X Bar Theory

X-Bar theory is part of Chomskian linguistics that is used to flesh out our Phrase Structure Rules. Essentially we can layer XPs inside XPs by labelling the top one XP and then each one after than X' (X-Bar).

Each XP can only have one actual XP (up the top), and one head X, but as many X's as necessary.

The PSRs are now updated.

Binary Branching

Each level of the phrase structure tree now has only two nodes at each level.

Types of Modifiers



Specifier Rule

$XP \rightarrow (YP) X’$
The specifier is the YP - a daughter of XP and a sister of X'.

Having revised D's to be part of a DP, subjects are now the only types of specifiers we have:

  • Possessors of 's genitives in DPs
  • Subjects of clauses of TPs
  • Small Clauses

Small Clauses

Small clauses are characterised by having no inflection or verbal predicate, so they have no TP and no CP (meaning the subject has to go into the specifier of the predicate).

e.g. "Bill considers Peter a fool"
There's no verb in "Peter a fool".


Adjunct Rule

$X’ \rightarrow (ZP) X’ \text{ or } X’ \rightarrow X’ (ZP)$
The adjunct is the ZP - a daughter of X' and a sister of X'.
(Adjuncts must be sisters to X's, not to the head).

A phrase can have multiple adjuncts, and they can be reordered relative to each other.

Complement Rule

$X’ \rightarrow X (WP)$
The specifier is the WP - a daughter of X' and a sister of X.

A Phrase can only have one complement, and the complement comes closest to the head.

The Complement/Adjunct distinction can capture ambiguity There is strong evidence for the C/A distinction in NPs and VPs.


If an XP modifies some head Y, then it must be dominated by some projection of Y (i.e., it must be dominated by Y’, …, Y’, YP).

Conjunction Rule

\begin{align} X^n \rightarrow X^n \text{ Conjunction } X^n \end{align}

e.g. "The red and blue house" vs "The red and cat".

Complements can be conjoined with complements, and adjuncts can be joined with adjuncts. They cannot be joined together.

"The student of linguistics and [of] philosophy with red hair and [with] a tattoo".
*"The student of linguistics and with red hair"

X-Bar Paramaters

When acquiring languages we have three switches, each of which has three states, for our x-bar settings.

Flat Structure to Bar Structure

Flat Structure - by which we mean the style of the phrase structure trees we were doing up until now - is kind of unsatisfactory for representing constituents. We're not sure where they're grouped, and are unable to replace them with "one".

TP Rule

TP, S, IP and AgrP are all the same thing (for our purposes).

So far we've had a TP with an optional T. This can't be correct, given X-Bar theory, so from now on the T must be obligatory (even if it is just Ø).

The TP specifier becomes the subject, and the complement the VP.

But what is the T head?

Predicates select based on the head - the head gives a category to the phrase, and determines what can be selected.

In a clause we select based on finiteness, and hence the T head of a TP must provide that feature.

Tense is represented in inflection (so Infl is another name for T), hence auxiliaries and agreement suffixes become the T head.


T lowering

This T head is great, except that there's a difference between inflectional suffixes and auxiliaries; suffixes must be attached to something, and auxiliaries are free.

To deal with the weird ordering we can hence end up with, we put a hacky measure in to 'lower' the suffixes and attach to the verb'.


This is the one exception to not breaking apart words during syntax.

Irregular Verb Morphology

"John runs" is easy, but "John ran" is not - the suffix is internal, not an affix.

What we do here is just state the it's a past tense marker, as the following



CP specifier is often empty, but can be used for wh- movement. More later on (chapter 11).

Every embedded clause is a CP, and all clauses have an obligatory C head (it may just be null Ø).

Null Question Marker

We can see from yes/no questions in other languages that special question markers are in complementary distribution with C, and hence know that our questions have a phonologically null complementiser that we use [+Q] for, and raise T up to in order to pronounce the question.

$\phi C_{[+Q]}$

If Test

We have a pronounce [+Q] for embedded clauses - "if".
Given SAI (Subject/Aux Inversion) is disallowed with if, we can say that SAI is a diagnostic for the presence of a C in English.
*"I wonder if has blah"

Evidence for Non-Questions Having a C

We know that non-questions must have a C (albeit null) because of the conjunction rule - anything conjoined with a CP must also be a CP with a C.

e.g. "You can lead a horse to water but can you make him drink?" <- the second clause has a null C (tested with SAI), therefore first must also.


Our specifier rule "XP -> (YP) X'" requires that the specifier be phrasal. Hence our determiners (commonly in specifier position), must be part of a phrase.

So the 'D' isn't a phrase - it's the head of a DP.


Evidence: 's Genitives

"'s" in "The man's coat" is not a suffix, it attaches to phrases. It's in complimentar distribution with determiners, which mean they must be the same thing:

  • "The man's coat" vs "The coat of the man"
  • (a 's genitive vs a free genitive).

DPs and NP Possessors

Once we say that "'s" modifies "coat", where do we put "The man"?

We change the whole way we've done NPs, to put them inside a DP:


N-Bar Structure


If we can replace the block above with the word "one", then the block must be a constituent. But under flat structure it's clearly not. What can we do?

We can replace the above Flat Structure tree with an N-Bar tree:


We can replace any N' node with "one". (Not heads or NPs, N's).

Complements vs Adjuncts in NPs


We can replace N's with 'one', but not Ns. Hence adjuncts can follow 'one', but complements can't.


PP's can only be complements to nouns if they start with 'of' (else adjuncts).

V-Bar Structure


We can replace the above block with the phrase "does so (too)". In order to better express this we come up with the following V-Bar Tree:


I.e. a recursive V' structure, with VP-V'-V'-V'-V'-V.

Any V' can be replaced with "does/did so (too)", but not VP or V. Only adjuncts can follow 'did so'.

Complements vs Adjuncts in VPs

In verb phrases the direct object is always the complement. Some verbs take multiple complements (e.g. 'give' and 'put' take two - "Beth gave the the apple to John"), and everything else is an adjunct.


Two V heads can be conjoined together, and two V's can be joined together, but they cannot be cross-joined. (E.g. 'eats and drinks sugar' or 'eats beans and tosses salads'). To be joined together they must be both either adjuncts or complements.

P-Bar Structure


Again, blocks of nodes within a PP can be replaced with "so", and hence we need P' structures, not flat P structures.


Adj and PP Adjunct vs Complements

AdjPs and PPs have much weaker adjunct vs complement tests.