BEYOND TREES: A Cartography of Human Syntax

 

A Cartography of Human Syntax beyond Trees

1 — The Partial Limits of Tree Syntax 

2 — Category Fluidity Across Languages

3 — Agreement as Alignment Systems

4 — Structured Omission 

5 — Word Order as Cognitive Pressure

6 — Multimodal Syntax 

7 — Rethinking Deep Structure 

Conclusion 

1 The Partial Limits of Tree-Based Syntax

1.1 The Foundational Commitment of Generative Syntax

Modern syntactic theory, particularly within the generative tradition, assumes that:

Sentences are hierarchically structured objects represented as trees.

This assumption is embedded in:

X-bar theory (Chomsky, 1970s–80s)

Minimalist Program (Chomsky, 1995–present)

standard constituency tests

For example, English sentence:

“The boy saw the dog”

is represented as:

[TP [DP The boy] [T’ T [VP saw [DP the dog]]]]

This representation encodes a strong theoretical claim:

linear order is derived from a hierarchical structure.

However, this assumption must be tested against typologically diverse languages.

1.2 Non-configurational Languages: Warlpiri

In Warlpiri (Pama–Nyungan family), word order is highly flexible.

Example:

Ngarrka-ngku ka wawirri panti-rni

man-ERG AUX kangaroo spear-NONPAST

“The man is spearing the kangaroo”

But permutation is possible:

Wawirri-ngku ka ngarrka panti-rni

Kangaroo-ERG AUX man spear-NONPAST

Key observation:

grammatical relations remain stable
word order does not encode hierarchy directly

Implication:

Hierarchical structure is not transparently mapped to linear order.

1.3 Polysynthesis and Morphological Saturation: Inuktitut

In Inuktitut:

“tusaa-nngit-su-qar-uma-lauq-tuq”

hear-NEG-PART-have-want-PST-3SG

“He did not want to have heard it”

Here:

a full clause is encoded morphologically

syntactic relations are internal to a single word

Problem for tree syntax:

Where is constituency located?

inside morphology?
or external to it?

Tree models struggle to represent internal syntactic complexity inside morphological units.

1.4 Discontinuous Dependencies: German

German shows long-distance scrambling:

dass den Mann die Frau gesehen hat

that the man.ACC the woman.NOM seen has

“that the woman saw the man”

Word order variations:

dass die Frau den Mann gesehen hat

Observation:

thematic roles remain stable
surface order is flexible
dependencies are non-local

Implication:

Linear order ≠ structural hierarchy mapping is not fixed.

1.5 The Problem of Constituency Tests

Traditional constituency tests include:

movement
coordination
substitution

But cross-linguistically:

some languages lack clear movement diagnostics

coordination behaves differently in polysynthetic systems

substitution effects vary widely

Conclusion:

constituency tests may be language-specific heuristics rather than universal diagnostics.

1.6 Alternative Modeling: Dependency Networks

Instead of trees, consider:

Model A: Tree structure

binary branching
hierarchical dominance
movement operations

Model B: Dependency network

relations between semantic roles
morphosyntactic alignment
discourse anchoring

Example:

Instead of:

      S
     / \
   NP   VP

We model:

agent → event
event → patient
discourse context → ordering

Insight:

Syntax may be better represented as a graph of dependencies rather than a hierarchical tree.

1.7 Interim Hypothesis

Based on cross-linguistic evidence:

Tree structures may represent a projection of dependency relations under specific cognitive and linearization constraints, rather than a universal architecture.

1.8 Theoretical Consequences

If this hypothesis holds:

(1) Movement theory is derivative

Movement becomes:

a repair strategy for linearization, not a core operation

(2) Phrase structure is optional

Not all languages require:

fixed constituency grouping

(3) Hierarchy is emergent

Hierarchy may arise from:

processing constraints
morphological packaging
discourse structuring

1.9 Interim Conclusion

The tree model is not false.

It is:

a highly successful local approximation of a broader system of relational organization.

The challenge is not to reject it, but to:

identify its domain of validity

and map its boundaries

 2 Category Fluidity Across Languages

2.1 The Classical Assumption: Lexical Categories as Universals

Most syntactic frameworks assume that human language is structured around stable lexical categories, typically:

Noun (N)
Verb (V)
Adjective (A)
Preposition/Adposition (P)

This assumption is deeply embedded in:

X-bar theory (Chomsky, 1981)
Minimalist Program (Chomsky, 1995–)
traditional typological grammar

Under this view, syntactic computation operates over:

pre-labeled lexical atoms belonging to fixed categories.

However, cross-linguistic research suggests that category stability may not be universal, but rather a property of specific language families (particularly Indo-European).

2.2 The Empirical Challenge: Category Fluidity in Global Languages

A growing body of typological research (e.g., Evans & Levinson, 2009; Croft, 2001) suggests that lexical categories vary significantly across languages.

Three recurrent patterns emerge:

Fluid categorization systems

Predication-based lexical organization

Context-driven category assignment

We examine each in turn.

2.3 Salish Languages: Predication Without Noun–Verb Division

In many Salish languages (e.g., St’át’imcets / Lillooet Salish), lexical items do not cleanly divide into nouns and verbs.

