2 The origins of syntactic structure

• More evidence for the existence of constituents
• A first attempt at building syntactic structure
• Projecting syntactic structure from lexical items
  • Elementary trees
  • The X' schema
  • Verbal subcategories
  • Deriving simple sentences
  • Deriving complex sentences
• The modification relation
  • The adjunction operation
  • Multiple adjuncts
  • A typology of syntactic dependents
  • More on the distinction between complements and adjuncts
• Notes

In Chapter 1, we presented various pieces of evidence for the structure dependence of human language, including the intuition that certain words belong together, the phenomenon of structural ambiguity, and the dependence of a noun phrase's interpretation on its structural relations to other noun phrases. These facts all raise a basic question---what is the origin of syntactic structure?

In this chapter, we begin to answer this question. After reviewing some further evidence for the existence of constituents, we explore, but ultimately reject, an idea about syntactic structure that readily comes to mind: namely, that it is the result of concatenating the words in a sentence into successively larger binary-branching constituents. We then introduce an idea that is linguistically more natural: namely, that the words that make up a sentence impose structural requirements that largely determine how they combine into phrases and sentences, much as the physical properties of atoms regulate how they combine into molecules. The modification relation, briefly mentioned in Chapter 1, introduces a bit more freedom into this combinatorial system and is therefore covered in a separate section.

More evidence for the existence of constituents

As was mentioned in Chapter 1, we have the intuition in sentences of our native language that certain words belong together---in technical terms, that several words can form a single constituent. In this section, we support this intuition with further evidence.

Consider the sentences in (1).

(1)	a.		Zelda sings arias.
	b.		His student sings arias.
	c.		That tall student sings arias.
	d.		The tall Finnish student sings arias.

In (1a), the subject of the sentence is a single word---Zelda---so it is obviously a constituent. In (1b-d), the sequences of italicized words consist of several words, yet they function as subjects of their sentences in the same way that the single word Zelda does in (1a). For instance, each of the italicized sequences is a possible response to the question in (2) (the curly brackets are a conventional way of indicating alternatives).

(2)	a.		Who sings arias?
	b.		{ Zelda, his student, that tall student, the tall Finnish student. }

A related fact is that each of the italicized sequences in (1) can function as the focus of the it-cleft construction illustrated in (3).

(3)

It is { Zelda, his student, that tall student, the tall Finnish student } that sings arias.

Finally, notice that where the referent of the italicized sequences has been introduced in previous discourse, the sequences themselves can be replaced by an appropriate pro-form, as in (4).

(4)	a.		What is { Zelda, his student, that tall student, the tall Finnish student } up to nowadays?
	b.		She sings arias.

The fact that the pro-form that substitutes for each of the italicized sequences in (4a) is a single word again provides compelling evidence that the sequences themselves are structural units.

It is worth pointing out, by the way, that the term 'pronoun' is potentially misleading because it suggests that pronouns substitute for nouns. In fact, what they substitute for is entire noun phrases, as is evident from the ungrammaticality of sentences like (5).

(5)	a.	*	His she sings arias.
	b.	*	That tall she sings arias.
	c.	*	The tall Finnish she sings arias.

A more accurate term for these lexical items would therefore be pro-noun phrases. The traditional term is so established in usage, however, that we continue to use it, at least as long as it is unlikely to cause confusion.

Constituents need not be subjects. In (6), for instance, the italicized words function as the object of the verb admit.

(6)	a.		They will admit Zelda.
	b.		They will admit his student.
	c.		They will admit that tall student.
	d.		They will admit the tall Finnish student.

Evidence for the constituenthood of the sequence in this position comes from question-answer pairs and sentences analogous to the ones just given, as in (7).

(7)	a.		Question	Who(m) will they admit?
	b.		Sentence fragment	{ Zelda, his student, that tall student, the tall Finnish student. }
	c.		It-cleft	It is { Zelda, his student, that tall student, the tall Finnish student } that they will admit.
	d.		Substitution of pro-form	They will admit her.

Constituents also need not be noun phrases. For instance, we conclude that the italicized sequences in (8) are constituents on the basis of the evidence in (9).

