3. The Clause

While the main body of this text has focused on Functional Features and their involvements within the Phrase--constituting much of our discussion thus far on a narrow scope of morpho-syntactic features at the Phrase level, we now turn our attention to broad scope syntax and look to see if traditional notions of Clause and Sentence levels might not be captured in an equally stimulating way. Although there may not be much in the way of any functional features at the Clause level, since a clause is ultimately involved with stringing phrases together, and, as we have seen, phrases have already exhausted the inherent roles played by features, there may, however, be other new and interesting ways to think about Clauses. The following section attempts to define and scrutinize the infrastructure of the clause by introducing two new and very important concepts that have come out of Generative Grammar: (i) the Empty Category and (ii) Movement.

The Clause is traditionally defined as an expression which minimally contains a Subject and Predicate (viz., DP & VP). Therefore, any environment which strings a DP along side a VP has the potential of forming a clause:

(167) DP + VP => Clause

The distinguishing factor between a clause vs. a basic sentence (since both are defined as sharing the same material, subject + predicate) is that (i) a basic sentence doesn't enter into an arrangement with a dependent clause (forming a complex sentence), and (ii) an independent clause does. Hence, if we are talking at a complex sentence level (with two or more clauses), then an otherwise basic sentence becomes an independent clause. And if we are talking only at a basic sentence level (with only one clause of which to speak), then we maintain the label of simple sentence. Consider the following italicized clauses below:

What you should note about the two token (complex) sentences is that there are two clauses per each sentence: the Head (or initial) Clause and the Complement (or final) Clause. One nice way to spot such complexity is to see whether or not there are projected functional features having to do with Tense and Agreement regarding two or more verbs in the overall sentence. If so, then you have a complex sentence with two [+Fin] clauses on your hands. In short, if there are two verbs with T/Agr features, then you have potentially two (basic) sentences (or one complex sentence) made up of two clauses. In sentence (168a), there are two verbs with functional Tense/Agreement features spelled out (thinks and is)--and hence, there are two clauses. Likewise with sentence (168b), there are two [+Fin] verbs sporting T & Agr: (studies and remains). So, spotting and counting feature projections within a sentence can actually guide the overall analysis at the clause and sentence levels. However, some clauses may contain no functional features whatsoever on the Verb (and at times may even be Verbless)--such clauses are referred to as Small Clauses (=SC). In fact, Small Clauses appear to contain no typical functional predicate/verb material associated with Inflection--that is, a SC may contain no Auxiliary or Infinitive "to" particle.

Consider the Small Clauses below and remember that even though they constitute (as a class) a sort of elliptical or shortened clause (as can be seen via our substitution test below) they are not fully fledged clauses: they maintain no full predicate value in light of the absence of a [+Fin] verb.

Small Clauses
Consider the following S(mall) C(lauses) below:

(169)

(a) I believe [The president incapable of telling the truth].
(b) I want [John off the team].
(c) Will she find [me a real bore]?

(170) SC Substitution Test for Full Clause status: (showing [+Fin] vs. [-Fin] Clause):

In the above examples, we see that a Small Clause that shows no verbs can readily be lifted to a full clause status by filling in the required predicate material--in this case, by projecting the relevant verb.

(171) Empty Category
Having briefly looked at the deficient structure of Small Clauses, there however appear to be times when there may be more to a Clause than first meets the eye. Again, by calling on our substitution test, we quickly find that what first might appear as an Infinitive Phrase (cf., italicized (170b, c) above, might actually have the internal structure of a clause (if not at the surface phonological level, at least on a more abstract covert level--say, at a level in which features reside.

