The structural theory of syntax known as X-bar Theory (q.v. or this will be incomprehensible) builds sentences recursively out of phrases that have a binary tree structure. There is a main branch down to a head word, and side branches qualifying the head. The topmost side branch is called the specifier:

       XP
      /  \
     /    \
   Spec   X'
         /  \
        /    \
       X    Comp

X denotes any category. While there can be multiple branches at the intermediate X' level, there is only one maximal projection XP, only one head X, at most one specifier (Spec), and at most one complement (Comp). The purpose and content of the specifier depend on what category X is.

Most words belong to lexical categories: noun, verb, adjective, or preposition. The head of these is notated N, V, A, or P, and the phrases they build are notated NP, VP, AP, and PP. In addition, some grammatical function words such as that and will are representatives of functional categories, the most important of which are C, I, and D (complementizer, inflection, and determiner). These have been integrated into X-bar Theory and are now considered to project their own phrases CP, IP, and DP, which also have specifier and complement positions.

Roughly speaking, the specifier of a lexical phrase is a qualifier inherent in it, whereas the specifier of a functional phrase is there to satisfy grammatical relationships. The relationship between a specifier and its head allows grammatical interaction, such as agreement between subject and verb, or between adjective and noun. A word may move into an empty specifier position for checking that it has the correct match of features. The detailed discussions that follow are subject to revision as alternative versions of X-bar Theory continue to be proposed. Much of it is problematic and unconvincing.

The specifier of an XP is variously notated Spec,XP or Spec-XP or [Spec, XP].

Spec,NP

The obvious candidate for the Spec,NP position is the determiners, words which in English include a, the, this, that, my, your, some, every, no. As the specifier is unique within a single phrase, specifiers cannot co-occur, and indeed none of the determiners co-occur: you can't say *the my car or *every these books.

Another class of expression that appears in front of nouns (in English) is genitive noun phrases: Mary's car; the Prince of Denmark's march; my mother's grey hair. These qualifiers preclude any of the determiners from appearing: the the and my in these examples belong inside the genitive noun phrase the Prince of Denmark and my mother, and do not qualify march or hair.

Unfortunately more recent theories agree that the D or determiner projects above the NP, so my mother's grey hair is a DP with an NP inside it, not an NP with a DP inside it.

Spec,VP

A specifier is a kind of subject: it is analogous to the subject of a sentence, and the actual subject is considered to be in the Spec,IP position. It is perhaps more natural to think of the subject as Spec,VP, and this is now widely held, in what is called the VP-Internal Subject Hypothesis. The subject DP starts inside the VP:

       VP
      /  \
     /    \
   Subj   V'
         /  \
        /    \
       V    Obj

In this position it is governed by the verb, and can be assigned the θ-role pertaining to the subject. Having been assigned the correct semantic interpretation, it then moves elsewhere to get its syntactic marking: it moves up into its surface position of Spec,IP.

Spec,AP

If adjectives project a phrase of their own, with a complement (as in proud of Mary), then they should have a specifier. It is unclear what this is: words like very have been proposed, but as you can be very very ill and really very clever, this seems to violate the uniqueness of specifiers.

Spec,PP

What can go in front of a preposition and still keep it preposition-like? Well there are a few such adverbs: just down the street; right across the room. Possible.

Another candidate is miscellaneous DPs that can qualify PPs: as in three minutes before midnight; a long way from Tipperary.

Spec,DP

Functional specifiers are for agreement, and as nouns and determiners or adjectives agree, it may be that Spec,DP is a landing site for checking. The only agreement that determiners or adjectives have in English is in the number of these, those.

Spec,CP

At last an interesting one. CP is the whole sentence or clause. Spec,CP is the landing site for material that moves to the front of the sentence or clause. There are three principal classes of fronting, and Spec,CP receives the two structure-preserving movements, but not the structure-building one. Also, it may be done overtly or covertly, depending on the language.

To dispose of the structural distinction first, a structure-preserving movement is one into a position that already exists, albeit empty, whereas a structure-building one creates some new tree structure. These latter kinds have an intuitively 'looser' grammatical dependence, shown by intonation: examples are topicalization like This book I like (though the other one is boring) and left shifting like This book, what do you think about it?

In contrast, a move into an existing position feels grammatically tighter: What did you read? is just as much a single unit as I read that book. The question word what starts as the underlying object of read, and moves into the Spec,CP position, leaving a trace in its original object position: Whati did you read ti?. The fronted question word binds its trace, so that they can be logically interpreted as co-referring.

Another kind of co-reference is between a quantifer such as every or some and the variable it binds. A sentence like Every child knew some song has two possible readings, depending on whether it was the same song they all knew. These are expressible as two different logical forms, with two ways of binding it: either Every child x, some song y, x knew y, or Some song y, every child x, x knew y.

In English the 'every' and 'some' elements remain in their normal subject and object positions. There is no fronting to Spec,CP here, is there? Aha! Welcome to syntax. The claim is that the brain actually does construct a level called LF or logical form in which the move has been made. The reason we don't hear it is that the derivation of phonetic form or PF splits off from the derivation of LF at a stage before the move takes place. But the move does still take place, with the quantifiers fronted by the time they get to be semantically interpreted.

Actually we do see a bit of evidence of this process in English. Notice that in questions there is verb inversion: John is reading something shows up as What is John reading?. Not only does the question move to Spec,CP but the auxiliary verb hops forward into the head C. Now this also happens with some quantifier-like expressions: Never have I seen the like; Nothing have I seen to compare.

Spec,IP

The I of a sentence is the inflection for tense and agreement. The whole sentence is regarded as IP, the maximal projection of I, then further wrapped up in a CP. The specifier of IP is the normal position for the subject of a clause. Under the VP-Internal Subject Hypothesis the subject starts in Spec,VP and moves into Spec,IP for agreement checking.

The original idea was that the subject was assigned nominative case in this specifier position, whereas verbs assigned accusative case to their complements and prepositions might assign other cases to their complements. To match this, verbs could agree with their subjects and (in some languages anyway) with their objects. So two kinds of relationship were involved, Spec-head and Head-complement. This was deemed theoretically unsatisfactory, and evidence was brought forward suggesting a better idea would be if all agreement was under the Spec-head relationship.

So IP was split into several nested phrases, relating to its subject-agreement, tense, and object-agreement properties. These were designated AgrS (or AgrS), T, and AgrO, and each projects a phrase AgrSP, TP, and AgrOP, and each of those has a specifier, which is the checking position for the particular agreement feature. The subject moves to Spec,AgrS to check or match its agreement features, and the object likewise does to Spec,AgrO.

Of course if this movement was all overt you would get a particular order of subject, verb, and object; whereas all six logically possible orders occur in various languages. So languages are regarded as differing in whether parameters for AgrS or AgrO movement are strong or weak: strong parameters force movement before SpellOut, the split between LF and PF, so actually get represented in the phonetic form, whereas weak movements happen covertly after SpellOut, on the transition to logical form. They are nevertheless required in order to satisfy the grammatical checking that makes the sentence logically interpretable.

I was made somewhat less confused by repeated reading of the relevant sections in:
Ouhalla, J. (1999) Introducing Transformational Grammar, Arnold.
Culicover, P. (1997) Principles and Parameters, Oxford.

You are quite at liberty to refuse to believe a word of this seemingly mad theory. Me, I have a 3-hour exam on this sort of thing in May, so I'm obliged to take it in.

Log in or register to write something here or to contact authors.