Example:

q’wátswa7

“be a child / childing / to be a child”

The same root can function as:

predicate

argument
nominal expression

Illustration (St’át’imcets pattern):

kwán-lhkan q’wátswa7

know-1SG child

“I know the child / I know (someone who is) a child”

But also:

q’wátswa7-lhkan

child-1SG

“I am a child”

Key observation:

There is no strict lexical boundary between:

naming an entity
predicating a state

Theoretical tension:

In standard syntax:

nouns are arguments
verbs are predicates

In Salish:

predication appears to be primary, and category secondary.

2.4 Riau Indonesian: Extreme Category Flexibility

Riau Indonesian (documented in typological linguistics literature, e.g., Gil, 2005) shows extreme lexical flexibility.

Example root:

makan (“eat”)

Can function as:

Predicate:

Dia makan

“He eats”

Noun-like usage:

makan sudah

“the eating is done / the meal is finished”

Modifier-like usage:

orang makan

“eating person / person who eats”

Crucial observation:

There is:

no overt morphological marking of category shift

category determined by syntactic position and discourse context

Implication:

lexical items do not carry fixed syntactic category features in the lexicon.

Instead, category emerges in structure.

2.5 Classical Nahuatl: Nominal–Verbal Continuity

Classical Nahuatl (Uto-Aztecan family) shows a systematic blur between nouns and verbs.

Example:

ni-pilli

1SG-child
“I am a child”

But also:

pilli

“child / to be a child”

And:

ni-cua

1SG-eat

“I eat”

Structural observation:

nominal and verbal forms are not strictly separated
predication is morphologically distributed
lexical roots are category-neutral at base level

2.6 Comparative Insight: Three Systems of Category Organization

Across these languages, we observe three distinct strategies:

Language Type	Category System
Indo-European (e.g., English)	rigid lexical categories
Salish	predication-centered system
Riau Indonesian	context-driven categorization
Nahuatl	morphologically flexible roots

2.7 Theoretical Problem for Standard Syntax

In Minimalist syntax:

lexical items enter derivation with category features (N, V, etc.)
syntax manipulates pre-labeled items
structure is built hierarchically over categories

But in the systems above:

category assignment is not stable prior to syntactic context.

This creates a core theoretical tension:

Either:

Universal categories exist but are invisible in surface forms

Or:

Categories are not universal primitives

2.8 Competing Theoretical Responses

(A) Lexicalist Defense (Traditional View)

Some formal approaches might argue:

categories exist at abstract LF level

surface flexibility is derived from movement or coercion

However:

no consistent morphological evidence supports hidden fixed categories in many cases

category behavior is stable at discourse level, not just surface

(B) Constructionist View (Goldberg, Croft)

Construction Grammar proposes:

grammatical constructions, not lexical categories, determine syntactic behavior.

This aligns more closely with observed data:

category emerges from constructional context

lexical items are underspecified

(C) Distributed Lexical Hypothesis (Proposed here)

We refine further:

lexical items may be best modeled as category-neutral semantic bundles, with syntactic category assigned dynamically through structural embedding.

2.9 Revised Hypothesis

Based on cross-linguistic evidence:

Lexical categories are not universal primitives of syntax, but emergent classifications arising from interaction between lexical semantics, morphosyntax, and constructional context.

2.10 Consequences for Syntactic Theory

If this hypothesis is correct, then:

(1) Syntax is not category-driven

It is:

relation-driven (dependency + structure)

(2) The lexicon is underspecified

Lexical entries may encode:

semantic core

argument structure potential

not syntactic category

(3) Phrase structure becomes derivative

If categories are unstable:

projection of phrases becomes construction-dependent, not universal

2.11 Implications for Typology

This reclassification suggests:

“noun” and “verb” may be language-specific analytical conveniences

not universal cognitive primitives

A more accurate typology may classify languages by:

degree of category rigidity

degree of predicative flexibility

constructional dependency strength

2.12 Counterargument: Why Category Systems Still Appear Universal

Formal syntax might respond:

even flexible systems show emergent noun/verb behavior in discourse

processing requires categorization for real-time parsing

Response:

This may indicate:

categories are processing-level stabilization mechanisms rather than lexical primitives.

2.13 Interim Conclusion

The cross-linguistic evidence suggests a gradual shift:

From:

“languages have nouns and verbs”

To:

“languages distribute predication, reference, and modification across different structural mechanisms, sometimes independently of fixed lexical categories.”

2 Recap STATEMENT

Lexical categories, long treated as foundational to syntactic theory, may instead represent:

emergent stabilizations of more primitive semantic and relational systems under specific structural constraints.

3 Agreement as Alignment Systems

3.1 The Classical Model: Agreement as Feature Checking

Within generative syntax, agreement is typically understood as:

a formal operation of feature matching between a probe and a goal.

In the Minimalist Program (Chomsky, 1995–present), agreement is implemented through:

φ-features (person, number, gender)
Agree relation between functional heads and noun phrases
feature valuation and deletion mechanisms

Example (simplified English):

The boy walks

(3rd person singular agreement on verb)

This model assumes:

agreement is a single unified syntactic mechanism

features are abstract and universal

agreement is structurally identical across languages

However, cross-linguistic evidence suggests that what is called “agreement” may actually consist of multiple structurally distinct systems that have been unified under a single theoretical label.

3.2 The Core Problem: The Overextension of “Agreement”

Across languages, phenomena grouped under “agreement” differ radically in:

morphological realization
semantic domain
syntactic locality
obligatoriness
directionality

This raises a foundational question:

Is “agreement” a single syntactic phenomenon, or a descriptive umbrella covering unrelated alignment processes?

