A Revolution in the Study of Language: From Corpora to Competence
Introduction
We begin with Chomsky’s opening remarks, which define the object of inquiry and state the fundamental problem that the entire work sets out to solve. Chomsky establishes that a language is a finite or infinite set of sentences, each finite in length and constructed out of a finite set of elements. The central task of linguistics, he argues, is to provide a grammar that generates all and only the grammatical sentences of a language, while excluding those that are ungrammatical.
“From now on I will consider a language to be a set (finite or infinite) of sentences, each finite in length, and constructed out of a finite set of elements.”
The immediate difficulty is that a speaker of a language can produce and understand an indefinitely large number of sentences, many of which have never been uttered before. A traditional grammar that merely lists sentences or patterns cannot account for this creative aspect of language use. Instead, Chomsky proposes that a grammar must be a device of some sort — a set of rules — that generates the sentences of a language. The term “generate” is used here in a precise sense: the grammar is to produce a structural description for every sentence, showing how it is built up from constituent parts. The work will develop formal models of such generative grammars, beginning with the simplest and moving toward those capable of capturing the full complexity of natural language.
This matters because Chomsky is redefining the very nature of linguistic inquiry. Instead of a science that classifies observed data (as in structuralist linguistics of the time), he is proposing a theoretical enterprise: the linguist must construct a hypothesis — the grammar — that explains the speaker’s ability to produce and understand novel sentences. The grammar is not a catalogue of utterances; it is a theory of the speaker’s knowledge.
Chapter 1: An Elementary Linguistic Theory
The first chapter lays out the basic concepts and exposes the inadequacy of the simplest possible model of grammar: a finite-state Markov process. Chomsky imagines a machine that produces sentences by moving from one state to another, each transition emitting a word. The machine is limited to a finite number of states and can only construct sentences by “left-to-right” dependencies — each word choice depends only on the current state and, through the state, on the limited history of the preceding words.
Chomsky then demonstrates that this finite-state model is fundamentally incapable of generating all grammatical sentences of English. He offers a simple proof using self-embedding structures. Consider sentences of the form: `If S₁ then S₂`, where S₁ and S₂ are themselves sentences. English allows an indefinite number of recursive embeddings: `If it rains, then if it snows, then we will stay home` — and this can be nested further. A finite-state machine, Chomsky argues, cannot handle such discontinuous dependencies because the relationship between `if` and `then` can be arbitrarily far apart, with an unbounded number of intervening `if…then` pairs. The number of states a finite-state machine would need grows without bound as the depth of embedding increases, which is impossible for a machine with a finite number of states.
The conclusion is sharp: English is not a finite-state language. Any grammar that can account for the full range of grammatical sentences must be more powerful than a finite-state device. Chomsky identifies that natural languages require a grammar with phrase structure rules, which can build hierarchical structures rather than simple left-to-right chains. A phrase structure grammar allows a constituent to be rewritten as a string of smaller constituents (e.g., `S → NP + VP`, `VP → V + NP`), and crucially, it permits recursion: a rule like `S → if S then S` embeds one `S` within another, creating an infinite generative capacity from a finite set of rules.
This chapter establishes the central methodological principle of the work: the linguist must choose a formal model of grammar that is powerful enough to describe natural languages but not so powerful that it overgenerates (i.e., produces ungrammatical sequences) or fails to capture relevant generalizations. The finite-state model is rejected as too weak; phrase structure grammars are introduced as the next candidate.
Chapter 2: The Independence of Grammar
The second chapter addresses the relationship between grammar and meaning. Chomsky makes a controversial argument: that grammaticalness is distinct from meaningfulness. A sentence can be perfectly grammatical and utterly nonsensical, and a meaningful sequence can be ungrammatical. He introduces a classic pair:
“Colorless green ideas sleep furiously.”
This sentence is semantically anomalous — green things cannot be colorless, ideas cannot literally sleep, and sleeping cannot be done furiously. Yet any speaker of English recognizes it immediately as a grammatical sentence of English, with a clear subject-verb-adverb structure. By contrast, the reversed sequence *”Furiously sleep ideas green colorless”* is both semantically and grammatically deviant. The difference, Chomsky argues, has nothing to do with meaning or frequency of occurrence; it is a purely formal property of the sequence’s structure.