(172) Consider the clauses below:

What is interesting here is that while each of the subjectless infinitive strings (172b,d) may appear to be of a phrasal classification, their substitution counterparts show a potential subject slot within the constituency--promoting its status from a single constituent phrase to a multi-constituent clause. (Recall, that two phrases create a clause: in this sense below, an empty subject DP (PRO) merges with the VP to form a clause). Hence, much in the same way that zero allomorphs were viewed as projecting features despite the fact that no element in that slot appeared on the surface, we too view (172b,d) as full clauses which happen to take a zero allomorph (of sorts) in the embedded subject position. In other words, we can argue that subjectless infinitive clauses are headed by a Null Subject and that even though these slots may be empty at the phonological surface level (=empty category), they are (psycho-linguistically) understood as subjects with the same referential and functional feature properties pertaining to all typical Pronouns. The null subjects may either have (i) binding properties that relate the PRO back to its controlling subject of the initial clause, or (ii) free properties that keep the PRO independent of the matrix clause. Control Binding properties are illustrated in (172b,d) while free pronoun properties are illustrated in (172a,c). Binding properties render the overall sentence as an Si-derivation since (i) the covert PRO and (ii) overt Subject of the matrix clause are simple extensions of one another. Conversely, free pronouns, sharing no indexing, are independent from one another--hence rendering the overall sentence as an Sii-derivation.

Using relevant Generative Grammar terminology, the zero allomorph {ø} as the subject of an Infinitive clause is referred to as a PRO-drop subject since the PRO-noun of the clause is missing. Note that PRO has an antecedent that crosses clausal boundaries--in (173b) and (173d). Also note that the covert PRO empty category of the Infinitive [-Fin] clause has its subject features tethered to the overt PROnoun of the Finite [+Fin] clause. Hence, these empty PRO subjects are understood as having a co-indexing of features with the overt subject, or in more technical terms, as having an antecedent relationship with the subject of the matrix clause. This otherwise implicit covert subject becomes quite explicit when examining the informal predicate logic of the following (S)entences below (with example (174c) showing antecedent binding):

(174) I/She want [x], where [x] is [something]

Moreover, note in (175) below how the otherwise implicit subject becomes clearly visible when the relevant Infinitive [-Fin] clause is paraphrased by its Finite [+Fin] clause counterpart. Typically, the process behind an overt subject projection forces the infinitive {to} to delete while promoting the matrix verb from having a [-Fin] status to having a [+Fin] status encompassing the full range of T/Agr features.

(175) PRO subjects and Paraphrasing an Infinitive Clause to a Finite Clause:

(a) He is sorry [ø to have spoken too softly].
(a') He is sorry [he has spoken too softly].
(b) I am hoping [ø to have a stipend this year].
(b') I am hoping [I have a stipend this year].
(c) She has been claiming [ø to be divorced for years].
(c') She has been claiming [she is divorced for years].

The fact that the bracketed {to} infinitive clause seems to map nicely onto its italic clause counterpart adds further support to the notion that a PRO empty subject is always at least syntactically present in all Infinitive Clauses. The fact that it may not be sounded simply speaks to concerns at the phonological level and needn't concern us at the morpho-syntactic level.

One additional argument that we can factor into all of this concerns the syntax of reflexive pronouns. Recall that co-indexing has become a hot topic for us regarding feature relationships and exchanges between two pronouns. Consider how such reflexive co-indexing might necessitate a PRO subject:

(176) PRO subject Infinitives with Reflexives

(a) Mary likes [PROi to test herselfi]
	(Mary controls PRO) & (PRO is the antecedent of herself)

(b) John needs [PROi to watch himselfi]
	(John controls PRO) & (PRO is the antecedent of himself)

In the above examples, PRO is controlled by the overt subjects (Mary/John) while (herself/himself) must have their antecedents within their own bracketed clause (for reasons having to with a general principle that says binding requires a local adjacency condition). Without a local PRO in the bracketed Infinitive clause, the binding operation would not be sufficiently close to carry and project the relevant co-referential features.

