Principles and Parameters (Chomsky)

How the framework works

The puzzle Chomsky's 1979 Pisa Lectures set out to solve was Plato's problem: how do children acquire a system as complex as a natural-language grammar from input that is finite, fragmentary, and free of correction? Pre-1979 generative grammar answered with rule systems — long lists of phrase-structure rules and transformations a child had to induce. By the late 1970s the rule lists had grown unwieldy and language-specific, and the acquisition story was failing.

The Principles and Parameters framework reframed the problem. Suppose Universal Grammar (UG) is a fixed core of principles — laws of grammatical computation — plus a small finite set of parameters — binary switches the child must set from input. Then acquisition is not rule-learning but parameter-fixing. A handful of switches, once set, generate a complete grammar.

The architecture has three parts:

1. Principles are invariant. Structure-dependence says operations target hierarchical units, not linear positions. Subjacency bounds movement across two cyclic nodes at most. The Binding Theory (Chomsky 1981) constrains the distribution of anaphors (himself), pronouns (him), and referring expressions (John). The Empty Category Principle licenses traces of movement.
2. Parameters vary across languages. The head parameter sets phrase-internal order. The null-subject (pro-drop) parameter governs whether subject pronouns may be dropped. The wh-movement parameter distinguishes English-type fronting from Mandarin-type in-situ. The configurationality parameter distinguishes English from non-configurational Warlpiri.
3. Lexicon holds idiosyncratic word-specific information — argument structure, irregular morphology, exceptions.

The cross-linguistic payoff is that a parameter setting cascades. Once a child fixes head-initial, every projection — verb-object, preposition-noun, noun-complement — falls into line. The economy of the explanation drove the framework's adoption.

Worked example: setting the head parameter

Greenberg's typological surveys (1963) revealed that languages tend to be either consistently head-initial (like English) or consistently head-final (like Japanese). The Principles and Parameters answer is that a single binary switch determines the orientation of all heads.

Consider the input a child receives:

• English child hears: "drink milk", "on the chair", "kicked the ball", "happy with the cake". Verb before object, preposition before noun, adjective followed by complement.
• Japanese child hears: miruku-o nomu (milk-OBJ drink), isu-no ue-ni (chair-GEN top-on), booru-o ketta (ball-OBJ kicked), keeki-de ureshii (cake-with happy). Object before verb, noun before postposition, complement before adjective.

A handful of utterances is enough to fix the parameter. In Lila Gleitman and colleagues' acquisition work, English-learning toddlers respect verb-object order from the earliest two-word stage; Japanese-learning toddlers respect object-verb order from the same stage. The cascade across categories is in place essentially from first multi-word combinations.

Subsequent typological work has complicated the simple binary picture. Mixed-headedness languages exist (German is head-final in VP but head-initial in PP). Mark Baker (The Atoms of Language, 2001) recast this as a hierarchy of parameters, with high-level macroparameters constraining clusters of microparameters. The head parameter remains the textbook illustration.

Worked example: English do-support as a parametric consequence

English negation and questions look bizarre cross-linguistically. "John does not eat sushi" — why the auxiliary does? Why not "John eats not sushi", parallel to French Jean ne mange pas de sushi?

The Principles and Parameters answer threads several pieces. In Romance and Old English, the lexical verb raises to a higher functional head (T or Infl) to combine with tense and negation. In Modern English, that verb-raising has been lost — only auxiliaries (be, have, modals) raise. When negation or question fronting demands an element in T, English inserts a dummy do. This is Pollock's (1989) classic analysis in Linguistic Inquiry.

The parametric switch is not idiosyncratic English oddity — it is a single setting (loss of V-to-T movement) with multiple visible consequences: do-support, subject-auxiliary inversion patterns, the placement of adverbs ("often eats" vs French "mange souvent"), and the position of negation. One parametric change in late Middle English explains all four phenomena. This is the kind of bundling the framework was built to capture.

Principles and Parameters vs other syntactic frameworks

	Principles & Parameters	Minimalism	LFG	HPSG	Construction Grammar	Systemic Functional
Originators	Chomsky 1981	Chomsky 1995	Bresnan, Kaplan 1982	Pollard, Sag 1994	Goldberg 1995, 2006	Halliday 1985
Universal Grammar	Rich, innate	Slimmer, mostly Merge	Yes, modular	Yes, schemas	Rejected — emergentist	Rejected — social
Cross-linguistic variation	Parameter setting	Lexical features	Lexical entries	Type hierarchy	Construction inventory	Register / context
Movement	Move-α	Internal Merge	Functional structure (no movement)	Feature percolation	None	None
Acquisition story	Switch setting	Feature acquisition	Lexical learning	Schema abstraction	Item-based learning	Functional learning
Computational tractability	Limited	Limited	Strong	Strong	Strong	Moderate
Major texts	GB Lectures (1981)	Minimalist Program (1995)	Bresnan 1982, 2001	Pollard & Sag 1994	Goldberg 1995	IFG (1985, 2014)

Parameter setting and the acquisition argument

The framework's strongest selling point was its solution to the acquisition problem. Children master grammar by age four or five from input that, by Chomsky's argument, is too impoverished to support induction of arbitrary rule systems — the Poverty of the Stimulus argument. Switch-setting circumvents the difficulty: there is nothing to induce, only a finite set of switches to fix.

