmelo reading this paper again.
I will discard the argument in [10] from the list of syntactic factors which have consequences for the analysis of [want NP to-V] and [want NP V-ing] since, as suggested by those who back up the analysis of the nominal constituent as the object, I contend that the examples with for belong to a different structure.

AN ANALYSIS OF THE [WANT NP NONFINITE-CLAUSE] CONSTRUCTION
In the previous sections I put forward arguments in favour of the syntactic status of the nominal phrase occurring after want in the constructions under analysis as either object of matrix want or subject of the subordinate infinitive or ing clause.Such syntactic factors, summarised in Table 1, could not decide on one analysis over the other since there were strong arguments in favour of the subject nature (the acceptability of expletives and the behaviour of idioms) and the object status (distribution of other constituents, e.g.adjuncts) of the noun phrase -other syntactic factors were not decisive at all.Pérez-Guerra (2003: 277), which deals with the analysis of secondary predicates such as resultatives and depictives, suggests the scale of transitivity in ( 42): (42) intransitive > monotransitive with no overt objects > monotransitive resultative with an overt fake or lexical 'object complement within VP' > complex-transitive with an object and an independent object predicative On the basis of such a hypothesis, the predicates of sentences including resultatives, such as ( 43), would occupy the third level of transitivity and would be analysed as in (43') below: Providing, first, that want is reticent about admitting complex-transitive subcategorisation (see section 2.1.2above) even though a few examples of this type were found in the corpora; second, that the noun phrase after want and before the nonfinite constituent has a twofold status since there are arguments in favour of monoclausality with either the matrix or the subordinate predicators; and, third, that the nonfinite clause seems to instance an extension of the monotransitive pattern, I shall suggest here that the nonfinite clause, be it an infinitive or an ing clause, occupies the third stage in he transitivity scale in (42) and must accordingly be analysed as an 'object complement within VP', thus implying 'extended monotransitivity'.In basic categorial terms, the analysis would be as follows: Even though the (rare) examples with expletives are left for further discussion, their analysis might fit the second stage in the transitivity scale in (42) above, that is, in the class of prototypical monotransitive subcategorisation.

OTHER SEMANTIC AND STRUCTURAL FACTORS
In this section I discuss other non-syntactic factors which are cited in the literature as significant in the explanation of variation between the infinitive and the ing forms in want-predicates.In the description of the behaviour of the constructions under investigation according to the variables just mentioned, I will refer to statistical findings from my data on the following corpora: -The Lancaster-Oslo/Bergen Corpus of British English, for use with digital computers (LOB), representing written British English in 1961.
-The Freiburg-LOB Corpus of British English (FLOB), for written British English in the early 1990s.
-The Diachronic Corpus of Present-day Spoken English (DCPSE), including ICE-GB (early 1990s) and the London-Lund Corpus (from late 1960s to early 1980s).
-The British National Corpus (BNC), for spoken and written British English from 1980s to 1993.
-A Standard Corpus of Present-Day Edited American English, for use with Digital Computers (Brown), a collection of American English published in 1961.
-The Freiburg-Brown Corpus of American English (Frown), for American English written in the early 1990s.
-TIME Magazine corpus, for American English from 1923 to the present.
This section is organised as follows.Section 3.1 is devoted to the semantics of the constructions.In section 3.2 I investigate the productivity of the patterns according to the force of the sentence.Section 3.3 tackles the connection between the spread of each of the alternative structures and the polarity of the sentences in which they occur.Finally, section 3.4 offers the results of dialectal, textual and historical variables.

