Syntax from Phrase Structure to Merge

 

Deep Syntax — From Phrase Structure to Merge: The Hidden Mechanics of Grammar

Intro

Welcome.

What follows is not a description of language in the ordinary sense. It is an attempt to trace what lies beneath it: the computational architecture that allows finite human minds to generate infinite structured expressions.

At first glance, syntax appears to be a system of rules governing sentence formation. But as we move deeper, from phrase structure rules to transformations, from constraints to features, and finally to Minimalism, we begin to see something more radical emerging:

language is not a set of rules about sentences, but a recursive system for building structure itself.

This essay develops that trajectory in two stages:

first, classical generative grammar as a system of trees and transformations;

then Minimalism, where nearly everything collapses into a single operation: Merge.

1. Syntax as a System of Expansion

Traditional generative grammar begins with a simple insight:

A finite set of rules can generate an infinite number of sentences.

We formalize this using phrase structure rules:

S → NP + VP
VP → V / V + NP / V + NP + NP / V + S
NP → Det + N / Proper N / N + PP
PP → P + NP
N → Adj + N / lexical entries

At first, this looks like a descriptive taxonomy. But its implications are deeper:

Grammar is not a list of sentences
It is a generative system

From a small rule set emerges potentially unbounded expression.

This is the first foundational claim of generative linguistics:

Finitude of rules, infinity of sentences.

2. Generation and Parsing — Two Directions of the Same System

Once rules exist, they can be used in two ways:

Generation

Start with S → expand until a sentence emerges.

Parsing

Start with a sentence → reconstruct the structure that could have generated it.

These are not separate cognitive activities. They are inverse processes over the same system.

This leads to a deeper claim:

The mind does not store sentences. It computes them.

Language, then, is not retrieval; it is derivation.

3. The Hidden Structure Problem

Consider:

“Mary saw the mouse in the house”

This sentence admits multiple structural interpretations.

Why?

Because linear order does not determine hierarchical structure.

This leads to a key realization:

Surface strings underdetermine syntax.

So grammar must involve:

hierarchical structure
invisible attachments
structural ambiguity resolution

Thus:

Grammar is not linear sequencing; it is hierarchical computation.

4. Deep Structure, Surface Structure, and Transformations

To explain discrepancies between meaning and form, classical generative grammar introduces two levels:

Deep Structure

Represents underlying semantic relations:

who did what to whom
predicate–argument structure

Surface Structure

Final expressed form:

word order
stylistic variation
syntactic rearrangement

Transformations connect the two.

Example:

Deep:

Mary saw the mouse

Surface:

The mouse was seen by Mary

This is not lexical substitution; it is structural reorganization.

5. Syntax as Movement and Constraint

Transformations include:

movement
deletion
insertion
reordering

But unrestricted transformations overgenerate possibilities.

So grammar introduces constraints:

locality restrictions
island constraints
structure preservation
movement limitations

This yields a key shift:

Grammar is not only generative, but it is also restrictive.

It does not merely produce sentences; it filters impossible ones.

6. Features — The Invisible Mechanism of Agreement

Modern theory replaces many surface rules with features:

[+plural], [−plural]
[+tense]
φ-features (person, number, gender)

Agreement is no longer a rule like:

“add -s in third person singular”

Instead:

it is a feature-checking process during derivation.

Example:

“The cats runs”

fails because:

subject = plural
verb = singular
feature mismatch → derivational crash

Grammar becomes:

a system of formal consistency checking.

7. Recursion — The Engine of Infinite Expression

Recursion allows structure to embed within structure:

NP → NP + PP
VP → VP + PP
N → Adj + N

Examples:

the cat
the furry cat
the very furry cat in the house near the river
the man who said that the woman who believed that…

This is not stylistic elaboration.

It is structural necessity:

recursion allows infinite generation from finite rules.

8. Ambiguity — A Structural Property, Not an Error

Ambiguity arises naturally because:

linear strings collapse structure
multiple trees map to one surface form
attachment sites are underspecified

Thus ambiguity is not noise.

It is:

a structural consequence of hierarchical compression.

Meaning is not ambiguous because language is flawed—it is ambiguous because structure is richer than surface form.

9. Parsing as Computation

Parsing is not interpretation alone.

It is:

reconstruction of hidden derivational structure from observable output.

This aligns syntax with broader computational problems:

decoding signals
reconstructing latent variables
interpreting compressed data

Thus syntax becomes:

a theory of human symbolic computation.

10. The Minimalist Turn — Why Syntax Becomes Simpler

Classical grammar contains:

phrase structure rules
transformations
constraints
feature systems

Minimalism asks a deeper question:

What is the smallest possible system that can generate language?

The answer radically simplifies everything.

11. Merge — The Core Operation of Syntax

At the center of Minimalist theory lies one operation:

Merge

Merge combines two elements into one structured object:

Merge(A, B) → {A, B}

That is all.

From this single operation, everything follows.

12. External Merge — Building Structure

External Merge combines separate items:

Merge(the, cat) → [the cat]
Merge(saw, [the mouse]) → [saw the mouse]
Merge(Mary, [saw the mouse]) → sentence structure emerges

There are no phrase structure rules.

Only:

recursive combination.

13. Internal Merge — The Origin of Movement

Movement is reinterpreted:

not displacement, but re-merging within structure.

Example:

Mary saw the mouse

→ Which mouse did Mary see?

The object is not moved—it is re-merged at a higher structural position.

Thus:

movement is an effect of recursion, not a separate operation.

14. Why Merge Generates Hierarchy

Merge inherently produces:

binary branching
nested structures
hierarchical embedding

Therefore:

hierarchy is not imposed on language; it is produced by the operation itself.

15. Features as Constraints on Merge

Not all combinations are allowed.

So Merge is regulated by features:

agreement conditions
case requirements
tense compatibility

Example:

“The cats runs”

fails because feature mismatch blocks derivation.

Grammar becomes:

a system of constrained structure-building.

16. Economy Principles — The Logic of Minimal Effort

Minimalism introduces optimization constraints:

Least Effort: choose simplest derivation
Last Resort: move only if necessary

Grammar behaves like an optimization system:

not maximal generation, but minimal computation.

17. The Collapse of Grammar into One Function

At its most abstract level:

all syntactic structure reduces to recursive Merge under feature constraints.

Everything else becomes:

derived
emergent
or descriptive shorthand

Phrase structure rules become summaries of repeated Merge.

Transformations become properties of re-merge.

Trees become visualizations of computation.

18. Language as a Mathematical System

Merge behaves like a recursive function:

takes two inputs
produces structured output
applies repeatedly
generates infinite hierarchy

Thus language resembles:

a naturally occurring formal system of symbolic computation.

Conclusion — What Syntax Finally Reveals

If we follow the trajectory from phrase structure to Minimalism, grammar undergoes a complete transformation:

from rules → to operations
from lists → to recursion
from transformations → to derivation
from complexity → to minimal computation

And at the end, one idea remains:

Language is not a set of structures. It is a single recursive act repeated endlessly.

If earlier linguistics asked what sentences are possible, Minimalism asks something deeper:

What is the smallest computational system that makes sentence formation possible at all?

And the surprising answer is:

almost nothing, just Merge constrained by features.

From this minimal core, everything else emerges:

hierarchy, meaning, ambiguity, and the infinite expressivity of human language.

Introduction to Linguistics: Syntax 3