3.3 Bantu Languages: Noun Class Agreement as Systemic Classification

In Bantu languages (e.g., Swahili), agreement extends beyond person/number/gender into noun class systems.

Example (Swahili):

ki-tu ki-kubwa ki-meanguka

7-thing 7-big 7-fall.PST

“The big thing has fallen”

Agreement occurs across:

noun
adjective
verb
demonstratives

Key observation:

agreement is not feature matching between two elements only
it is a system-wide classificatory alignment

Theoretical tension:

In Minimalism:

agreement is local (probe-goal)

In Bantu systems:

agreement is distributed across the entire clause architecture

3.4 Australian Languages: Spatial and Relational Agreement

In many Australian Aboriginal languages (e.g., Arrernte, Warlpiri systems), agreement can encode:

spatial orientation
deixis
directionality of motion

Example pattern (generalized from Warlpiri-type systems):

Verb forms may vary depending on:

direction of movement
spatial relation of participants
discourse positioning

Key insight:

Here, “agreement” is not φ-feature based, it encodes:

geometric and spatial relations between participants in discourse space.

Implication:

Feature theory based solely on person/number/gender is insufficient.

3.5 Tibeto-Burman Languages: Evidential Alignment

In several Tibeto-Burman languages (e.g., Tibetan varieties, Sherpa, etc.), verbal morphology encodes:

source of information
epistemic status
evidential certainty

Example (Tibetan-style system, simplified):

“He went”
may differ depending on whether:
speaker saw it
inferred it
heard it reported

Crucial observation:

This “agreement” is not with arguments but with:

epistemic stance of the speaker

Theoretical consequence:

Agreement is not necessarily argument-structure-driven.

It may align:

clause content with speaker epistemology.

3.6 Cross-System Comparison: Three Types of Alignment

We can now distinguish three fundamentally different systems:

(A) Referential Agreement (Indo-European type)

person
number
gender
NP–verb alignment

(B) Classificatory Agreement (Bantu type)

noun class systems
global concord across clause
category-wide alignment

(C) Epistemic / Spatial Alignment (Australian & Tibeto-Burman types)

evidentiality
spatial orientation
discourse anchoring

3.7 Theoretical Breakdown of the Unified Agreement Model

The Minimalist assumption is:

all agreement reduces to φ-feature valuation under Agree.

But empirical diversity suggests:

Problem 1: Feature Heterogeneity

Not all agreement involves φ-features.

Problem 2: Domain Expansion

Agreement extends beyond argument structure.

Problem 3: Directional Variation

Some systems are:

upward (NP → verb)

downward (functional head → NP)

or system-wide (Bantu concord chains)

3.8 Revised Hypothesis: Agreement as Alignment Mechanisms

We propose a revised model:

“Agreement” is not a single syntactic operation, but a class of alignment mechanisms that synchronize grammatical elements under different functional pressures.

These pressures may include:

referential tracking
classification stability
spatial mapping
epistemic grounding
discourse cohesion

3.9 Formal Consequence: Collapse of Feature Uniformity

If this is correct:

(1) φ-features are not universal primitives

They are one subtype of alignment feature system.

(2) Agree is not a single operation

It may be a family of structurally distinct relations.

(3) Syntax is modular at the alignment level

Different alignment systems operate semi-independently.

3.10 Alternative Modeling Framework: Alignment Networks

Instead of a single “Agree” operation, we propose:

Alignment Network Model

A clause consists of interacting alignment layers:

Referential layer (who/what)
Classificatory layer (type/shape/category)
Epistemic layer (certainty/source)
Spatial layer (position/orientation)

Each language:

selects and weights different alignment layers differently.

3.11 Counterargument: Minimalism’s Possible Defense

A Minimalist response might argue:

all these systems are reducible to feature valuation
differences are morphological, not syntactic
underlying architecture remains uniform

Response:

This position risks:

collapsing all grammatical variation into post-hoc abstraction, without independent diagnostic justification.

If everything is “features,” then:

the concept of feature loses explanatory power

3.12 Interim Synthesis

Cross-linguistic evidence suggests:

agreement is not structurally uniform
feature types are heterogeneous
alignment is multi-systemic rather than single-operation based

Thus:

“agreement” may be a descriptive label for multiple independently evolved grammatical alignment strategies.

3 FINAL STATEMENT

Agreement, traditionally treated as a cornerstone of syntactic uniformity, may instead represent:

a convergence of distinct alignment systems that operate across different grammatical dimensions—referential, classificatory, spatial, and epistemic.

The challenge for syntactic theory is not to unify these systems prematurely, but to:

map their diversity before attempting theoretical compression.

4 The Syntax of Omission: Structured Absence in Human Language

4.1 The Classical Problem: Omission as Deletion

Within formal syntax, omission is typically analyzed as:

ellipsis (VP-ellipsis, TP-ellipsis)
deletion under identity
recoverability via LF reconstruction
null categories (pro, PRO, traces)

Example (English VP ellipsis):

John will eat pizza, and Mary will __ too.

Standard analysis assumes:

underlying structure is fully present but partially deleted at PF.

This creates a strong theoretical commitment:

absence is always derived from presence.

However, cross-linguistic evidence suggests that omission may not always be “deletion from a full structure,” but sometimes a primary grammatical option encoded in the system itself.