SEMANTIC ACCOUNT
This section is devoted to the exploration of the semantic contrast between [want NP to-V] and [want NP V-ing] when both constructions are possible with the same (basic) interpretation.To give an example, since I shall focus on variation between constructions which are semantically, that is, truthconditionally, equivalent, out of the four interpretations which, according to Quirk et al. (1974: 880) are associated with (50), only the interpretation in (50a) will fall within the scope of the present investigation since this is the one that also corresponds to ( 51) -( 50b) is an example of a nonfinite purpose clause whose covert subject corefers with the girl; (50c) is another example of a purpose clause whose subject corefers with the subject of the matrix clause, that is, he; and (50d) is an example of a postmodifying infinitive clause, already discussed in Section 1.
(50) He wants a girl to finish the cleaning.a.'He wants that a girl will finish the cleaning'.b.'He wants a girl in order that she will finish the cleaning'.c. 'He wants a girl in order that he can finish the cleaning'.d.'He wants a girl who will finish the cleaning'.
(51) He wants a girl finishing the cleaning.
That there exists an aspectual or modal difference between the infinitive and the ing constructions has been pointed out in the literature from Palmer (1974: 195) onwards (see, for example, Quirk et al. 1985Quirk et al. : 1191Quirk et al. , 1195)), whereas scholars such as Andersson (1985) have objected to that claim and have claimed that the two patterns are semantically equivalent.Recent studies on this instance of variation have focused on the semantic differences between such constructions.
In a recent paper, Wherrity and Granath (2006) maintain that (1) and (2) above are two distinct surface structures which portray a semantic difference: whereas the basic meaning of the ing form in (2) is 'process' (with variants such as vividness, immediacy, on-goingness, non-futurity or experienced activity), 14 the to form is associated with futurity and conceptual remoteness (see Gonzálvez García 1999: Section 5.3 for the semantics of the to-infinitive raising 14.Wherrity and Granath (2006) claim that on-goingness and vividness are compatible since they maintain that the main discourse function of the ing construction is to make the narrative more vivid by presenting an event as taking place over time.In this respect, they offer example (i) as an illustration of such a function: (i) The help desk informed me that they had decided to "keep it quiet" because they did not want too many people travelling to the Reading Festival.(The Guardian, 1998).To my knowledge, [want NP V-ing] is used in example (i) simply as an alternative to [want NP to-V], with no significant semantic differences.In this vein, as agreed by my native informants, the meaning of (i) with the [want NP V-ing] construction does not imply that the organisation wants to avoid people's constant travelling to the Festival.construction; see Fanego 2004b: section 5 for an overview of different semantic explanations of to and ing construction couched in functional and cognitive frameworks as well as for constructive criticism of such proposals). 15Such a semantic difference is somehow in keeping with, for example, Quirk et al.'s (1985Quirk et al.'s ( : 1191Quirk et al.'s ( and, similarly, 1195) ) claim that "[a]s a rule, the infinitive gives a sense of mere 'potentiality' for action, (...) while the participle [author: the ing form] gives a sense of the actual 'performance' of the action itself" (see Mair 1990: 85-86 for the application of Quirk et al.'s claim to subject infinitive clauses). 16According to Biber et al. (1999: 757), the difference between the infinitive and the ing clauses lies on the more "hypothetical or potential" meaning of the latter construction with respect to the former.In their view, for verbs that can control both infinitive and ing clauses, such as want, these semantic factors are responsible for the choice.

FORCE
Whereas there does not seem to hold any connection between (illocutionary) force and the spread of the [want NP to-V] construction, in their ( 2006) paper, Wherrity and Granath maintain that the [want NP V-ing] pattern occurs commonly in imperative, proclamatory and exhortative sentences.Examples (44) to (46) (Wherrity and Granath's examples (8), (12) and ( 16)), on the one hand, illustrate respectively such uses and, on the other, evince the preference of the ing construction for negative polarity (see section 3.3 in this respect).
15.The telic interpretation of the infinite construction would agree with the directional origin of to, which by Late Old English or Early Middle English lost its adverb or prepositional meaning 'towards' and grammaticalised as an infinitive marker (see, in this respect, Fischer 1992: 317ff, Traugott 1992: 241ff or Fanego 2004b: 27).From a different perspective, Smith and Escobedo (2001: 561) claim that the insertion of to in the [V NP to-V] construction iconically separates the times of the matrix and the subordinate clause (versus the ing construction, in which no lexical material detaches both clauses).Wierzbicka (1988: 165) notes that the future orientation, related to the purposive interpretation of to adverbial clauses, "should be regarded as part of the semantic invariant of all TO complement constructions".
By contrast, the 'in-process' meaning conveyed by the ing construction would derive from the actional interpretation which the Old English suffix -an, from which -ing descends, added to verbs in order to form abstract nouns (see Fanego 2004b: 27).
16.The fuzziness of the aspectual interpretation associated with the ing construction can be exemplified by resorting to Huddleston et al. (2002Huddleston et al. ( : 1232) ) when they say that the "gerund-participial with want2 generally has a progressive interpretation but in non-affirmative contexts it can be nonprogressive".To give some examples, standing is claimed to have progressive meaning in (ia), on a parallel with (ib), whereas bringing in (ii) would be non-progressive: (i) a.I want them standing when the Minister enters.b.I want them to be standing when the Minister enters.(ii) I don't want you bringing the dog with you.
(44) "You haven't shaved again: you look a right mess.I don't want you coming to see me like that".(The Guardian, 1996).
(45) "I've only been here eight months and I am just trying to get out.I don't want my son growing up around here".(The Guardian, 2000).
(46) In Pensacola, he was welcomed at a rally of nearly 10,000 supporters by the blue grass musician Ricky Skaggs, who told the crowd: "We don't want Hollywood deciding who our next president is".(The Guardian, 2004).
Out of the 148 examples of the [want NP V-ing] pattern in my corpus data, 113 occur in declarative sentences and 6 are registered in interrogative sentences.Since the ascription of a given example to the semantic types recognised by Wherrity and Granath is not straightforward, I cannot corroborate Wherrity and Granath's findings for the [want NP V-ing] construction and cannot conclude that the illocutionary force of the sentence plays a role in the distribution of this pattern.