There are other examples of [-Fin] V(erb) clauses worth discussing. These clauses tend to go without the Infinitive {to} particle thus making it hard sometimes to evaluate the status of the clause. Recall in our discussions above regarding [-Fin] verbs, that there were two other forms of Infinitives in addition to the {to} form: (i) the 'Bare V ' form, and the (ii) 'ing V' form. Consider the two additional [-Fin] forms making up their respective clauses below:

First of all, let's recognize that only the bracketed constituents in (177a,c) instantiate a clause by the mere fact that they project at DP and a VP (as stipulated in (167) above). Note that examples (177b,d) fail to provide us with the stipulated DP & VP rule. Therefore, examples (177b,d) are simple interrogative sentences which either encode a Double DP indirect and direct object (respectively) as in (177b), or a PP projection as in (177d). In either case, examples (177b,d) don't constitute a double clause projection. Sentences (177a,c) however do constitute a double clause projection since both required elements (DP and VP) are present:

It's not difficult to see that for such clauses all relevant functional features having to do with Case and T/Agr are set to a [-] specification. Case is set to a default (automatic setting) [-Nom] Accusative Case (him) while the T/Agr features on the verb go missing (write/walking).

Consider the paraphrase clauses below showing promotion of features--elevating the reduced [-Fin] clause to a unreduced [+Fin] clause now showing its full range of feature specification [+Nom, +T/Agr]:

(181) Summary of [-Fin] Clauses

* Note that the "for" in Example (181c) has an entirely different syntactic function than the commonly understood preposition "for". Above, for functions as a Complementizer in that it serves to introduce the subsequent clause: the complement of for is the clause. Consider the different functions between Prepositional and Complementizer for below noting that one nice way to tell the distinction is by invoking a movement operation (see movement in §4 below) by forming a wh-question among the two lexical items. The Prep(ositional) Phrase (PP) headed by preposition for can undergo movement (fronting) and be positioned at the beginning of the sentence whereas the Comp(lement) for may not. More specifically, an interrogative expression like who/what/which one? can be pre-posed in the front of a sentence (with or without for) only if for functions as a preposition. Consider the distributional fall out of the two functions of for under such movement:

(182) Preposition "for";

(183) Compliment "for";

Similarly, only a Comp for can be substituted by a that-clause (though note the feature change regarding the [+/-Finite] feature of the selected matrix clause):

(184)

Features on Comp
Since Complementizers are classed as functional categories, their internal make-up must also include some amount of feature specificity. For instance, the Complementizers (that) encodes a [+Fin] (Finite) Feature by virtue of the fact that they exclusively select a Finite Clause containing both a [+Nom] case on the (pro)noun as well as [+T/Agr] features on the verb. By contrast, the Comp (for) encodes the specific feature of [-Fin] (non-finite) since it selects an Infinitive clause. Consider the feature distributions of the clause types below:

In order for the Complementizer That to serve its role as a [+Fin] Comp, it is required to select both a (i) [+Nom] subject along with (ii) a [+Tense] Finite Verb. In contrast, consider the for-clause feature selection below:

In addition to the specific feature of Tense, the Agreement feature also enters into a [+Fin] Clause Projection--recalling, the [+Fin] requires both +T/+Agr as well as [+Nom] case. Returning our attention to the empty category (PRO) presented above, we quickly see that Agreement must be preserved between the overt subject and the empty PRO of the matrix clause--the Agreement features have to do with Number & Person. Consider the Agr relation in following clauses below:

(188) Summary of [+Fin] Clause

(189) Movement
As we have seen above, movement operations also can help define the Clause level of sentence structure. On a basic level, conditions placed on constituency require the whole clause to stay together in the event of some form of movement. Constituency here is defined as a (self-contained) structural unit or expression out of which a sentence is built up: e.g., Phrase-level constituency (as in PP, VP, DP) as well as Clause-level constituency (as in Dependent vs. Independent Clause). Consider how the following clause constituents maintain their cohesive structure even though they appear to have been scrambled within a sentence:

(190) Clause Movement

(a) I decided to wait till Friday to buy my supplies [since the book store was so crowded].
(b) Since the book store was so crowded, I decided to wait till Friday to buy my supplies

Notice below that one cannot move just a part of the clause and break the constituency:

(191) Clause-level Constituency Violations

(a) I decided to wait till Friday to buy my supplies since the book store was (so crowded).
(b) *So crowded, I decided to wait till Friday to buy my supplies since the book store was.
(c) I decided to wait till Friday to buy my supplies since the book store (was so crowded).
(d) *Was so crowded, I decided to wait till Friday to buy my supplies since the book store.