4.2 Pro-Drop Systems: Omission Without Loss of Reference

In languages such as Spanish, Italian, Japanese, and Turkish, subject omission is grammatically licensed.

Example (Spanish):

Ø hablo español

speak.1SG Spanish

“I speak Spanish”

No overt subject is required because:

verbal morphology encodes person/number
discourse context resolves reference

Key observation:

The subject is not “deleted” in derivation, it is:

never obligatorily externalized in the first place.

4.3 Japanese: Argument Omission as Discourse Structure

Japanese allows extensive omission of arguments:

Ø tabeta

ate

“(Someone) ate”

Context determines interpretation:

speaker

addressee

third party

But crucially:

omission is not exceptional

it is structurally normal

Example:

Ø kitto Ø shitteiru

probably know

“(I/you/he/she) probably knows (it)”

Implication:

Syntax does not require overt argument realization for grammatical completeness.

4.4 Turkish: Morphology-Driven Referential Recovery

Turkish allows pro-drop with rich agreement morphology:

Gel-di-m

come-PST-1SG

“I came”

The verb encodes:

tense
subject person
number

Key point:

Reference is recovered from morphology, not syntactic presence.

4.5 Ellipsis vs Non-Expression: A Critical Distinction

A central problem in syntactic theory is the conflation of:

(A) Ellipsis (derivational absence)

Structure is present but unpronounced.

(B) Non-expression (structural optionality)

Element is not required in representation.

Cross-linguistically, evidence suggests:

VP ellipsis behaves differently across languages
argument omission is often systemic, not reconstructive

Hypothesis:

Not all omission is ellipsis. Some omission reflects alternative structural encoding strategies.

4.6 Warlpiri and Non-Configurational Omission

In Warlpiri (Pama–Nyungan), argument structure is often encoded independently of surface order.

Example:

Ngarrka-ngku ka wawirri panti-rni

man-ERG AUX kangaroo spear

“The man is spearing the kangaroo”

But:

arguments may appear dislocated

clausal structure is distributed

Key insight:

What appears as “omission” is often:

distributed realization across morphology, clitics, and discourse structure.

4.7 Structured Absence: The Core Hypothesis

We propose a stronger reinterpretation:

Omission is not absence of structure, but a grammatically licensed mode of structural realization.

This implies:

syntax does not always require overt projection of all arguments

structural completeness is not dependent on phonological realization

4.8 Three Types of Omission Systems

Cross-linguistic data suggests at least three distinct systems:

(1) Morphological Recovery Systems

(e.g., Turkish, Spanish)

agreement morphology encodes arguments

omission is licensed by feature richness

(2) Discourse-Based Recovery Systems

(e.g., Japanese, Chinese discourse contexts)

reference tracked via discourse continuity

syntax allows radical ellipsis

(3) Distributed Argument Systems

(e.g., Warlpiri, polysynthetic languages)

argument structure is encoded across multiple morphological elements

no single locus of realization

4.9 Theoretical Conflict with Standard Minimalism

Minimalist syntax assumes:

full argument structure exists in narrow syntax

omission is PF deletion or LF reconstruction

However:

Problem 1: Overgeneration of hidden structure

Assuming full structure everywhere risks:

unfalsifiability

descriptive redundancy

Problem 2: Typological mismatch

Some languages do not behave as if “full structure” is ever fully constructed.

Problem 3: Morphology–syntax entanglement

In polysynthetic systems:

argument structure is not separable from morphology

4.10 Revised Hypothesis: Omission as Structural Strategy

We propose:

Omission is a grammatical strategy for distributing information across morphological, syntactic, and discourse systems, not a deletion from a fully realized syntactic object.

This reframes omission as:

economy of expression

structural compression

information distribution strategy

4.11 Implications for Theoretical Syntax

If this view is correct:

(1) Null elements are not uniform

“pro,” “PRO,” and ellipsis may not be the same type of object.

(2) Recoverability is not universal constraint

Some languages do not require syntactic recoverability in the same way.

(3) Syntax is not fully compositional in overt form

Meaning may be distributed across:

morphology

context

discourse history

4.12 Counterargument: The Generative Position

A generative linguist might argue:

all omission is syntactic presence with PF non-realization

covert structure preserves universality

differences are surface variation

Response:

This risks:

reducing empirical diversity to unobservable uniformity.

A theory gains strength not by making all data look the same, but by:

allowing principled variation

4.13 Interim Synthesis

Across languages, omission behaves as:

morphologically licensed (agreement-based)

discourse-conditioned (context-based)

structurally distributed (polysynthetic systems)

Thus:

omission is not a single phenomenon but a class of structurally distinct strategies for non-realization of grammatical elements.

4 FINAL STATEMENT

Omission is not the absence of syntax.

It is:

one of the primary ways in which syntax is realized across languages—through controlled non-expression, structural distribution, and morphological substitution.

5 Word Order as Cognitive Pressure

5.1 The Classical Typological Model: Word Order as Parameter

In formal syntax and typology, word order is commonly treated as a structural parameter:

SVO (English, Mandarin)
SOV (Japanese, Turkish, Korean)
VSO (Classical Arabic, Irish)

Within generative grammar, this is often modeled as:

movement operations applied to an underlying universal structure

or, in some variants:

parameter settings of head directionality or functional projections

This yields a strong theoretical assumption:

word order is primarily a syntactic property.