(8)	a.		Adverb phrase:	It happened only recently.
	b.		Prepositional phrase:	It happened { an hour ago, on Tuesday, over the weekend. }
(9)	a.		Question	When did it happen?
	b.		Sentence fragment	{ Only recently, an hour ago, on Tuesday, over the weekend. }
	c.		It-cleft	It is { only recently, an hour ago, on Tuesday, over the weekend } that it happened.
	d.		Substitution of pro-form	It happened then.

For constituents other than subjects, there is a final piece of evidence for constituenthood. In English, the initial constituent of a declarative sentence is ordinarily the subject, as in (10). This ordinary word order is referred to as canonical. But English also allows noncanonical word orders, as in (11), in which we can think of a constituent as having moved from its canonical position to a position preceding the subject. Noncanonical word orders sometimes require special discourse contexts to be felicitous; if so, we have added appropriate material in parentheses.

(10)			Canonical	They will call the top three candidates early tomorrow morning.
(11)	a.		Noncanonical	Early tomorrow morning, they will call the top three candidates.
	b.			The top three candidates, they will call early tomorrow morning. (The others, they will contact later on this week.)
	c.			(They said they would call the top three candidates early tomorrow morning, and) call the top three candidates early tomorrow morning, they will.

A first attempt at building syntactic structure

Concatenation from left to right, as illustrated so far, yields so-called left-branching structures. In such structures, it is the left branch of a branching node that contains complex substructure, and right branches are always nonbranching. It is also possible to apply concatenation from right to left. Doing so results in right-branching structures, in which it is right branches that contain complex substructure and left branches that are nonbranching. (14) gives the right-branching structures that result from concatenating the words in (1a-d) from right to left.

(14)

a.

b.

c.

d.

Given the possibility of building syntactic structure by concatenation in either direction, we might ask ourselves whether the direction of concatenation matters, and if so, which direction is the correct one for English.

In order to address these questions, it is crucial to have a clear understanding of the role of tree structures in syntactic theory. For a mathematician, the formal properties of tree structures are interesting in their own right, but for a syntactician, the interest of tree structures lies in the fact that they are representations, or models, of a particular domain of inquiry---namely, syntactic structure, or the relations among syntactic elements. In other words, the graphic structure of a tree on the page is intended as a model of the syntactic structure in the mind. In any good model, the properties and the structure of the model correspond in simple and straightforward ways to the structure of the domain of inquiry. Such close correspondences allow us to state observations and generalizations about the domain of inquiry without unnecessary complication. Eventually, these correspondences may even enable us to use our understanding of the model's formal properties to generate hypotheses and to discover generalizations concerning the domain of inquiry that otherwise would have gone unnoticed.

Given these considerations, a natural requirement to impose on tree structures is that in (15).

(15)

A sequence of elements is represented in tree structure as a constituent iff it is exhaustively dominated by some node.

Observe now that in the left-branching structures in (12b) and (13), the subjects are correctly represented as constituents, as indicated by the boxed nodes. In other words, these left-branching structures are consistent with the requirement in (15), in contrast to the right-branching structures in (14). We might therefore be tempted to conclude that concatenation in English proceeds from left to right.

However, such a conclusion would be premature in light of sentences like (6), repeated here as (16).

(16)

They will admit { Zelda, his student, that tall student, the tall Finnish student. }

If you use the grammar tool in concatenation to build the relevant left-branching and right-branching trees, you can see for yourself that in this case it is the right-branching trees that correctly represent the object as a constituent.

Finally, to complete the paradox, a simple sentence like (17) shows that English is neither a consistently left-branching nor a consistently right-branching language.

(17)

The tenor will sing the first aria.

Both the tenor and the first aria are constituents, as demonstrated by the evidence in (18) and (19).

(18)	a.		Question	Who will sing the first aria?
	b.		Sentence fragment	The tenor.
	c.		It cleft	It is the tenor that will sing the first aria.
	d.		Substitution of pro-form	He will sing the first aria.
(19)	a.		Question	What will the tenor sing?
	b.		Sentence fragment	The first aria.
	c.		It cleft	It is the first aria that the tenor will sing.
	d.		Substitution of pro-form	The tenor will sing it.
	e.		Movement	The first aria, the tenor will sing. (The second one is for a bass.)