Chomsky extends this argument to a broader point: a grammar cannot be based on statistical approximation to the observed sentences of a language. He shows that if we construct sentences by picking words with probabilities based on how often they appear after one another in real texts, we will never arrive at the grammatical/ungrammatical distinction. A sequence like *”Colorless green ideas sleep furiously”* would have near-zero probability in any corpus, yet it is perfectly grammatical. The grammar must be based on formal rules, not on the statistics of usage.
This chapter matters because it liberates grammatical theory from dependence on semantics and probability. For Chomsky, the grammar is an autonomous system — it can be studied in its own right, without reference to meaning or context. The goal of syntactic theory is to characterize the formal patterns that define grammaticalness, independent of what sentences “mean.” This separation is foundational for the rest of the work, which will develop purely formal syntactic models.
Chapter 3: An Elementary Linguistic Theory (Continued)
The third chapter returns to the task of building a grammar for a fragment of English. Chomsky now presents a phrase structure grammar in concrete detail, using a small set of rewrite rules (also called immediate constituent rules). The grammar begins with the initial symbol `Sentence` and successively rewrites it into strings of symbols until only actual words remain. A sample fragment:
`Sentence → NP + VP` `NP → T + N` (where `T` is a determiner like `the`, `N` is a noun) `VP → Verb + NP` `Verb → Aux + V` (where `Aux` includes tense and modals) `T → the, a` `N → man, ball, book, etc.` `V → hit, took, read, etc.`
Chomsky shows how this set of rules generates sentences like `the man hit the ball`. The derivation proceeds stepwise: `Sentence` → `NP + VP` → `T + N + VP` → `the + N + VP` → `the + man + VP` → `the + man + Verb + NP` → `the + man + Aux + V + NP` → `the + man + Past + V + NP` → `the + man + Past + hit + NP` → `the + man + Past + hit + T + N` → `the + man + Past + hit + the + N` → `the + man + Past + hit + the + ball`. The final string `the man past hit the ball` is then converted (by a morphophonemic rule) into `the man hit the ball`.
This simple grammar, Chomsky notes, is insufficient. It cannot generate passive sentences like `the ball was hit by the man`, or questions like `did the man hit the ball?`, or sentences with negation, or complex verb phrases with modals and perfective auxiliaries. But the purpose is not to give a complete grammar of English; it is to demonstrate the formal mechanism of phrase structure. A phrase structure grammar provides a constituent analysis of each sentence — it assigns a labeled tree structure showing how smaller parts combine into larger units. The tree for `the man hit the ball` would show `NP` dominating `the` and `man`, `VP` dominating `Verb` and `NP`, and so forth.
Chomsky then introduces a critical limitation of phrase structure grammars: they cannot easily handle certain kinds of relationships between sentences. Consider the active-passive pair: `the man hit the ball` and `the ball was hit by the man`. These two sentences are clearly related: they contain the same nouns and verb, but the grammatical roles (subject, object) are swapped, and the verb form changes. A phrase structure grammar can generate both sentences, but it would do so through completely separate derivations, with no way to express their systematic relationship. This is a key motivation for the more powerful model Chomsky will develop in subsequent chapters.
The chapter also introduces the concept of transformational rules. These are operations that map one syntactic structure (a phrase marker) into another phrase marker. For example, a passive transformation could take the phrase marker of an active sentence and rearrange its constituents to produce the passive counterpart. Transformations are not part of the phrase structure grammar itself; they are an additional layer that operates after the phrase structure rules have generated a deep structure.
This is a pivotal moment in the book. Chomsky is arguing that a grammar must have two distinct levels: a deep structure (generated by phrase structure rules) that captures the underlying grammatical relations (subject, object, etc.), and a surface structure (produced by transformations) that corresponds to the actual form of the sentence. The active-passive relationship, for instance, can now be captured: both sentences share the same deep structure, but the passive transformation applies to produce the surface structure of the passive. This two-level conception is the core of what will become transformational-generative grammar.