(e) He must have worked late into the night for him to be so tired.
(f) For him to be so tired, he must have worked late into the night
(g) *The night For him to be so tired, he must have worked late into.
(h) * Late into the night for him to be so tired he must have worked.

Note that this same constituency condition is placed on the Phrase-level as well:

(192) Phrase-level Constituency Violations

(a) She does like to see [DP which films]?
(b) Which films does She like to see?
(c) *Films does She like to see which?

(d) There are many good research journals [PP on the second floor of the library].
(e) On the second floor of the library, there are many good research journals.
(f) *Of the library, there are many good research journals on the second floor.

(g) The soldiers stood at attention [VP hoping to get praise from their sergeant]
(h) Hoping to get praise from their sergeant, the soldiers stood at attention.
(i) *Stood at attention Hoping to get praise from their sergeant the soldiers.

The actual VP here is elliptical and could be more correctly referred to as a dependent progressive clause--e.g.,

(j) The soldiers stood at attention (while they were) hoping for praise.

Coming on the heels of example (192j) above, what the various substitution tests show is that very often what might seem to be less than a clause is actually a clause once you consider the possibility that an ellipses has occurred. Consider such elliptical clauses which appear at first glance to be of a fragmented phrase (without a subject):

(193) (a) I saw the accident while driving home.

3.1 Independent & Dependent Clause

In this section, let's review those defining aspects that separate an Indep(endent) Clause from a Dep(endent) Clause. First of all, the grammatical term Independent means precisely that "Independent": its full meaning is not bound to some outside intra/inter-sentential source. In simple terms, an Independent clause is a clause that can stand on its on (as a possible sentence). A Dependent Clause (containing a subject and main verb) on the other hand must however link-up with some additional outside clause information in order to maintain its full and potential meaning. An Independent Clause in this sense then is tantamount to being a reduced simple sentence (albeit a simple sentence which is complicated by the fact that it serves in proximity to another (dependent) clause). Perhaps the best way to understand the workings of this tag-team dual structure is simply by looking at various examples of the two clauses at work. Consider the following Indep vs. Dep Clauses below:

The (Independent) clauses--He walked to the meeting, We all met at the pub for a beer, I saw an accident--can stand alone as potential basic sentences; however, their matrix (Dependent) clauses--because the bus drivers went on strike, after we spoke about the syntax project, while I was driving home--cannot. There is no complete or fully cohesive subject/predicate meaning behind the following fragmented sentences and when spoken or written down (punctuated as a sentence), they are referred to as Sentence Fragment errors:

Given a closer look, one notices that what in fact makes the fragments above Dependent is the insertion of a sole Sub(ordinate) Conj(unction), and the feature conditions with which they put on the complement clause. In other words, it is possible to drop the Sub Conj here and attain a completely perfect Indep(endent). clause/basic sentence: e.g., The bus drivers went on strike. We spoke about the syntax project. I was driving home. In a funny sort of way, it is by "adding" something onto an otherwise Independent clause that we render the clause Dep(endent). Typically, one imagines the converse--that sentence fragments are formed due to something incomplete or missing--here, quite the opposite is the case. The insertion of the Sub Conj renders an otherwise good sentence incomplete.

In addition to serving as a conjunctive particle, unifying two separate but related clauses, the conjunction also carries with it certain functional features. Below, we examine the roles of such functional features associated with both Sub Conjunctions as well Relatives.

