Syntax as Cognitive Computation

 

Syntax as Cognitive Computation: Merge, Structure Dependence, and the Limits of Pedagogical Grammar

A Generative Reassessment of Grammar Instruction in Higher Education

This paper reconsiders the status of grammar in higher education through the lens of generative syntax. It argues that dominant pedagogical models in English Language Teaching (ELT) systematically misrepresent grammar by reducing it to surface-oriented, correctional practices. In contrast, the Minimalist Program (Chomsky 1995) conceptualizes grammar as a computational system grounded in a small set of operations, primarily Merge, which generates hierarchical structure from discrete lexical elements. We demonstrate that key properties of human language- structure dependence, recursion, and displacement, are inaccessible to correction-based pedagogical models because they operate at the level of derivational computation rather than surface performance. The paper formalizes this misalignment and proposes a structural reorientation of grammar instruction toward generative competence rather than prescriptive accuracy.

1. Introduction: The Misobjectification of Grammar

In most institutional settings, “grammar” is treated as a normative system regulating linguistic correctness. This view, however, diverges sharply from the generative tradition, in which grammar is defined as an internalized computational procedure (I-language) that maps lexical arrays to structured expressions (Chomsky 1986, 1995).

We assume the following distinction:

Pedagogical Grammar (PG): a rule-based system for evaluating surface well-formedness

Generative Grammar (GG): a formal system generating hierarchical syntactic objects via recursive computation

This paper argues that PG is not a simplified version of GG but a fundamentally distinct epistemological object.

2. Grammar as I-Language

Within the Minimalist framework, a language L is defined as:

$$L = \langle \text{Lexicon}, \text{Computation} \rangle$$

where computation is minimally constituted by a single operation:

$$\text{Merge}(x, y) \rightarrow \{x, y\}$$

This operation recursively applies to its own output, yielding unbounded hierarchical structures.

Thus, grammar is not a set of descriptive rules but:

a derivational system generating structured representations in the mind.

This interpretation aligns with Chomsky (1995) and subsequent Minimalist refinements (Hornstein, Nunes & Grohmann 2005).

3. Structure Dependence and the Failure of Linear Grammar

A core property of human syntax is structure dependence (Chomsky 1971). Consider the transformation:

(1) Declarative

 The student is reading the book.

(2) Interrogative

 Is the student reading the book?

The grammatical rule is not:

invert the first auxiliary in linear order

but rather:

target the structurally highest T node in the derivation.

This can be represented as:

$${[}_{CP}C{[}_{TP}DPTVP]]$$

Movement applies to structural configuration, not string position.

Pedagogical grammar, however, typically encodes rules as linear heuristics, thereby eliminating the explanatory domain of syntax.

4. Merge and the Derivation of Hierarchy

The fundamental operation Merge constructs hierarchical structure:

External Merge:

$$\text{Merge}(X,Y)\to \{X,Y\}$$

Internal Merge (movement):

$$\text{Merge}(X,\{X,Y\})\to \{X,\{X,Y\}\}$$

This yields displacement effects, including wh-movement and raising constructions.

Example derivation:

(3)

The student who solved the problem understood the theory.

Simplified structure:

$${[}_{DP}\text{The student}{[}_{CP}who{[}_{TP}tsolvedtheproblem]]]{[}_{VP}understoodthetheory]$$

This representation captures hierarchical embedding and traces of movement absent in surface pedagogy.

5. Recursion as a Computational Property

Recursion is defined as the capacity of a system to embed its own output as input:

Merge(x, Merge(y,z))

This yields unbounded generative capacity, consistent with Hauser, Chomsky & Fitch (2002), who argue that recursion is a core property of the Faculty of Language in the Narrow sense (FLN).

Pedagogical simplifications that equate recursion with “complex sentences” eliminate its computational status and reduce it to stylistic variation.

6. Competence, Performance, and Empirical Misalignment

Chomsky’s (1965) distinction is defined as:

• Competence: idealized generative system
• Performance: real-time linguistic output under cognitive constraints

Formally:

$$Performance = f(Competence, Memory, Attention, Noise)$$

Educational systems, however, evaluate only performance approximations, leading to systematic misidentification of syntactic competence with surface correctness.

This produces what may be termed a derivational blind spot: absence of observable error is incorrectly interpreted as evidence of underlying grammatical knowledge.

7. The Poverty of the Stimulus and Pedagogical Limits

The poverty of the stimulus argument (Chomsky 1980) holds that:

$$\text{Input}\to ̸\text{Grammatical Competence}$$

Given the insufficiency of environmental data, grammar must be constrained by innate structural principles (UG).

Pedagogical models based on imitation:

$$\text{Exposure}\to \text{Repetition}\to \text{Acquisition}$$

are therefore theoretically incomplete, as they fail to account for:

• underdetermination of structure
• rapid acquisition of hierarchical constraints
• absence of negative evidence in input

8. The Fallacy of Correctionism

We define correctionism as:

the assumption that grammatical knowledge is equivalent to the elimination of surface errors.

However, correction operates only over:

$$\text{Surface Representation (σ)})$$

not over:

$$\text{Derivational Structure}(D)$$

Thus:

$$Correction\; (\sigma )\; \ne \; Modification(D))$$

Correction can alter output without altering the underlying grammar.

9. Toward a Generative Pedagogy (Theoretical Implication)

A structurally adequate pedagogy would align instruction with generative principles:

• grammar as structure-building system
• instruction as derivational awareness training
• writing as externalization of syntactic computation
• reading as reconstruction of hierarchical representations

This implies a shift from:

• rule enforcement → structure generation
• error correction → derivational diagnostics
• template reproduction → recursive production

10. Conclusion

Grammar, in the generative tradition, is not a pedagogical instrument but a model of cognitive architecture.

Pedagogical grammar, by contrast, is an epiphenomenal system that operates at the surface of linguistic output.

The consequence is a fundamental theoretical asymmetry:

The more grammar is reduced to correctness, the further it is removed from its cognitive reality.

To recover adequacy, grammar instruction must be re-situated within the computational theory of syntax.

Ultimately, the question is not whether learners can produce correct sentences, but whether instruction reflects the fact that:

human language is a recursive, structure-dependent computational system generated by the mind.

References

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

Chomsky, N. (1971). “Deep Structure, Surface Structure, and Semantic Interpretation.”

Chomsky, N. (1980). Rules and Representations. Columbia University Press.

Chomsky, N. (1986). Knowledge of Language. Praeger.

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

Hauser, M. D., Chomsky, N., & Fitch, W. T. (2002). “The Faculty of Language: What Is It, Who Has It, and How Did It Evolve?” Science, 298(5598), 1569–1579.

Hornstein, N., Nunes, J., & Grohmann, K. K. (2005). Understanding Minimalism. Cambridge University Press.