Chapter 5: Finite State Languages (Continued from Part 1)
We are now in the second half of Chapter 5, where Chomsky has established that natural languages cannot be described as finite state languages and turns to a mathematical characterization of what such languages actually are.
Chomsky presents the concept of a finite state Markov process more formally. A finite state language is produced by a machine that can be in any one of a finite number of states, and each time it moves from one state to another it produces a symbol. The transition from one state to another is determined by a fixed set of probabilities. A sentence in the language is a sequence of symbols produced by a path from the initial state to a final state. The language itself is the set of all such sequences. Chomsky then introduces the key concept of dependency sets: a finite state language can only generate dependencies that do not have a “nesting” or “center-embedding” structure beyond a fixed depth. This is because the machine has no memory of which choices it made earlier—it only knows its current state.
The author then provides a rigorous mathematical proof that English cannot be a finite state language. He constructs a specific type of sentence structure in English—self-embedded constructions—and shows that they generate patterns that no finite state machine can produce. A self-embedded sentence is one where a phrase is inserted inside another phrase of the same type, such as “If S₁ then S₂” where S₁ itself contains another “if…then” construction. The crucial property is that in English, the number of “if”s must match the number of “then”s, and they must be paired in order. In a sentence like “If if it rains then the game will be cancelled then we will go home,” the first “if” must pair with the last “then”, the second “if” with the first “then”, and so on. This non-finite-state character arises because the machine would need to remember an unbounded number of previous choices to ensure proper pairing.
“Clearly, English is not a finite-state language, since it permits nested dependencies of arbitrary depth.”
This proof is significant because it establishes a fundamental limitation on what kind of mathematical description can capture natural language. A finite state grammar can only handle languages where dependencies are limited to a fixed distance. But English allows unbounded nesting, meaning there is no upper limit on how deeply one clause can be embedded within another. Therefore, any adequate grammar of English must be more powerful than a finite state machine. Chomsky is not just criticizing a particular theory; he is demonstrating a mathematical impossibility.
Chapter 6: On the Goals of Linguistic Theory
We now move to Chapter 6, where Chomsky steps back from specific grammatical mechanisms to ask what linguistic theory itself should aim to achieve. This chapter establishes the criteria by which different grammars and theories can be evaluated.
Chomsky introduces three levels of adequacy for linguistic theory. The first is observational adequacy: a grammar that achieves this level correctly describes the data—that is, it distinguishes grammatical from ungrammatical sentences in a given corpus. The second is descriptive adequacy: a grammar at this level not only describes the data but also captures the linguistic intuition of the native speaker, including judgments about ambiguity, paraphrase relations, and structural relationships between sentences. The third and highest level is explanatory adequacy: a linguistic theory that achieves this provides a principled basis for selecting among competing grammars that are observationally and descriptively adequate. It explains how a child, given limited and imperfect data, can acquire the correct grammar.
The author argues that most structuralist grammars of the period achieved at best observational adequacy. They could catalogue the sentences of a language, but they could not explain why certain sentences are related or why certain structures are impossible. A theory that achieves explanatory adequacy must, Chomsky argues, impose strong constraints on what counts as a possible grammar. It must provide a evaluation measure that, when given two grammars compatible with the data, selects the one that best captures the speaker’s intuitions.
“A grammar of a language purports to be a description of the ideal speaker-hearer’s intrinsic competence.”
This is a crucial distinction. Chomsky is drawing a line between competence—the tacit knowledge a speaker has of their language—and performance—the actual use of language in concrete situations. Performance is subject to memory limitations, distractions, and errors, but competence is the underlying system of rules. A grammar is a model of competence, not a record of performance. The goal of linguistic theory is therefore to characterize the nature of this competence, which is universal across human languages. This shift from describing data to explaining the cognitive basis of language is the central move of the chapter.
Chapter 7: Phrase Structure
We now enter Chapter 7, where Chomsky begins a systematic exposition of phrase structure grammar as a formal system. This chapter presents the precise mechanisms of phrase structure and shows both their power and their limitations.