However, this view struggles to explain why:

word order flexibility correlates with discourse conditions

processing constraints systematically influence linearization

languages with similar syntax differ in ordering under cognitive load

This suggests that word order may not be purely syntactic.

5.2 The Processing Perspective: A Competing Hypothesis

Psycholinguistic research has long shown that sentence comprehension is constrained by:

working memory limitations
incremental parsing strategies
prediction and surprisal effects
attentional salience

This raises a fundamental question:

Could word order be partially shaped by cognitive optimization rather than syntactic configuration alone?

5.3 English vs Japanese: A Processing Contrast

Consider:

English (SVO):

The boy ate the apple.

Japanese (SOV):

男の子がりんごを食べた

boy-NOM apple-ACC ate

Structural observation:

Both encode identical semantic relations:

agent
patient
event

But differ in:

argument placement
verb position
dependency resolution timing

5.4 Incremental Processing Asymmetry

In SVO languages (English):

verb appears early
predicate structure is immediately available
argument completion is post-verbal

In SOV languages (Japanese):

arguments are accumulated before predicate
verb arrives late, completing structure

Cognitive interpretation:

These are not arbitrary configurations, they reflect:

different strategies for managing incremental interpretation under memory constraints.

5.5 German Scrambling: Word Order as Load Redistribution

German allows relatively free scrambling:

dass den Mann die Frau gesehen hat

that the man.ACC the woman.NOM seen has

“that the woman saw the man”

versus:

dass die Frau den Mann gesehen hat

Key observation:

grammatical roles remain stable
linear order shifts based on discourse emphasis

Processing interpretation:

Word order is used to:

reduce ambiguity
distribute information load
align with discourse focus

5.6 Warlpiri and Non-Fixed Linearization

In Warlpiri (Pama–Nyungan):

Ngarrka-ngku ka wawirri panti-rni

man-ERG AUX kangaroo spear

But constituents can appear in variable order.

Key insight:

grammatical relations are not encoded in fixed linear positions
morphology and case marking carry structural information

Implication:

linear order is not the primary carrier of syntactic relations.

5.7 Cognitive Pressure Hypothesis

We now propose a broader model:

Word order variation is partially shaped by cognitive pressure systems governing memory load, predictability, and discourse efficiency.

These pressures include:

(1) Memory constraint pressure

earlier placement of predictable elements reduces load

(2) Predictability optimization

high-surprisal elements are positioned strategically

(3) Discourse salience alignment

topical elements appear earlier or later depending on prominence

(4) Structural disambiguation pressure

word order reduces ambiguity in less morphologically rich systems

5.8 Cross-Linguistic Distribution of Pressure Systems

Morphologically rich languages (e.g., Turkish, Finnish):

freer word order
reliance on case marking
reduced linear constraints

Morphologically light languages (e.g., English):

stricter word order
reliance on position for interpretation

Intermediate systems (e.g., German):

mixed strategy: case + word order flexibility

5.9 Reframing Typology: From Parameters to Optimization

Traditional typology treats word order as:

fixed parametric choice (SVO vs SOV etc.)

We propose instead:

word order emerges from optimization over competing cognitive and communicative constraints.

This shifts typology from:

categorical classification

to

gradient constraint interaction model

5.10 Formal Syntax vs Processing Models: A Core Tension

Generative syntax assumes:

structure determines interpretation
linear order is derived

Processing models suggest:

interpretation emerges incrementally
linear order shapes processing load

The unresolved tension:

Is syntax primary and processing secondary, or are they co-constitutive?

5.11 Toward an Integrated Model: Syntax-Processing Coupling

We propose a hybrid hypothesis:

Word order is the surface outcome of an interaction between syntactic constraints and real-time processing optimization.

This implies:

syntax does not operate in isolation
linearization is co-determined by cognitive architecture
variation reflects different weighting of constraints across languages

5.12 Counterargument: The Autonomy of Syntax

A formal syntactician might argue:

word order differences are purely syntactic derivations
processing effects are performance-level, not competence-level
grammar must remain independent of cognition

Response:

While methodological separation is useful, empirical patterns suggest:

persistent alignment between grammatical structure and processing efficiency across languages is difficult to dismiss as accidental.

5.13 Interim Synthesis

Across languages, word order variation reflects:

structural constraints (syntax)
morphological encoding (case/agreement systems)
discourse structure (topic/focus)
cognitive processing pressures (memory, prediction)

Thus:

word order is not a single parameter, but a multi-factorial equilibrium state.

5 FINAL STATEMENT

Word order is not simply a syntactic configuration.

It is:

the surface resolution of competing pressures between structure, cognition, and discourse organization.

6 Multimodal Syntax: Extending Grammar Beyond Speech

6.1 The Speech-Centric Assumption in Linguistic Theory

Modern syntactic theory, from its earliest structuralist foundations through generative grammar, has largely treated language as:

a system encoded in linear verbal sequences.

Even when acknowledging phonology, morphology, and discourse, the primary object of syntactic analysis remains the spoken or written sentence.

This creates a deep methodological bias:

syntax is assumed to be exclusively verbal and linear.

However, extensive work in interactional linguistics, gesture studies, and cognitive semiotics suggests that this assumption is empirically incomplete.

