Direct object of the verb "want"

Question

John wants Jim to write a letter to the mayor

According to Cambridge, the verb "want" can take the pattern of obj+to-infinitive as a complement, so that means "Jim" here is the direct object of the verb.

However, Ron Cowan in The Teacher's Grammar of English says that "Jim" here is not the direct object of the verb but is the subject in the to-clause. This is because the answer to the question of "What does John want?" is "For Jim to write a letter to the mayor" not merely "Jim". Another test that supports this conclusion is that a passivization can only occur in the complement clause but not in the main clause.

1.John wants Jim to write a letter to the mayor^OK

2.John wants [a letter to be written to the mayor (by Jim).]^OK

3.*[Jim is wanted (by John)] to write a letter to the mayor.^{not OK}

So my question, what is really the direct object of the verb "want"?

Answer 1

John wants Jim to write a letter to the mayor.

This is a catenative construction in which "Jim" is direct object of "want" (and the 'understood' subject of the subordinate "write" clause". The catenative complement of "want" is "to write a letter to the mayor", not "Jim to write a letter to the mayor", for the latter is not a constituent, but a sequence of direct object + complement.

The fact that we can’t passivise is a lexical property of "want": there are a fair number of exceptions to passivisation (cf. "John would like them to help him", but not *"Them to help him would be liked by John"!)

"Jim" is called a raised object: the verb that "Jim" relates to syntactically is higher in the constituent structure than the one it relates to semantically.

Answer 2

If you think about it, it's clear what John actually wants is

Jim [write] a letter to the mayor

After "John wants...", we need to express it as a noun phrase, which means that [write] must be expressed either as a gerund (-ing) or a to-infinitive. It is actually possible to use either with want: we use the to-infinitive form for a single action, and a gerund form for a continuous (progressive) situation. The to-infinitive form is much more widely used.

John wants Jim to write to the mayor
John wants Jane running around afer him

It is also possible to use an implied to be (a silent copula?) to indicate a continuous (progressive) situation:

John wants Jane [to be] at his beck and call

That's looking at things from a semantic viewpoint. When you look at it from the grammar point of view, it seems that Jim, not the write-phrase, is the object. This makes no difference with a name, but if we substitute a pronoun, we have to use an object-pronoun him rather than a subject-pronoun he:

John wants him to write to the mayor

For this reason, when the Cambridge Dictionary complilers got to work on it, they itemised the most common noun clause forms using obj to represent the word in this position, because they have to describe it in a way that people can choose the correct pronoun form.

[ + obj + to infinitive ] Do you want me to take you to the airport?
[ + obj + past participle ] This package - do you want it sent today?
[ + obj + adj ] Do you want this pie hot?
[ + obj + -ing verb ] I don't want you

So, semantically Jim is the subject of the noun-phrase, which in turn is the object of the want-clause, but grammatically the word Jim is the object of the want-clause. In the same way that physicists have wave-particle duality, linguists have this.

Answer 3

It seems that there is currently consensus that in "John wants Jim to write a letter to the mayor" "Jim" is in some way a direct object (not just a subject), but there is not consensus about the exact way in which it is a direct object.

As Bill J mentioned, the phenomenon is called "raising to object" or " subject-to-object raising"; there is a short Glottopedia article about it: http://www.glottopedia.org/index.php/Raising_to_object

Here is something I found online that seems useful: Lingering Challenges to the Raising to Object and Object Control, Jeffrey T. Runner, Syntax 9:2 (2006)

Runner says

There are currently three main approaches to the so-called (subject) raising to object construction illustrated in (1):

1. Cindy believes Marcia to be a genius.

The COVERT RAISING account assumes that in the surface string ‘Marcia’ is in the embedded clause; at LF [Logical Form] Marcia (or the relevant features associated with Marcia) raises to a main clause position where it is syntactically licensed (by checking case features) (e.g., Lasnik & Saito 1991; Chomsky 1995). This account assumes that the main verb ‘believes’ appears in VP (or vP). The OVERT RAISING account differs from the covert raising account in assuming that the movement to syntactically license the embedded subject is part of the overt syntax (Lasnik & Saito 1991; Koizumi 1993; Runner 1995); thus, this account additionally assumes that the main verb ‘believes’ moves to a position higher than the landing site of ‘Marcia’. This position is usually assumed to be VP- (or vP-) external. The third approach is the LEXICALIST account of the construction—initially proposed by Bresnan (1982)—I will focus here on the Head-driven Phrase Structure version (see, e.g., Pollard & Sag 1994; Sag, Wasow & Bender 2005). This account shares with the overt raising account the assumption that ‘Marcia’ appears in the main clause in the surface string; it differs from both raising accounts, though, by assuming a monostratal syntax, which means that though ‘Marcia’ is the object of ‘believes’ in the phrasal syntax, it is also associated with the syntactic and semantic features of the embedded predicate (‘to be a genius’) by a kind of coindexing called structure-sharing, and not movement.

What all of these accounts have in common, which make them different from the previous “exceptional case-marking” (Chomsky 1981) account—which claims that the embedded subject ‘Marcia’ is never part of the main clause—is that at some level of representation the embedded subject ‘Marcia’ is a kind of direct object in the main clause.(1-2)

I also found a chapter online that seems relevant, "Raising, Control, and Empty Categories," which seems to be from Andrew Carnie's Syntax: A Generative Introduction.