The consistently left-branching structure in (20a) correctly represents the subject as a constituent, and the consistently right-branching structure in (20b) correctly represents the object as one, but neither representation manages to represent both facts at once.

(20)

a.

b.

The impasse that we have arrived at suggests that the concatenation approach to syntactic structure that we have been exploring is not headed in the right direction. The reason for this is actually not far to seek: the concatenation approach is really just a gussied-up version of the beads-on-a-string approach to sentence structure already mentioned in Chapter 1. Both the beads-on-a-string approach and the concatenation approach, for instance, are incapable of representing the structural ambiguity of sentences like (21) in an insightful way.

(21)

I enthusiastically recommend this candidate with no qualifications.

Specifically, neither approach is able to associate each of the two interpretations with a structure of its own.

Projecting syntactic structure from lexical items

In this section, we introduce an approach to syntactic structure according to which individual words are themselves represented as small pieces of syntactic structure, or elementary trees. The idea is to be able to generate the phrases and sentences of a language by composing these elementary trees with one another in well-defined ways.

In this view, the words of a language are comparable to the atoms of physical matter. Atoms do not combine into molecules just because they happen to be next to each other; rather, each atom has particular properties that govern its combinatorial possibilities and that have to do with its internal structure (for instance, the number of electrons on the atom's outermost shell and the relative number of protons and electrons).

In developing the approach to syntactic structure just sketched, we begin by focusing on verbs and then generalize the approach to the point where we can build simple sentences as well as complex sentences containing subordinate clauses. For expository reasons, we treat noun phrases, prepositional phrases, adjective phrases, and adverb phrases as unanalyzed constituents in this chapter, leaving a detailed discussion of their internal structure until Chapter 3.

Elementary trees

What information needs to be represented in an elementary tree---say, for a lexical item like constructed? A first important piece of information is the fact that constructed belongs to the syntactic category of verbs (see Verbs for some basic information and terminology). For instance, it can combine with two noun phrases to form a sentence, whereas lexical items belonging to other syntactic categories cannot.

(22)	a.		Verb		They constructed a new dormitory.
	b.		Adjective	*	They constructive a new dormitory.
	c.		Noun	*	They construction a new dormitory.

Second, the elementary tree for constructed needs to represent the fact that it combines with two noun phrases---no more and no less, as illustrated in (23).

(23)	a.	*	They constructed.
	b.	ok	They constructed a new dormitory.
	c.	*	They constructed the students a new dormitory.

Both of these pieces of information are represented in the two elementary trees in (24). V', the label for the intermediate node in (24a), is read as V-bar. Notice, by the way, that what corresponds to the traditional category 'verb phrase' in (24a) is the V' node, not the VP node. From a semantic point of view, it is useful to think of VPs as denoting situations.

(24)

a.

b.

Does it matter which elementary tree we choose to represent constructed? Given (25), the answer is yes.

(25)	a.		First Drexel constructed a new dormitory,
	b.		and then Penn did so.

The grammaticality of substituting the pro-form did so in (25b) shows that the sequence constructed a new dormitory is a constituent. The elementary tree in (24a) represents this fact, whereas the one in (24b) doesn't. The elementary tree in (24b) is therefore inadequate on factual grounds, or empirically inadequate.

The X' schema

As we show later on in this chapter and in Chapter 3, the form of the elementary tree in (24a) can be extended to other syntactic categories. In other words, (24a) is an instantiation of the general schema in (26), which is known as the X' schema (read: X-bar schema) of phrase structure; X, Y, and Z are variables over syntactic categories.

(26)

We say that a lexical item---for instance, constructed in (24a)---projects the syntactic structure in the elementary tree. The lexical item is the head of the projected structure. The lexical item's syntactic category is called the lexical projection of the head. V' is the head's intermediate projection, and VP is the head's maximal projection (sometimes also called phrasal projection). The head and its three projections form what we will call the spine of the elementary tree. The three projections have the bar levels indicated in (27).