6.2 The Empirical Reality: Language as a Multimodal System

Human communication is inherently multimodal. Across natural interaction, meaning is distributed across:

speech
gesture
gaze
posture
facial expression
prosody

This has been extensively documented in:

McNeill (1992, 2005): gesture-speech unity hypothesis
Kendon (2004): interactional gesture systems
Goldin-Meadow (2003): homesign systems
Levinson (2006): spatial language and embodied cognition

Key observation:

linguistic meaning is often co-expressed across multiple channels simultaneously.

6.3 Gesture and Syntax: Parallel Structuring Systems

Research in gesture studies shows that gestures are not random supplements to speech, but structurally aligned with it.

Example (narrative speech with gesture):

A speaker describing motion:

“The ball rolled down the hill”

Accompanied by gesture:

hand motion downward following temporal structure of “rolled down”

Key finding:

Gestures often encode:

argument structure

spatial relations

event segmentation

even when not linguistically expressed.

6.4 McNeill’s Gesture–Speech Unity Hypothesis

McNeill proposes that:

gesture and speech originate from a single integrated cognitive process.

This implies:

gestures are not post-lexical additions

but co-expressive components of thought formulation

Implication for syntax:

If gesture participates in encoding meaning structure:

syntactic representation may not be fully contained within verbal linearity.

6.5 Gaze as Grammatical Coordination

In conversational analysis, gaze direction plays a structural role in:

turn-taking
referent tracking
topic shift marking

Example:

speaker gaze toward object → establishes referential anchoring

gaze shift → signals discourse transition

Key insight:

Gaze operates as:

a real-time grammatical coordination mechanism in interaction.

6.6 Prosody as Structural Encoding

Prosodic features include:

intonation
stress
rhythm
pitch contour

These are not merely phonetic embellishments.

Example:

“He didn’t steal the money”

can have different interpretations depending on stress placement:

HE didn’t steal it (someone else did)

he didn’t STEAL it (but maybe borrowed it)

Observation:

Prosody encodes:

focus structure
negation scope
pragmatic contrast

Implication:

prosody functions as a parallel syntactic system encoding structural distinctions.

6.7 Facial Expression and Interactional Grammar

Facial expressions are systematically used in:

question marking
epistemic stance
discourse alignment

Example:

raised eyebrows → interrogative or uncertainty marking

head nod → affirmation or backchanneling

Key insight:

These are not random emotional signals, they are:

structured communicative markers integrated into discourse grammar.

6.8 Homesign Systems: Syntax Without Speech

Research on deaf individuals without exposure to conventional sign languages (Goldin-Meadow) shows that:

consistent gesture systems emerge spontaneously

stable argument structures appear

spatial organization encodes grammatical relations

Example:

Homesign systems often encode:

agent-patient relations

temporal sequencing

event structure

without spoken input.

Critical implication:

grammatical organization does not depend on spoken language.

6.9 Sign Languages: Fully Grammatical Multimodal Systems

Established sign languages (e.g., ASL, BSL) demonstrate:

hierarchical structure
morphological processes
agreement-like spatial indexing
simultaneity of grammatical encoding

Key distinction:

Unlike spoken languages:

multiple syntactic relations can be expressed simultaneously in space.

This challenges:

linearity assumptions in syntax
sequential derivation models

6.10 The Linear Bias Problem in Syntactic Theory

Most syntactic frameworks assume:

structure unfolds linearly
derivations proceed step-by-step
elements are ordered sequentially

However, multimodal data suggests:

linguistic structure may be inherently multi-dimensional rather than linear.

6.11 Toward a Multimodal Syntax Model

We propose:

Syntax is not confined to linear verbal sequences but is distributed across coordinated semiotic channels that jointly encode grammatical relations.

This includes:

verbal structure
gestural structure
prosodic structure
spatial interactional structure

6.12 Structural Integration Hypothesis

Rather than treating modalities as separate layers, we propose:

multimodal elements participate in a single integrated syntactic system, where different modalities encode different dimensions of grammatical structure.

For example:

speech → propositional content

gesture → spatial mapping

prosody → focus and scope

gaze → discourse anchoring

6.13 Counterargument: The Modularity View

A traditional syntactician might argue:

gesture and prosody are performance systems
syntax is purely abstract and amodal
multimodality belongs to pragmatics, not grammar

Response:

This position struggles to explain:

systematic alignment between gesture and syntactic structure
grammatical consistency in sign languages
cross-linguistic integration of prosody into meaning

6.14 Interim Synthesis

Across linguistic systems:

meaning is distributed across multiple channels
structural alignment is consistent across modalities
grammar cannot be fully reduced to linear speech

Thus:

syntax may be better understood as a multimodal coordination system rather than a purely verbal computation.

6 FINAL STATEMENT

Syntax is not confined to words.

It is:

a distributed system of structural coordination across speech, gesture, gaze, prosody, and spatial interaction—each contributing a distinct dimension of grammatical organization.

7 Rethinking Deep Structure: From Hidden Representations to Distributed Architecture

7.1 The Historical Role of Deep Structure

The distinction between deep structure and surface structure emerged as one of the foundational ideas in early generative grammar (Chomsky, 1960s–1970s).

In its classical formulation:

Deep structure encodes basic syntactic relations (argument structure, thematic roles)

Surface structure reflects phonological realization after transformations

Example (simplified):

“John is easy to please”

is derived from a deeper representation in which:

“someone pleases John”

transformations reconfigure argument alignment

This framework established a powerful claim:

meaning is derived from an underlying syntactic configuration that is not directly observable.