(198) Feature Selection of Complement Clause

Starting with the fragment in (197a) above, the Subordinate Conjunction "because" (maintaining the semantics/logic proposition of cause and effect) could be said to carry a formal feature within its lexical entry (as part of its sub-categorization) which requires it to introduce only a dependent clause as its complement. In other words, owing to this lexical feature specification, by definition, there is no way that the lexical class of words which make-up subordinate conjunctions could ever be a Head of an Independent clause. Consider the full range of feature specification for the list of Subordinate Conjunctions below showing both the fact that they must introduce a Dep clause (notated as +Dep Comp(lement) as well as their semantic/adverbial scope [time, cause/effect, manner etc.):

(199) Summary of Subordinate Conjunctions

What is important to understand here is that each of the Sub Conj listed above (i) shares a common feature that requires the conjunction to introduce a Dependent Clause, as well as (ii) holds a specific lexical (idiosyncratic) feature which helps to project a specific meaning. The former common feature (serving to introduce a Dep Clause) is notated above as [+Dep clause Comp], while the latter individual feature is notated as e.g., [+Time], [+Manner], etc. Take the [+Time] feature for instance, clearly the Conj. after introduces a concept of "time" into the proposition--e.g.,

(201) Difference between Subordinate Conjunction & Preposition

There is often a considerable amount of confusion in deciding whether or not a word such as before or after is a Conj(unction) or a Prep(osition). For starters, recall that as part of the common feature specification (as outlined in (199) above), there is the stipulation that Sub. Conjunctions must be [+Dep Clause Comp]--that is, a Dependent Clause must ensue. So, to this extent, feature specificity may in fact help us determine an appropriate (fine-grained) grammatical status of an otherwise homophonic lexical item. Consider the paired Sentences (=S) below showing both Dependent Clauses (=D.C.) and Independent Clause (=I.C.):

(203) Relative Pronouns and their features: (who, whom, whose, which, that)

Feature specificity can likewise help in determining precisely how a relative Pronoun works in conjunction with in referential DP. Since both Subordinate Conjunctions and Relatives work to connect an I.C. to a D.C./relative clause, a certain amount of feature specificity applies. First of all, let's consider the range of lexico-semantic features for the following Relative Pronouns:

(204) Table: Relative Pronouns and Features

That is now accepted in standard English (along with who/m) as a modifier of a [+Person] DP. In this sense, the relative that may be denoted as holding two distinct feature values [+/-Pers].

Clauses introduced by such Relative Pronouns are called relative clauses. These pronouns have the flavor of acting like an adjective to the extent that they modify their referential DP. Consider the relative clauses below along with their specific feature projections:

(a) [The woman [SVO who kissed John] is an IBM executive.
	Who: Features: [+Per] [+ Subj]
(a') [Subj Who = (The woman)] [V kissed] [Obj John]
=> SVO embedded clause,

The relative who in (205a) above--taking on its subjecthood status via the referential DP The woman--becomes a derived active subject embedded in an SVO sentence with appropriate [+Person] and [+Subject] features. In one sense, what we have here are two SVO structures: (i) the larger SVO giving us The woman is an IBM executive, and (ii) the smaller embedded svo giving us Who/(She) kissed John. However, consider how the relative functional/grammatical feature changes in the sentence below:

(a) [The woman] [OSV whom John kissed] is an IBM executive.
	Who: Features: [+Per] [+ Subj]
(a') [Subj John] [V kissed] [Obj whom (=the woman/her)]

In (206) above, The woman now becomes a passive derived object when examining its role within the embedded clause--e.g., John kissed her (the woman).

(NB. To a certain degree, syntactically speaking, the DP positioning of The woman is commonly referred to as Topicalization (i.e., fronted), and is therefore removed from being a straightforward [+Nom] DP-subject: viz., the co-indexing of whom in the embedded clause plays this out).