(27)			Projection	Bar level	Example
			Lexical	0	V
			Intermediate	1	V'
			Maximal	2	VP

The sister of the intermediate projection---ZP in (26)---is called the specifier. Each elementary tree has at most one specifier; as we will see later, elementary trees can lack a specifier altogether. In English, the specifier of VP---abbreviated as Spec(VP)---is the intermediate projection's left sister, but in a VOS language like Malagasy (recall Exercise 1.3), it is the right sister. Following traditional terminology, the sister of the head---YP in (26)---is called its complement (note the spelling with e rather than i---the idea is that complements complete the meaning of the head). As we will see in the next subsection, some lexical items lack a complement. The issue of whether a head can have more than one sister has been the subject of much debate in the syntactic literature; we postpone a discussion of this issue until a later chapter. The specifier and complement positions of a head are its argument positions. In other words, an elementary tree consists of a spine and argument positions.

The terms 'specifier' and 'complement' are used to refer to structural positions, as just described, as well as to the expressions that substitute into these positions in the course of building syntactic structures for whole sentences. If it is necessary to avoid confusion, we will speak of 'specifier/complement position' versus 'specifier/complement expression.'

Verbal subcategories

			Examples	Verbal subcategory
(28)	a.		Borges died.	Intransitive
	b.		Mona Lisa smiled.
	c.		They waited.
(29)	a.		They constructed / a new dormitory .	Transitive
	b.		They demolished / the house .
	c.		The lions devoured / a Thompson's gazelle.
(30)	a.		I gave / my friends / a ride.	Ditransitive
	b.		Let's lend / the neighbors / our lawnmower.

In connection with the different verbal subcategories, it is important not to confuse the syntactic notion of transitivity with the related, but not identical, semantic notion of valency, or adicity. For instance, from a semantic point of view, a verb like eat always expresses a dyadic (= two-place) predicate---specifically, a relation between two participants in an eating event, one doing the eating (the agent) and one being eaten (the so-called patient). When eat is used transitively (We ate dessert), both of these participants are represented in syntactic structure, but when it is used intransitively (We ate), only the agent is represented in syntactic structure, whereas the patient is not. In other words, in the intransitive use of eat, there is a mismatch between the two-place character of the eating relation and its intransitive representation in syntactic structure. When interpreting a sentence with such a mismatch, a hearer or reader resolves the mismatch by supplying a patient with an appropriate generic meaning (in this case, 'food'). It is worth repeating, however, that this understood patient is not available in the syntax.

An additional potentially confusing factor is that the same term, argument, is used in both a semantic and a syntactic sense. A semantic argument is a central participant in some relation denoted by a predicate, typically (though not always) expressed by a verb. A syntactic argument is the expression of such an argument in a sentence. As we have just seen in the case of eat, semantic arguments are not always represented as syntactic arguments.

Because of the potential for mismatches of the sort just discussed, it is actually quite rare for verbs to belong to just one syntactic subcategory. Two dyadic verbs that, like construct, are obligatorily transitive (recall (23)), are demolish and devour, as shown in (31) and (32).

(31)	a.	*	The workers demolished.
	b.	ok	The workers demolished / the house.
	c.	*	The workers demolished / the company / the house.
(32)	a.	*	The lions devoured.
	b.	ok	The lions devoured / a Thompson's gazelle.
	c.	*	The lions devoured / themselves / a Thompson's gazelle.

The more usual situation is illustrated in the following examples. For instance, (33) shows dyadic verbs being used both transitively and intransitively.

		Transitive	Intransitive
(33)	a.	We ate / dessert.	We ate.
	b.	Please don't interrupt / the meeting.	Please don't interrupt.
	c.	I like to knit / sweaters.	I like to knit.
	d.	He is reading / a book.	He is reading.

Conversely, in addition to an intransitive use, certain monadic (= one-place) verbs can be used transitively in the cognate object construction (so called because the verb and the object are cognate, or etymologically related).

		Intransitive	Transitive
(34)	a.	Borges died.	Borges died / an easy death.
	b.	Mona Lisa was smiling.	Mona Lisa was smiling / her mysterious smile.

It is also possible to use some triadic (= three-place) verbs not just ditransitively, but transitively and even intransitively (the parenthesized adverb in (35a) is not a complement, but rather a modifier). Notice that, in principle, either of the two complements can be omitted.