Chomsky defines a phrase structure grammar as a set of rewrite rules that replace one symbol with a sequence of other symbols. Each rule has the form A → Z, where A is a single non-terminal symbol (like S for Sentence, NP for Noun Phrase, VP for Verb Phrase) and Z is a non-empty string of symbols. The grammar begins with the initial symbol S and applies rules until only terminal symbols—the actual words—remain. For example, the rules S → NP + VP, NP → T + N (where T is an article and N is a noun), and VP → V + NP can generate the sequence “the man hit the ball” by successive applications. The set of all sequences generated by these rules is the terminal language.
Chomsky then introduces the crucial notion of derivation and the phrase marker it produces. A derivation is a sequence of strings, starting with S and ending with a terminal string, where each step replaces one symbol according to a rule. The phrase marker is a labeled tree or bracketing that shows the hierarchical structure of the sentence at each point in the derivation. For example, in the sentence “the man hit the ball,” the phrase marker shows that “the man” is a constituent (NP), “hit the ball” is a constituent (VP), and “the ball” is a sub-constituent (NP) within the VP. This hierarchical structure is essential for capturing grammatical relations.
“A grammar is a finite set of rules that generates an infinite set of sentences.”
This chapter makes clear that phrase structure grammars are much more powerful than finite state grammars. They can generate languages with nested dependencies, like “if…then” constructions, because the rewrite rules can be applied recursively. A rule like S → if S then S allows a sentence to contain another sentence within it. However, Chomsky then begins to show the limitations: phrase structure grammars, as defined here, cannot handle certain kinds of discontinuous constituents or movement phenomena. For instance, in a passive sentence like “the ball was hit by the man,” the object “the ball” has been moved to subject position—a transformation that cannot be captured by simple rewrite rules alone. This sets the stage for the introduction of transformations in the next chapter.
Chapter 8: The Explanatory Power of Linguistic Theory
Chapter 8 demonstrates how the theoretical framework developed so far can be applied to solve specific problems in English grammar. Chomsky shows that a phrase structure grammar alone is insufficient and that transformations are necessary to achieve explanatory adequacy.
Chomsky begins with the problem of auxiliary verbs in English. A sentence like “John has been reading” contains three auxiliary elements: have + -en, be + -ing, and the main verb. The order of these auxiliaries is fixed: modal (like “can”) comes before perfect (“have + -en”), which comes before progressive (“be + -ing”), which comes before the main verb. A phrase structure grammar can generate these sequences by rules like Aux → C (have + -en) (be + -ing), where C is the tense marker. But this rule must also account for negation and question formation. In a question like “Has John been reading?”, the first auxiliary element moves to the front of the sentence. In a negative like “John has not been reading,” the negation marker “not” appears after the first auxiliary.
Chomsky then introduces the transformational rule for forming yes-no questions: Tq: NP — Aux — X → Aux — NP — X. This rule takes a declarative sentence with a subject (NP), an auxiliary (Aux), and a remainder (X), and moves the auxiliary to the front. The deep structure—the phrase marker before the transformation—is the same for both the declarative and the question. The difference is produced by applying or not applying the transformation. This is a key demonstration that transformations capture systematic relationships between sentence types.
“The phrase structure rules generate a deep structure; the transformational rules map this deep structure onto a surface structure; and the phonological rules map the surface structure onto a phonetic representation.”
The chapter then extends this analysis to more complex phenomena. Chomsky shows how the negative transformation (Tnot) and the passive transformation (Tpass) interact with the auxiliary system. The passive transformation, for example, moves the object into subject position, inserts the passive auxiliary “be + -en,” and optionally adds “by + NP” at the end. Critically, these transformations must apply in a fixed order: the auxiliary rules must apply before the rules that form questions or negatives, because the transformation needs to know which auxiliary element to move. This ordering is a fundamental property of transformational grammar.
Chapter 9: Transformational Analysis of Some Sentence Types
We are now in Chapter 9, where Chomsky provides a detailed transformational analysis of several specific constructions in English, demonstrating the explanatory power of the framework.