Consider the various examples of "Correct Feature Spell-outs" and "Feature Crashes" for the following Relative Clauses below:

(207) Correct Relative Feature spell-outs: No Feature Crash

(a) I saw the boy who has been hanging around break the window with a base-ball.
	Who: Features: [+Per] [+ Subj]
(a') The boy/ He has been hanging around...
He => Subject, correct relative feature spell-out of [+Nom] case who.
	=> No feature Crash

(b) The player whose cleat fell off kept playing.
	Whose: Features: [+Per] [+ Poss]
(b') His cleat fell off. Both His & whose are correctly spelled-out with a corresponding [+Poss] feature: His forming a [+Poss] DP and whose forming a [+Poss] Relative.
	=> No feature Crash

(c) The dog which bit the post-man has been put to sleep.
	Which: Features: [-Per] [+ Subj]
(c') The dog is an inanimate subject thus takes the relative which.
	=> No feature Crash

(208) Examples of Feature Crash

(a) I saw the boy *whom has been hanging around...
	Whom: Features: [+Per] [+ Obj]
(a') The boy/ *Him has been hanging around...Him is incorrectly placed in the subject slot: improper relative feature spell-out of [-Nom] case whom.
	=> Feature Crash

(b) The player *who cleat fell off kept playing.
	Who: Features: [+Per] [- Poss]
(b') His cleat fell off. The relative must be correctly spelled-out with a corresponding [+Poss] feature: *Who cleat is that? (= Whose cleat is that?)
	=> Feature Crash

(c) The boy *which has been hanging around...
	Which: Features: [-Per] [+Inanim]
The boy is [+Person] so should be marked with either a Who or That relative.
	(i) The boy that...
	(ii) The boy who...

In addition to the relatives that take on the role of DP modifier, there is a class of relatives that serve in an Adjectival capacity--such relatives are called Relative Adverbs.

(209) Table: Relative Adverbs and their Features

Note that such relatives don't take on modifier status in the sense that they can neither substitute nor correspond with a referential DP:

(a) The boy [who/that/ *where/*when/*why...]

(b) The dog [that/which/ *where/*when/*why...]

In other words, adverbial relatives don't necessitate an antecedent in an adjacent clause but rather can function within their own clause. We can see this when we rephrase a relative clause as a Independent clause: while such adverbial are indeed relatives, since they have an antecedent DP, they seem to be unlike adjectives to the extent they can be rephrased as an Independent clause whereas Adjectival Relatives cannot.

(211) Adverbial Relative as Independent Clause

(a) I took the test [ when there were very few people around].
	=> functions with antecedent in a matrix clause

(b) Can you explain (the reason) [why he left]?

(b') He left for some reason. => functions within its own clause

In addition, Adverbial Relatives seem to contain a sufficient amount of proposition material so that they may very well stand on their own without any antecedent referential DP. Unlike Adjective Relatives that seem to require a DP antecedent in which to modify, Adverbial Relatives can seemingly carry on a grammatical function without a DP referential--suggesting that the adverbial relative in such elliptical structures may take on some of the functional features of the omitted DP:

(a) I need to stay at [a place] where I can write.
		Features: [+Place]
(a') I need to stay ø where I can write.
	(i) I can write here/there. (=at a place)

(b) I recall [the time] when John took a strange class about the history of furniture.
		Features: [+Time]
(b') I recall ø when John took a strange class about the history of furniture.
	(i) John took a strange class then. (= at a time)

(c) I never knew [the reason] why he took that class.
		Features: [+Reason]
(c') I never knew ø why he took that class.
	(i) He took the class for some reason. (=for a reason)

This apparent cross between a relative and a DP brings us to the Interrogative--where "wh-words" can (i) either merge in conjunction with an (obligatory) overt Noun as its complement forming a D+N (DP) (Prenominal), or (ii) function in isolation with a zero allomorph as its complement forming a D+ø (DP) (Pronominal). Recall that this cross between a relative (with an antecedent) and a subject/object DP was evident in our "IBM executive" example found in (205) above where who served both as (i) a relative with an antecedent in the adjacent clause (=The woman) and (ii) a subject-DP on its own right (playing the role of a subject in an embedded clause) (= Who kissed John).