		Ditransitive	Transitive	Intransitive
(35)	a.	I gave / my nephew / a Tinker Toy set.	They gave / everything they had.	They gave (generously).
	b.	I sent / my friends / the pictures.	I sent / the pictures.	---
	c.	We teach / college students / syntax.	We teach / college students. We teach / syntax.	We teach.
	d.	He told / me / the whole story.	He told / me. He told / the whole story.	Don't tell.

Finally, certain basically dyadic verbs can be used ditransitively.

		Transitive	Ditransitive
(36)	a.	I baked / a delicious cake.	I baked / my friends / a delicious cake.
	b.	The lions killed / a gazelle.	The lions killed / themselves / a gazelle.
	c.	She sang / a lullaby.	She sang / her baby / a lullaby.

In the remainder of this section, we discuss the elementary trees for intransitive verbs (as mentioned earlier, we postpone a discussion of how to represent ditransitive verbs until a later chapter). Consider an optionally intransitive verb like eat. The elementary tree for transitive eat is analogous to the tree for constructed and is given in (37).

(37)

For intransitive eat, the two trees in (38) come to mind.

(38)

a.

b.

The choice between the two trees is not self-evident, and (38b) might at first seem preferable because it is simpler (in the sense of containing fewer nodes). Nevertheless, we will prefer (38a) on the grounds that its adoption results in a simplification of the grammar as a whole---that is, a simplification not just of the elementary trees themselves, but also of rules and definitions stated over them. First, with transitive verbs, the pro-form do so can substitute for the verb and its complement, but not for the verb alone, as illustrated in (39). (Why do so substitution should be restricted in this way will not concern us here; we will simply rely on the restriction as a matter of fact.)

(39)	a.		First the children	ate	dinner,
			and then the adults	did so.
	b.	*	First we	ate	pasta,
			and then we	did so	dessert.

With intransitive verbs, do so substitution gives the results in (40).

(40)			First the children	ate,
			and then the adults	did so.

Given the elementary tree in (38a), the restriction concerning do so substitution can be stated succinctly as in (41).

(41)

Do so substitutes for instances of V'.

Given (38b), the corresponding generalization, which is stated in (42), is both more difficult to formulate and less straightforward.

(42)

Do so substitutes for instances of V' or of V without a complement.

A second, related reason to prefer (38a) is that it permits the succinct definition of the notion of specifier in (43a) rather than the longer one in (43b).

(43)	a.		Specifiers are sisters of intermediate projections.
	b.		Specifiers are sisters of intermediate projections or of lexical projections without a complement.

In concluding this section, we make explicit a point that is implicit in the preceding discussion. Obligatorily transitive and obligatorily intransitive verbs project a single elementary tree of the right type. Verbs that can be used either transitively or intransitively project two elementary trees. More generally, we allow lexical items of any category category to project one or more elementary trees, as needed.

Deriving simple sentences

In a sentence like (44), the verb waited contains the bound morpheme -ed, which expresses past tense.

(44)

He waited.

If tense morphemes were invariably expressed on the verb in this way, then complete structures for full sentences could be derived by substituting appropriate structures into the argument positions of the verb's elementary tree. However, this is not a general solution, because tense is not always expressed as a bound morpheme on the verb. For instance, future tense in English is expressed by the modal will, as in (45), the future tense counterpart of (44).

(45)

He will wait.

Even more strikingly, present and past tense in English is alternately expressed as an appropriate form of auxiliary do in do support contexts, as shown in (46), and as a bound morpheme on the verb otherwise.

(46)	a.		Negation:	He { doesn't, didn't } wait.
	b.		Question:	{ Does, Did } he wait?
	c.		Emphasis:	He { does, did } wait.

These facts raise two related questions. First, how can we represent sentences like (45) and (46), in which tense is expressed as a free morpheme? (We postpone discussion of questions and negated sentences until later chapters.) Second, how can all sentences be represented in a syntactically uniform way, regardless of the morphological status of tense as a free or bound morpheme? The reason that we want a syntactically uniform representation is that given a semantics of tense that is oversimplified but sufficient for our purposes, both past and future are semantically parallel: they are functions that take situations (denoted by VPs) as their arguments and return situations that are located in time, either before or after the time of speaking.

We begin by answering the first question in several steps. First, in line with the general approach to syntactic structure that we have been developing, modals and auxiliaries, like all lexical items, project elementary trees. For instance, the elementary trees for will and auxiliary did are shown in (47).