Chomsky begins with declarative sentences and their relationship to imperatives. An imperative like “Close the door!” is related to the declarative “You close the door.” The deep structure of the imperative includes a subject “you” and a future tense marker. The imperative transformation deletes the subject and also deletes the future marker, producing the imperative form. This analysis explains why imperatives are understood as having “you” as the understood subject, and why they cannot be negated in the same way as declaratives without auxiliary verbs.
The author then examines passive sentences in more detail. He shows that the passive transformation must be stated in terms of the deep structure relations, not the surface structure order. The deep structure of a passive sentence like “The ball was hit by John” is the same as that of the active “John hit the ball”: it has the subject “John,” the verb “hit,” and the object “the ball.” The passive transformation reorders these elements and changes the verb form. Crucially, this explains why only transitive verbs (verbs that take an object) can appear in the passive: the passive transformation requires an object in the deep structure to move into subject position.
“The passive transformation applies to any string of the form NP — Aux — V — NP — X.”
Chomsky then turns to coordinate structures—sentences joined by “and” or “but”—and shows that conjunction reduction is a transformation that deletes repeated elements. For example, the sentence “John sang and John danced” is reduced to “John sang and danced” by deleting the second occurrence of “John.” This transformation explains why certain conjunctions are ambiguous: “old men and women” can have two different deep structures—one where “old” modifies both “men” and “women” (from “old men and old women”) and one where it modifies only “men” (from “old men and women”). The ambiguity is resolved at the level of deep structure, before the conjunction reduction transformation applies.
The chapter concludes with an analysis of relative clauses, such as “the man who came to dinner.” Chomsky argues that these are derived from embedded sentences via transformations: “the man who came to dinner” comes from “the man — the man came to dinner,” where the second “the man” is replaced by “who” and moved to the front. This analysis accounts for the fact that relative clauses have the same internal structure as declarative sentences, and it explains why the relative pronoun must match the head noun in certain respects. These analyses collectively demonstrate that transformations provide a unified account of diverse sentence types, revealing underlying regularities that phrase structure alone cannot capture.
Chapter 9 (Continued): Transformational Analysis of Some Sentence Types
LOCATE: Continuing within Chapter 9, Chomsky moves from the analysis of interrogatives, negatives, and passives to the treatment of verb phrases containing auxiliaries, and then to more complex constructions involving nominalizations and gerunds.
PRESENT: Chomsky examines the placement of the verbal affix in sentences with auxiliary verbs. He observes that in declarative sentences, the main verb appears with a tense marker: “John came” or “John has come.” In the passive, the auxiliary “be” takes the tense, and the main verb appears as a past participle: “John was seen.” Chomsky proposes a systematic rule: the affix (tense marker, participial ending) is always attached to the immediately following verbal element. In “John came,” the past tense affix attaches to “come,” yielding “came.” In “John has come,” the affix attaches to “have,” producing “has,” and the main verb “come” appears in its participial form. In “John is coming,” the affix attaches to “be,” giving “is,” and the main verb appears with the progressive suffix “-ing.” This rule, which Chomsky formulates as Affix Hopping, ensures that the verb morphology is always correctly generated from the underlying phrase structure. The transformation moves the affix to the right onto the first verb or auxiliary in the string, applying after the phrase structure rules have generated the base forms.
Chomsky then turns to nominalizations, such as “John’s proving the theorem” or “the proof of the theorem by John.” He argues that these are not derived directly from the phrase structure rules but are instead related to underlying sentences via transformations. For instance, “John’s proving the theorem” is transformationally derived from the sentence “John proves the theorem.” The transformation nominally the verb (“proves” to “proving”) and inserts a possessive marker on the subject, all while preserving the basic argument structure (subject-verb-object). Chomsky notes that this analysis explains why nominalizations have the same internal grammatical relations as the corresponding sentences: in both “John proves the theorem” and “John’s proving the theorem,” “proves” (or “proving”) relates to “the theorem” as verb to object, and to “John” as subject. Without the transformational derivation, a phrase structure grammar would need separate rules for each nominalization pattern, missing the systematic relationship.