GRAMMATICAL POLARITY
The studies on the connection between polarity and the type of nonfinite clausal complementation governed by want of which I am aware have focused on the [want NP V-ing] pattern since the [want NP to-V] is, in this respect, unmarked and is found extensively in both affirmative and negative sentences.
In a corpus of two million words of British and American English, Andersson (1985) found one single example of [want NP V-ing] (Andersson's 'PrP' construction) in a sentence without negation.On the other hand, Wherrity and Granath (2006) report that the vast majority of the [want NP V-ing] examples occur in negative sentences. 17  Several remarks seem in order here in the light of the data in Table 2. First, want is followed by an infinitive clause in 134 examples per one million words, 18 whereas the normalised frequency (n.f.) for the examples in which want governs an ing form is only 8 examples per million words. 19This signifies that the to-infinitive clauses constitute the preferred complementation alternative selected by want, at least when there is an intervening noun phrase between want and the nonfinite clause.Second, the distribution of the data per text type might imply that more examples are found in the fictional texts (117 examples of the infinitive pattern per one million words occur in the fiction subcorpus and only 17 in the informative samples; almost all the examples of the ing constructions have been found in the imaginative texts).However, such a difference (contra Andersson 1985) has not proved to be statistically significant (p = 0.7024). 20Third, more examples of want followed by nonfinite complementation 18. Mair (1990: 111) finds 366 examples of the [V NP to-V] construction in the Survey of English Usage corpus, 80-90 percent of the sentences with want being included in this pattern, a percentage which is in line with, for example Andersson (1985), already discussed -other verbs which are found in the [V NP to-V] pattern in Mair's (1990) study are: allow, enable, expect and get (41-50 percent); like (31-40 percent); cause and require (11-20 percent); and have, order, wish, etc. (with frequencies lower than 10 percent).The specific proportions of [V NP to-V] in the spoken and written subcorpora are, respectively, 393.25 (175 occurrences in 445,000 words) and 483.54 (191 in 395,000 words) per one million words.Unfortunately, Mair does not provide the specific frequencies for the want examples.
19. Aarts and Aarts (1995)  have been retrieved from the British subcorpus (83 occurrences per one million words versus 59 in the collection of American English).Unfortunately, since Andersson does not provide the overall frequency for want in his corpus, one cannot investigate whether this remark is justified simply by the wider occurrence of want in British English or whether this indicates that clausal complementation after want is more popular in British English.In this vein, the preference for want controlling a to-infinitive form, as already mentioned, is also corroborated when one takes into account the dialectal distribution of the construction (97.57and 90.67 percent of the examples include the infinitive form in, respectively, British and American English).Even though the proportion of constructions including the ing form is apparently greater in American English (3.03 percent of the examples in which want controls a nonfinite clause belong to the ing type in the British texts, and the percentage for American English is 9.32), the chi-square results demonstrate that such a difference is not statistically significant (p = 0.1283).My data corroborate the same conclusion since the distribution of ing versus to-infinitive complementation in LOB-FLOB versus Brown-Frown is not significant either (LOB-FLOB: 9 and 168 instances of, respectively, [V NP V-ing] and [V NP to-V] versus Brown-Frown: 7 and 183 examples, respectively; p = 0.6886).
In their study on want complementation after 1993, Wherrity and Granath (2006: Table 1) registered 842 examples of the ing construction in their corpus of approximately 255 million words, that is, with a normalised frequency of 3.3 examples per million words.Apart from the raw percentage itself, which unfortunately is not contrasted with the frequency of the [want NP to-V] alternative, Wherrity and Granath's paper is certainly illuminating because it attests the apparent progressive increase of the ing construction from 1993 onwards. 21In Figure 2 I offer Wherrity and Granath's (2006) results on the basis of normalised frequencies per 1 million words -the figures used by Wherrity and Granath are not normalised and, in consequence, their findings may be slightly distorted.The chi-square test does not however support the visual output in Figure 2 and reveals that the diachronic difference is not statistically significant (p = 0.7176). 22 Figure 2. Wherrity and Granath's (2006) frequencies for the [want NP V-ing] construction after normalisation.