(47)

a.

b.

Second, we build a core structure, or predicate-argument structure, for a sentence by taking the elementary tree for its verb form in (48a) and filling the empty argument position, yielding (48b).

(48)

a.

b.

Next, we substitute the structure in (48b) into each of the elementary trees in (47), yielding (49).

(49)

a.

b.

Finally, we derive the correct word order by moving the subject from Spec(VP) to Spec(IP). This yields the structures in (50).

(50)

a.

b.

A few remarks are in order about the process of movement just introduced. Movement is best understood as a way of representing the fact that a single phrase is pronounced once in a particular position, yet can satisfy more than one syntactic function in a sentence. In the case at hand, he satisfies two distinct functions. First, it is a semantic argument of the verb wait---specifically, verb's agent argument. Second, it is the grammatical subject of the entire sentence. That these functions are distinct is shown by the existence of passive sentences, where a nonagent argument functions as the grammatical subject (in italics), and where the agent is expressed as a nonsubject, as in (51). (The various types of semantic arguments (agent, beneficiary, patient, and so on) are treated in more detail in a later chapter.)

(51)	a.		The book	was sent	to Mary	(by John).
			Patient		Beneficiary	Agent
	b.		Mary	was sent	the book	(by John).
			Beneficiary		Patient	Agent

In order to clearly express the phrase's double function, we do not simply move the phrase from one position to another. Instead, movement leaves a trace in the phrase's original position, and the two positions share an index. In the syntactic literature, indices for movement are represented by the same alphabetical subscripts as binding indices. For clarity, we diverge from this practice and use the natural numbers as binding indices, and the lowercase letters i, j, k, and so on, as movement indices. A constituent and its traces of movement (possibly none, in the absence of movement) are referred to as a chain. The elements of a chain are its links. Higher links in a chain are often referred to as the antecedents of lower ones. The motivation for this terminology is that traces of movement, just like reflexive pronouns, must be c-commanded by their antecedents.

We are now in a position to answer the second question posed earlier---namely, how can sentences be represented in a syntactically uniform way regardless of the morphological expression of tense? A simple answer to this question is possible if we assume that English has tense elements that are structurally analogous to auxiliary do, but not pronounced, as shown in (52); the square brackets are a convention to indicate silent elements.

(52)

a.

b.

Elementary trees as in (52) make it possible to derive structures for sentences in which tense is expressed as a bound morpheme on the verb along the same lines as for sentences containing a modal or auxiliary do. In (53), we illustrate the derivation of He waited.

(53)

a.

b.

c.

d.

Select elementary tree for verb

Substitute argument

Substitute (53b) in elementary tree for tense

Move subject from Spec(VP) to Spec(IP)

Deriving complex sentences

(54)	a.		We will ask if she left.
	b.		They believe that he came.

Although sentences with complement clauses can become quite long and complex (recall the instances of recursion in Chapter 1), deriving structures for them proceeds along the lines already laid out in the previous subsection. If and that are both complementizers, so called because they have the effect of turning potentially independent sentences into the complements of a matrix verb, and they project the elementary trees in (55).

(55)

a.

b.

Given elementary trees like (55), we can derive complement clauses like (54a) as in (56). This in turn allows the derivation of the entire matrix clause in (57).

(56)	a.				b.
			Elementary tree for complement clause verb				Substitute argument
	c.				d.				e.
			Substitute (56b) in elementary tree for tense (52b)				Move subject in complement clause				Substitute (56d) in elementary tree for complementizer (55a)

(57)	a.				b.
			Elementary tree for matrix clause verb				Substitute arguments, including clausal complement (56e)
	c.				d.
			Substitute (57b) in elementary tree for modal (47a)				Move subject in matrix clause

Given the representation in (57d), we can now characterize recursive structures as in (58).

(58)	a.		A structure is recursive iff it contains at least one recursive node.
	b.		A node is recursive iff it dominates a node distinct from it with the same label.

The modification relation

By contrast, the prepositional phrase after lunch, the adverbial phrase very rarely, and the noun phrase every day seem less closely related to the verb. In particular, they refer not to an additional type of event participant, but instead modify the event by specifying the time at which or the frequency with which it takes place. Such phrases are called modifiers or adjuncts.