7.2 The Theoretical Commitment

Deep structure theory rests on three assumptions:

There exists a universal underlying syntactic representation

Surface variation is derived from transformations

All languages share this underlying level of representation

This yields a strong architectural claim:

syntax is fundamentally hidden and derivational.

7.3 The Empirical Pressure: Cross-Linguistic Diversity

However, typological research reveals persistent tensions:

(A) Polysynthetic languages (e.g., Inuktitut)

argument structure is morphologically fused
no clear separation between syntax and morphology
derivational reconstruction becomes non-trivial

(B) Non-configurational languages (e.g., Warlpiri)

word order is flexible
grammatical relations are encoded via case and discourse
structural hierarchy is not transparently recoverable

(C) Discourse-driven systems (e.g., Japanese)

interpretation relies heavily on context
omission and inference are structurally licensed
overt structure does not fully determine meaning

Key tension:

In many systems, there is no independent evidence for a unique, recoverable “deep” representation.

7.4 The Reconstruction Problem

A central challenge for deep structure theory is:

how do we empirically reconstruct a representation that is, by definition, not observable?

Different diagnostics (movement, binding, reconstruction effects) yield:

inconsistent results across languages
theory-dependent interpretations
varying degrees of applicability

This leads to a methodological concern:

deep structure risks becoming an inferred abstraction rather than a constrained empirical object.

7.5 Minimalism and the Internalization of Deep Structure

The Minimalist Program partially resolves this by eliminating deep structure as a separate level.

Instead:

structure is built derivationally
operations (Merge, Move) construct representations dynamically
LF/PF interfaces replace deep/surface distinction

However, this raises a new issue:

deep structure disappears as a level, but its explanatory role persists implicitly.

For example:

argument structure is still assumed to exist pre-syntactically
thematic roles still guide derivation
hierarchical relations remain foundational

Thus:

deep structure is removed as a label, but retained as a functional assumption.

7.6 Cross-Linguistic Challenge: Distributed Structure

Across diverse languages, we observe that structural relations are often:

split across morphology
distributed across discourse
reconstructed contextually

Example pattern:

In polysynthetic systems:

a single morphological unit encodes what would require an entire clause in English

In discourse-heavy systems:

referents are maintained without overt syntactic realization

Key implication:

structural information is not always centralized in a single syntactic representation.

7.7 Revised Hypothesis: Structure Without a Single Deep Level

We propose a more conservative but powerful reformulation:

what has been called “deep structure” may not correspond to a single universal representational level, but to a distributed set of constraints that jointly determine interpretation.

These constraints may include:

lexical semantics
morphological encoding
discourse continuity
pragmatic inference
structural alignment systems

7.8 From Hierarchical Depth to Distributed Architecture

Instead of a vertical model:

Deep Structure → Transformations → Surface Structure

we propose a distributed model:

Multiple interacting systems jointly produce interpretive structure

These systems include:

(1) Morphosyntactic encoding layer

(2) Argument structure system

(3) Discourse tracking system

(4) Prosodic and multimodal system

No single layer is primary.

7.9 Reinterpreting “Depth”

We redefine “depth” not as a structural level, but as:

the degree of abstraction required to integrate multiple grammatical systems into a coherent interpretation.

Thus:

depth is not a place in structure

but a property of interpretive integration

7.10 Theoretical Consequences for Syntax

If this model is correct:

(1) No single universal deep structure

Languages may not share a single representational substrate.

(2) Derivations are not universal

Some systems may not be derivational in the same sense as others.

(3) Syntax becomes integrative

Syntax is not a generator of structure but a coordinator of multiple subsystems.

7.11 Counterargument: The Necessity of Hidden Structure

A generative linguist might argue:

without deep structure, we cannot explain uniformity across languages
transformations require a base representation
hierarchical dependencies demand underlying structure

Response:

Uniformity may instead arise from:

cognitive constraints
processing pressures
shared communicative goals

not necessarily from a single representational level.

7.12 Interim Synthesis: From Depth to Distribution

Across Chapters 1–6, a converging pattern emerges:

constituency is not universal
categories are not fixed
agreement is not a single system
omission is structurally licensed
word order reflects cognitive pressure
syntax is multimodal

Therefore:

the assumption of a single underlying syntactic depth becomes increasingly difficult to maintain as a universal claim.

7 Excellence Statement

The notion of deep structure, historically central to syntactic theory, may be best reinterpreted not as a universal representational level, but as:

a theoretical compression of multiple interacting systems that jointly produce observable linguistic structure.

Syntax, under this view, is not a projection from a hidden base.

It is:

the emergent coordination of distributed grammatical, cognitive, and discourse-level systems.

CONCLUSION 

Across this manuscript, a consistent reorientation has been developed:

From:

trees → networks

categories → emergent functions

agreement → alignment systems

omission → structured absence

word order → cognitive pressure

deep structure → distributed architecture

speech-only syntax → multimodal grammar

To a unified perspective:

Syntax is not a single structural object but a distributed system of coordination across linguistic, cognitive, and interactional domains.

BIBLIOGRAPHY: Beyond Trees: A New Cartography of Human Syntax

I. FOUNDATIONAL GENERATIVE SYNTAX

Chomsky, N. (1957). Syntactic Structures. Mouton.