Steven Pinker's The Language Instinct (1994) popularized the case. Pinker's argument: complex inferences (auxiliary inversion, anaphora, constraints on movement) are mastered without explicit instruction, despite input that does not unambiguously cue the right rule. A built-in parameter space explains the speed and uniformity of acquisition.

Charles Yang's variational model (Knowledge and Learning in Natural Language, 2002) made the acquisition mechanism formal. The child entertains multiple candidate grammars; each parses incoming input with some probability of success; weights shift toward grammars that fit. Parameters with strong signal (head direction) converge in months; parameters with weak signal (residual V2, German) converge over years. Yang's model fits acquisition data closely and motivates a cleaner version of switch-setting than Chomsky's original triggering-experience proposal.

Counterarguments and the construction-grammar challenge

Adele Goldberg's Constructions: A Construction Grammar Approach to Argument Structure (1995) launched the most serious modern alternative. Goldberg argued grammar is not principles-plus-parameters but an inventory of constructions — pairings of form with meaning at every level from morpheme to clause. The ditransitive construction Subj V Obj1 Obj2 contributes its own meaning ("X causes Y to receive Z") independent of the verb. Argument structure is constructional, not lexical.

Functionalists (Givón, Bybee, Croft) argue grammatical regularities emerge from frequency and use, not innate switches. Statistical learning over corpora reproduces many grammatical patterns without postulating Universal Grammar.

Typologists raise empirical worries. Rizzi's classic pro-drop cluster — null subjects, free inversion, that-trace, expletive drop — fails to bundle in many languages. Greek allows null subjects without the inversion pattern. Brazilian Portuguese is losing pro-drop while keeping inversion. Evans and Levinson's The Myth of Language Universals (Behavioral and Brain Sciences, 2009) marshalled cross-linguistic data to argue that the parameter space is too small to capture observed diversity.

Defenders respond that parameters were never meant as exhaustive descriptive devices. They are heuristics organizing variation. The Minimalist reformulation — variation lives in lexical features on functional heads — addresses the typological challenges by giving up the macro-cluster claim while keeping the principles-plus-variation architecture.

Variants and successors

• Government and Binding (1981–1995). The classic articulation. Modules: X-bar theory, theta-theory, case theory, government, binding theory, bounding theory, control theory, ECP. Rich, intricate, and the basis of most graduate syntax courses through the 1990s.
• Minimalist Program (1995–). Chomsky's The Minimalist Program dismantled GB modules, replacing them with Merge plus interface conditions. Parameters are recast as features on functional heads. The architecture is leaner; the empirical coverage debated.
• Cartographic syntax (Cinque, Rizzi 1990s–2000s). Builds elaborate hierarchies of functional projections (TopP, FocP, FinP) with fixed cross-linguistic order. Parameters select which projections are pronounced.
• Macroparameters (Baker 2001). The Polysynthesis Parameter, the Head-Directionality Parameter, and others organized into hierarchies. Makes clear empirical predictions about clusters.
• Microparametric variation (Kayne 2000–). Variation lives in tiny lexical-feature differences between dialects. Closer Italian dialects differ in dozens of microparameters.

Common pitfalls in interpreting the framework

• Treating parameters as literal psychological switches. Chomsky's original talk of "switches" is heuristic; the cognitive reality is open. Parameters are research tools that organize cross-linguistic generalizations.
• Expecting all classical clusters to bundle. Rizzi's pro-drop cluster fails empirically in several languages. Modern work treats clustering as gradient, not categorical.
• Conflating Government and Binding with Minimalism. They are stages of one research program but differ substantively in machinery. GB has government, traces, indices, and modules; Minimalism reduces to Merge plus features.
• Reading "Universal Grammar" as identical sentences cross-linguistically. UG is the architecture of grammar — what kinds of rules are possible, how acquisition proceeds — not a set of universal sentences. Surface patterns vary; the system that generates them is constrained.
• Assuming parameters explain everything in typology. Word-frequency effects, areal contact, sociolinguistic conditioning, and semantic shifts shape language change in ways orthogonal to parametric architecture.
• Mistaking critique of clusters for refutation of UG. Even if classical parameter clusters fail, the architectural claim — invariant computation plus finite variation — survives multiple reformulations.

Legacy and current status

Principles and Parameters reorganized syntactic theory in the 1980s and remains the dominant framework in generative linguistics, in its Minimalist successor form. Comparative syntactic work in Romance, Germanic, and Bantu languages continues to be parametric in spirit. The framework's acquisition story shaped psycholinguistics for two decades. Its typological program, with Mark Baker's macroparameters and Roberts and Holmberg's parameter hierarchies, is active research.

The competitor frameworks — LFG, HPSG, Construction Grammar, Cognitive Grammar, Systemic Functional Grammar — have grown alongside and offer real alternatives, but no single replacement has achieved the explanatory unification of acquisition, typology, and theoretical syntax that Principles and Parameters offered. The 2009 Evans–Levinson critique sharpened the empirical questions; the 2010s saw nuanced parameter-hierarchy responses. The debate is unresolved and active.