Chomsky extends this analysis to gerundive nominals, such as “the flying of the plane” versus “the flying plane.” He notes a crucial ambiguity: “the flying of the plane” can refer to the act of flying (a gerund) or to a specific plane that flies (a participle). The gerundive reading is nominal in nature and preserves the verbal relations (the plane is the object of flying), while the participial reading is adjectival, modifying “plane.” The transformational account captures this by deriving the gerund from a sentence (“someone flies the plane”) and the participle from a reduced relative clause (“the plane that flies”). This distinction is not easily captured by phrase structure rules alone, which would assign the same surface structure to both. Chomsky writes:
“The transformational approach seems essential for explaining the systematic relations between active and passive sentences, between declarative and interrogative sentences, and between sentences and nominalizations.”
CLARIFY: The Affix Hopping rule is a masterful piece of analysis: it shows that the apparent complexity of English verb morphology (irregular past tenses, participial forms, auxiliary placement) can be reduced to a single, simple transformation operating on a uniform underlying structure. This systematization is a powerful demonstration of how transformations explain surface diversity from deep uniformity. The treatment of nominalizations and gerunds expands the reach of the theory beyond full sentences to phrases, showing that the same underlying sentence structures can yield many surface forms through different transformations. Chomsky thus establishes that transformations are not just optional frills but are central to the generative power of the grammar, connecting sentences to their related nominal forms and accounting for structural ambiguities that phrase structure cannot resolve.
Chapter 10: Some Transformations in English
LOCATE: In this chapter, Chomsky offers a more formal and systematic treatment of the transformations that operate in English, organizing them into a coherent system and discussing their ordering, conditions, and interactions.
PRESENT: Chomsky begins by distinguishing between obligatory transformations and optional transformations. Obligatory transformations must apply to every derivation; if they do not, an ungrammatical string results. Optional transformations are chosen by the speaker to produce different sentence types. For example, the Affix Hopping transformation is obligatory: the underlying form “John Past come” must have the affix moved to the verb, yielding “John came”; without it, we get the ungrammatical “John Past come.” Similarly, the Passive transformation is optional: it applies only when the speaker chooses to produce a passive sentence. Chomsky provides a list of the major obligatory transformations in English, including: Subject-Auxiliary Inversion for questions, Do-Support (which inserts “do” when an affix cannot hop onto a verb, as in “Does John come?” derived from “John Past come” where “do” carries the tense), and Agent-Delction (which removes the “by” phrase from a passive, producing “John was seen” from “John was seen by someone”).
Chomsky then turns to the ordering of transformations. He demonstrates that transformations must apply in a fixed sequence; applying them in the wrong order yields incorrect results. For example, the Passive transformation must apply before Subject-Auxiliary Inversion in order to produce a passive question like “Was John seen by Mary?” If Subject-Auxiliary Inversion applied first to the active “John saw Mary,” we would get “Did John see Mary?” and then the passive could not produce the correct form. Similarly, Affix Hopping must apply after certain other transformations that insert or move verbal elements, such as the passive “be” and the progressive “be” (which is inserted by a transformation from an underlying “ing” suffix). Chomsky formalizes this by stating that transformations are linearly ordered in a transformational cycle: the output of one transformation becomes the input to the next, and the order cannot be permuted.
Chomsky also discusses conditions on transformations, which restrict their application. The most important of these is the A-over-A Condition, which states that if a transformation applies to a constituent of type A (e.g., a noun phrase), it must apply to the highest constituent of type A on the relevant structural level, not to one embedded within it. This prevents, for example, moving a noun phrase out of a larger noun phrase: we cannot form “Who did you see [a picture of ___]?” by extracting “who” from inside the noun phrase “a picture of who” — only the entire noun phrase “a picture of whom” can be moved, giving “A picture of whom did you see?” (or, with the transformation applied to the whole NP, “Whom did you see a picture of?” — though this latter case involves more complex conditions). The A-over-A Condition ensures that transformations operate in a constrained way, preventing overgeneration and preserving the structure of the sentence.