CONCLUDING REMARKS
This paper has tackled the analysis of the major nonfinite complementation patterns governed by want in Present-day English.Special attention has been paid to the [want NP V-ing] construction, whose ratio of occurrence is very low in English (approximately 6 percent of the examples in which want is followed by a nonfinite constituent in Andersson (1985), 4.95 percent in Aarts and Aarts (1995), 3.3 instances per million words in Wherrity and Granath (2006) and 148 examples in my corpus data from seven large corpora).As a consequence of the low proportion of [want NP V-ing], want is not even included in Biber et al.'s (1999) top-ten list of verbs controlling ing clauses.
The existence of strong arguments in favour of the status of the nominal constituent after want as object of the matrix clause and subject of the nonfinite predicate has led to the analysis of such a nominal element as the object of want, followed by an extended complement materialised by a nonfinite clause (or Tense Phrase) whose subject is a PRO category coreferring with the matrix subject.
Both the analysis of the semantics of the two patterns of complementation in those examples in which the constructions can be interchanged and the productivity of both structures according to variables such as force, text-type idiosyncrasy, dialectal distribution and diachronic evolution have shown that the differences between [want NP to-V] and [want NP V-ing] are not statistically significant in contemporary English.In consequence, Smith and Escobedo's (2001: 561) claim that "to vs. -ing complements are semantically motivated and not arbitrary" has not been corroborated as regards the nonfinite complementation of want.By contrast, I have supported the view that polarity plays a significant role in the selection of [want NP V-ing] since most of the ing complements occur in negative sentences.
(v) The existence of parallel patterns controlled by want with and without depending nonfinite clauses has been used as an argument in favour of the object behaviour of the intermediate noun phrase: (27) a.I want a new computer.
resemblance between want and intend: (iii) relation with passive infinitivals: a. Pat persuaded both candidates to be interviewed by Liz.[meaning different from (i)].b.Pat {intended / wanted} both candidates to be interviewed by Liz.[meaning identical to (ii)].(iv) selection restrictions: a. *Liz persuaded the spotlight to intimidate Pat.b.Liz {intended / wanted} the spotlight to intimidate Pat.(v) dummies: a. Pat {intended / wanted / *persuaded} there to be one student on the board.b.Pat {intended / wanted / *persuaded} it to be easy to obtain a pass grade.(vi) simple/complex choice: a. *Liz persuaded to leave.
In their corpus, which consists of approximately 255 million words from British newspapers dated 1993-2004 (The Times 1993, and The Guardian and The Observer 1996-2004), 83.13 percent of the ing examples (700 out of 842 examples) are registered in sentences with negative polarity; 5.46 percent (46 examples) are retrieved from interrogative sentences and 11.4 percent (96 examples) occur in affirmative clauses.Such percentages seem to reveal that the variable which plays the most significant role in the distribution of the [want NP V-ing] construction in the press genre from 1993 to 2004 is '± interrogative undertake a corpus-based analysis of the patterns [want NP to-V] and [want NP V-ing], which leads to the following frequencies: 96 examples of [want NP to-V] and 5 occurrences of the [want NP V-ing] construction in a corpus of post-1985 printed English (TOSCA) which comprises approximately one million words.20.The number of examples which I have attested in my corpus data is so small per text type that no statistical significance test can be applied to my figures.

Table 1 .
Syntactic status of the noun phrase.