Chomsky, N. (1965). Aspects of the Theory of Syntax. MIT Press.

Chomsky, N. (1975). The Logical Structure of Linguistic Theory. Plenum Press.

Chomsky, N. (1981). Lectures on Government and Binding. Foris Publications.

Chomsky, N. (1995). The Minimalist Program. MIT Press.

Chomsky, N. (1964). Current Issues in Linguistic Theory. Mouton.

Halle, M., & Chomsky, N. (1968). The Sound Pattern of English. Harper & Row.

II. SYNTACTIC THEORY & FORMAL MODELS

Baker, M. (2001). The Atoms of Language. Oxford University Press.

Radford, A. (2004). Minimalist Syntax. Cambridge University Press.

Carnie, A. (2013). Syntax: A Generative Introduction. Wiley-Blackwell.

Haegeman, L. (1994). Introduction to Government and Binding Theory. Blackwell.

Kayne, R. (1994). The Antisymmetry of Syntax. MIT Press.

Lasnik, H. (2000). Syntactic Structures Revisited. MIT Press.

III. TYPOLOGY & CROSS-LINGUISTIC DIVERSITY

Greenberg, J. H. (1963). Universals of Language. MIT Press.

Greenberg, J. H. (1974). Language Typology: A Historical and Analytic Overview. De Gruyter.

Comrie, B. (1989). Language Universals and Linguistic Typology. University of Chicago Press.

Croft, W. (2003). Typology and Universals. Cambridge University Press.

Evans, N., & Levinson, S. C. (2009). “The myth of language universals.” Behavioral and Brain Sciences.

Haspelmath, M. (2010). “Comparative concepts and descriptive categories in cross-linguistic studies.”

IV. CONSTRUCTION-BASED & NON-LEXICALIST APPROACHES

Goldberg, A. (1995). Constructions: A Construction Grammar Approach to Argument Structure. University of Chicago Press.

Croft, W. (2001). Radical Construction Grammar. Oxford University Press.

Fillmore, C., Kay, P., & O’Connor, M. (1988). “Regularity and idiomaticity in grammatical constructions.”

Hilpert, M. (2014). Construction Grammar and its Application to English. Edinburgh University Press.

V. MORPHOLOGY–SYNTAX INTERFACE & POLYSYNTHESIS

Mithun, M. (1999). The Languages of Native North America. Cambridge University Press.

Baker, M. (1996). The Polysynthesis Parameter. Oxford University Press.

Hale, K. (1983). “Warlpiri and the grammar of non-configurational languages.”

Inuktitut linguistic studies (various field grammars, Inuit languages corpus tradition).

VI. AGREEMENT, FEATURES & ALIGNMENT SYSTEMS

Corbett, G. (2006). Agreement. Cambridge University Press.

Preminger, O. (2014). Agreement and Its Failures. MIT Press.

Bantu noun class systems literature (Demuth, A., & Givón, T.).

Aikhenvald, A. Y. (2004). Evidentiality. Oxford University Press.

VII. DISCOURSE, PRAGMATICS & OMISSION

Huang, Y. (2000). Anaphora: A Cross-Linguistic Study. Oxford University Press.

Givón, T. (1983). Topic Continuity in Discourse. John Benjamins.

Hinds, J. (1982). English and Japanese: A contrastive study (1978–1982). Annual review of applied linguistics, 3, 78-84.

Prince, E. (1992). “The ZPG letter: Subjects, definiteness, and information status.”

Kageyama, T. (1982). Word formation in Japanese. Lingua, 57(2-4), 215-258.

Kuno, S., & Takami, K. I. (2004). Functional constraints in grammar.

Kuno, S. (1978). Japanese: A Characterization of the Language.

VIII. PROCESSING, COGNITION & WORD ORDER

Frazier, L., & Fodor, J. (1978). “The sausage machine: A new two-stage parsing model.”

Hawkins, J. A. (1994). A Performance Theory of Order and Constituency.

Gibson, E. (1998). “Linguistic complexity: Locality of syntactic dependencies.”

Christiansen, M. H., & Chater, N. (2016). The Now-or-Never Bottleneck.

IX. MULTIMODALITY, GESTURE & INTERACTION

McNeill, D. (1992). Hand and Mind: What Gestures Reveal about Thought.

McNeill, D. (2005). Gesture and Thought.

Kendon, A. (2004). Gesture: Visible Action as Utterance.

Goldin-Meadow, S. (2003). The Resilience of Language.

Levinson, S. C. (2006). “On the human ‘interaction engine’.”

X. SIGN LANGUAGES & SPATIAL SYNTAX

Sandler, W., & Lillo-Martin, D. (2006). Sign Language and Linguistic Universals.

Emmorey, K. (2002). Language, Cognition, and the Brain: Insights from Sign Language Research.

XI. PHILOSOPHICAL & META-THEORETICAL FOUNDATIONS

Saussure, F. de (1916). Cours de linguistique générale.

Lakoff, G., & Johnson, M. (1980). Metaphors We Live By.

Derrida, J. (1967). Of Grammatology (for structural critique context only).

Fodor, J. (1983). The Modularity of Mind.

XII. COMPUTATIONAL & FORMAL LINGUISTICS (SUPPORTING FRAMEWORKS)

Grishman, R., et al. (1994). “Comlex Syntax: Building a Computational Lexicon.”

Boleda, G. (2019). “Distributional semantics and linguistic theory.”