CLARIFY: This chapter marks a significant advance in formalization: Chomsky is no longer simply illustrating transformations but is specifying their properties as a formal system. The distinction between obligatory and optional transformations clarifies the relationship between underlying structure and surface form: the base generates the deep structure, obligatory transformations ensure grammaticality, and optional transformations produce stylistic and grammatical variation. The ordering of transformations and the A-over-A Condition introduce the idea that transformations are not arbitrary but are subject to universal constraints — a concept that will become central to later work in generative grammar. Chomsky is building a theory that is not just descriptive but explanatory, limiting the power of the grammar to what is linguistically possible.
Chapter 11: The Explanation of Syntactic Structure
LOCATE: Chomsky now steps back from specific analyses to address a fundamental question: what does it mean to explain a syntactic structure, and how does the transformational theory achieve this explanation?
PRESENT: Chomsky argues that an adequate linguistic theory must do more than describe the data — it must provide explanations for why certain patterns occur and others do not. He contrasts the transformational approach with the taxonomic approach of structural linguistics, which merely classifies surface patterns. For example, the traditional structuralist analysis of passive sentences would list the passive as a separate structure, with its own immediate constituent analysis, unrelated to the active. Chomsky argues that this is merely descriptive: it tells us what the passive looks like but not why it exists or how it relates to other sentence types. The transformational theory, by deriving the passive from the active via a rule, explains why the passive has the same semantic relations (agent, object, verb) as the active, why it uses the same verb, and why its form is systematic. This, Chomsky contends, is genuine explanation.
Chomsky introduces a key concept: levels of adequacy for grammatical descriptions. The lowest level is observational adequacy: the grammar correctly lists the observed sentences of a language. The next is descriptive adequacy: the grammar correctly assigns structural descriptions to sentences, capturing the native speaker’s intuitions about ambiguity, phrase structure, and relations between sentence types. The highest level is explanatory adequacy: the grammar not only describes the data but does so in a way that reveals general principles that hold across languages, and that can be learned by children. The transformational theory, Chomsky argues, achieves explanatory adequacy because its concepts — phrase structure, transformations, recursive rules — are universal and are not learned from the surface data alone. Children do not hear explicit instruction about transformations or deep structure; they must innately possess these concepts to acquire language so rapidly and uniformly. Chomsky writes:
“The grammar of a language must be a device that generates all of the grammatical sequences of the language and none of the ungrammatical ones.”
Chomsky then discusses the role of simplicity in choosing between competing grammars. He argues that a simpler grammar — one that uses fewer rules or more general principles — is preferred because it is more likely to reflect the underlying psychological reality of language. For example, the transformational grammar of English is simpler than a phrase structure grammar that attempts to handle all sentences with its own rules, because the transformational grammar captures generalizations (like the relation between active and passive) with a single rule rather than with separate sets of phrase structure rules. Simplicity, then, is not just an aesthetic criterion but a heuristic for discovering the correct theory.
CLARIFY: This chapter is the theoretical core of the book. Chomsky is arguing that the goal of linguistics is not merely to list sentences or classify patterns but to discover the universal principles that underlie all human languages. The criteria of observational, descriptive, and explanatory adequacy provide a yardstick for evaluating theories: only transformational grammar, in Chomsky’s view, can reach the highest level. The appeal to simplicity and innateness foreshadows the later development of Universal Grammar, the idea that the basic architecture of language (including the concept of transformations) is biologically given. This is a radical departure from behaviorist or structuralist approaches, which treat language as a learned set of habits or patterns.
Chapter 12: The Boundaries of Transformational Grammar
LOCATE: In the final chapter of the main text, Chomsky reflects on the limitations and open questions of the transformational approach, acknowledging areas where the theory is incomplete and setting the agenda for future research.
PRESENT: Chomsky begins by noting that the transformational grammar he has outlined is a preliminary sketch, not a complete description of English syntax. Many phenomena remain unaccounted for, including coordinate structures (“John and Mary left”), comparative constructions (“John is taller than Mary”), and complement clauses with different types of embedding (“John believes that Mary left” vs. “John believes Mary to have left”). He suggests that these will require more complex transformations or additional base rules, but he is confident that the general framework is extensible.