The adjunction operation

As we saw in the last section, arguments substitute into syntactic positions that are part of elementary trees. Modifiers, being related to a head by a distinct and intuitively less close relation, are not integrated into syntactic structure by this operation of substitution, but rather by a second tree operation called adjunction. We illustrate adjunction in what follows in connection with (59a). For simplicity, we focus on its predicate-argument structure, omitting the projection of the silent present tense element.

As we will see in later chapters, adjunction is used for other purposes than to integrate modifiers, but it is always a two-step process that targets a particular node. When the purpose of adjunction is to integrate the modifier of a head, as it is here, the target of adjunction is the head's intermediate projection, as indicated by the box in (60a). The first step in carrying out adjunction is to make a clone of the target of adjunction that immediately dominates the original node, as in (60b). The second step is to attach the tree for the modifier as a daughter of the newly created clone, as in (60c).

(60)	a.				b.				c.
			Select target of adjunction				Clone target of adjunction				Attach modifier as daughter of clone

Deriving the rest of the structure for the entire sentence proceeds as outlined earlier, as shown in (61). It is not necessary, incidentally, for adjunction to precede the rest of the derivation.

(61)	a.				b.				c.
	Substitute arguments				Substitute (61a) in elementary tree for tense				Move subject

The recursive intermediate projection that is introduced by adjunction is empirically motivated by the facts of do so substitution. As (62) shows, do so can substitute for the verb, its complement, and the modifier, indicating that all three form a constituent.

(62)

We drink green tea { after lunch, very rarely, every day, } and they do so, too.

In addition, do so can substitute for the verb and the complement without the modifier, as shown in (63).

(63)	a.		We drink green tea after lunch, and they do so after dinner.
	b.		We drink green tea very rarely, but they do so quite frequently.
	c.		We drink green tea every day, and they do so every weekend.

This is evidence that the verb and the complement form a constituent on their own. A tree structure with the topology of (61c) represents the facts in both (62) and (63) in accordance with the principle in (15), repeated here.

(64)

A sequence of elements is represented in tree structure as a constituent iff it is exhaustively dominated by some node.

Multiple adjuncts

(66)	a.				b.				c.
	Select lower V' as target of adjunction				Clone lower V'				Attach modifier

A typology of syntactic dependents

(67)			Relation to head	Sister of ...	Daughter of ...
			Complement	Head	Intermediate projection
			Adjunct	Intermediate projection	Intermediate projection
			Specifier	Intermediate projection	Maximal projection

More on the distinction between complements and adjuncts

(72)			Every time I see him, ...
	a.		... he's eating.
	b.		... he's eating a six-inch steak.
(73)	a.		He ate, and I did so, too.
	b.		He ate a hamburger and french fries, and I did so, too.
	c.	*	He ate a hamburger and french fries, and I did so six samosas.

Although (69b) is false, (69a) does have the consequence in (74).

(74)

If a syntactic dependent is not a complement, it is not obligatory.

The two valid generalizations in (69a) and (74) can be summarized succinctly as in (75).

(75)	a.		Obligatory syntactic dependents are complements.
	b.		Adjuncts are optional.

read as V-bar

Why is "V'" read as V-bar when it contains not a bar, but a prime symbol? The reason is that in the 1970s, when the idea of bar levels was introduced, the various projection levels were distinguished by horizontal bars over a syntactic category. Heads had no bars, intermediate projections had one, and maximal projections had two. But back in the days of typewriters, such overbars were cumbersome to type (you typed the symbol, --* rolled up the platen a bit, backspaced, typed an overbar *--, repeated from --* to *-- for each overbar, and then rolled the platen down again the right amount). Overbars are also expensive to typeset, and they still aren't part of the standard character sets for HTML documents such as this one. Therefore, it was and continues to be economical and convenient to substitute prime symbols for overbars. However, linguists have failed to update their terminology, and so the old term 'bar' is still with us. In this connection, note also that 'XP' is generally used for X'' (read, X-double-bar).

yielding (49)

The representations in (49) look like they might be the structures for the questions Will he wait? and Did he wait? As we will see later on, however, the trees for questions contain an additional layer of structure.