(214) Table: Interrogatives (Features/Status)

What is interesting about the feature distinctions above is that "Wh-word" with a [-Pron] status cannot stand alone and project a DP, say without a Noun complement (to its left). In other words, we could say that such [-Pron] Wh-words contain an abstract complement feature that requires them to project with an overt Noun complement [D+N]--as opposed to [+Pron] Wh-words which contain no such complement feature (requirement), thus allowing the Wh-word to stand alone within a DP. (NB. "Wh-word" here refers to the fact that all Question words in English (save how, poor chap, which has undergone a sort of 'w-h' inversion) begins with the letter/sound "Wh"). Consider below how the [+/-DP] Feature Status affects the grammaticality of the following sentences. In addition to this Pronominal Feature, apparently there are some wh-words which cannot form DP expressions at all, but rather express Adverbial/predicate information (about a subject): such wh-words could be said to hold a [-DP] feature status. Consider the range of wh-word distributions below showing feature projections along with apparent Feature Crashes. (Feature Crashes are nothing more than instances of a grammatical "mismatch" between inherent features of particular words.)

(215) Wh-word status: Features [+DP] [+Pron]

(217) Wh-word status: Features [-DP] [-Pron]

(Since Wh-words Where/Which/How take on a strict [+Adverbial/-DP] status, we needn't fret over their DP status--the status of [-DP] can be given in the paradigm as a way to reflect their distributional qualities over their [+DP] counterparts).

In summary, we can classify the two aforementioned types of wh-words as either projecting a (i) [+/-Pronominal] feature, or (ii) a [+/-DP] feature. This way of classifying similar word types falls out naturally from a Feature-Checking Theory of language.

(218) Prenominal "wh-words" forming Constituents

Having now looked at the Prenominal [-Pron] DPs above, one other interesting grammatical phenomenon comes to light. When we consider a prenominal DP expression such as Which film in the sentence below, we instantly see that the wh-word (which) must remain adjacent to its complement Noun (films). In other words, whenever a prenominal wh-word is fronted (via movement) to the front of the sentence, the Noun complement must follow suit. This requirement of (D)-(N) Head-Complement Adjacency is viewed as being initially necessitated by the need to check-off the formal grammatical features pertaining to the [-Pron] DP. The fact that the phrase must remain whole, say even after the checking-off of features, has to do with Chain Integrity. Consider the Wh-phrase (DP) fronting below:

(a) Mary wishes to see which film?
(b) Which film does Mary wish to see?
	(a') Mary wishes to see [DP [D which] [N film]]?
	(b') [Which film] does Mary wish to see [DP [D which] [N film]]?

(c) * Which does Mary wish to see film?          Feature Crash (DP break-up)
(c') *[Which] does Mary wish to see [DP [D t] [N film]]?

What is happening above is that the Determiner Phrase (DP) must maintain its integrity as a unified phrase--that means holding the components that make up the phrase together. The idea behind this kind of self-adhesive stringing of words in a phrase is referred to as forming a "Chain" (being that certain words are inextricably linked to other words in forming a constituent, and that these whole constituents somehow are critical in conveying the essential meaning of the phrase: 'only entire chains enter into human computation' ). In addition, the role the features play within the internal configuration of the phrase may also contribute to this "chaining" effect--if we conclude that in order to properly match-up features (and proceed to check them) the words must be adjacent to one another. So in a strict sense, the adjacency condition (on checking) is seen as continuing even after the actual checking has taken place (before movement) for reasons having to do with Human Computational (Interpretability) of Language (CHL). Consider below how a break in the DP Chain would violate CHL Interpretability:

(220)

While reflecting on the ill-formed split DP structure as represented in (220) above on one hand, consider how the full range of such sentence types might violate this "Chain Principle" on the other.