Chomsky then discusses recursion in more detail. He points out that the recursive property of the base rules (the ability to embed sentences within sentences) is what gives language its infinite capacity. Without recursion, a grammar could generate only a finite set of sentences, which would be inadequate for natural language. The transformational component, by operating on the output of the recursive base, preserves and extends this infinity. Chomsky emphasizes that recursion is not just a formal property but a psychological one: speakers can produce and understand sentences of unlimited length and complexity because the grammar recursively generates them from a finite set of rules.
Chomsky also considers the question of phonological interpretation. He notes that the output of the syntactic component (the surface structure) must be input to a phonological component that converts it into sound. The transformational grammar interacts with phonology by rearranging morphemes (e.g., moving the affix to the verb) and by inserting or deleting elements (e.g., the dummy “do” or the passive “by”). The phonological rules then apply to the resulting string, assigning stress, intonation, and segmental phonemes. Chomsky acknowledges that this interface is only sketched in his work, but he argues that the transformational framework provides the necessary syntactic input for a complete theory of sound.
Finally, Chomsky addresses potential objections to the theory. One objection is that transformations are too powerful: they can generate any string from any other string, making the theory unfalsifiable. Chomsky responds that transformations are not arbitrary; they are constrained by structural conditions (like the A-over-A Condition) and by the requirement that they preserve the meaning of the underlying structure. A transformation that changed the subject-object relations, for instance, would be prohibited. Another objection is that the theory is too abstract: it posits hidden structures (deep structure) that cannot be directly observed. Chomsky counters that all scientific theories posit unobserved entities (like atoms or genes), and that the test of a theory is its predictive and explanatory power. The transformational theory, he argues, has already demonstrated this power by explaining ambiguities and structural relations that other theories cannot capture.
CLARIFY: Chomsky is both honest about the theory’s limitations and confident in its future. He explicitly sets a research program: to extend the transformational analysis to more constructions, to refine the conditions on transformations, and to integrate syntax with phonology and semantics. The chapter serves as a programmatic conclusion, inviting linguists to test and develop the theory. The discussion of recursion and infinity is foundational: it reminds the reader that language is not a finite set of sentences but a generative system. The defense against objections anticipates decades of debate about the abstraction and power of generative grammar, and it establishes Chomsky’s commitment to a rationalist, hypothesis-driven science of language.
Chapter 11: The Explanation of Syntactic Structure (Continued)
We are in the penultimate chapter of the work, where Chomsky consolidates the explanatory power of his theory by demonstrating how transformational analysis resolves fundamental problems in English syntax that phrase structure grammar alone cannot handle.
The central problem Chomsky addresses is the passive transformation in English. He begins by noting that active and passive sentences like “the boy hit the ball” and “the ball was hit by the boy” are not simply unrelated paraphrases but stand in a systematic grammatical relation. A phrase structure grammar would need to generate the passive sentence directly from the constituent structure, generating the Noun Phrase “the ball” as subject and “the boy” as the object of the preposition “by.” But this approach fails to capture the fact that the passive sentence is derived from the same underlying grammatical relations as the active. Chomsky writes:
“We have been assuming that the grammar of English is neatly divided into the two parts: phrase structure and transformations. In this chapter we shall consider some evidence that this assumption is essentially correct.”
To demonstrate this, Chomsky lays out the transformational rule for the passive in detail. The rule takes a structure of the form NP1 — Aux — V — NP2 and converts it into the structure NP2 — Aux + be + en — V — by + NP1. The symbol “en” represents the passive participle morpheme. This single transformation accounts for all passive sentences in English, regardless of the specific verb. The transformation operates on the terminal string generated by the phrase structure component, which includes the abstract morpheme “en” as part of the auxiliary system. The passive rule then detaches the first NP, moves the second NP to subject position, inserts “be” before the verbal affix “en,” and attaches the original subject NP as the object of the preposition “by.”
Chomsky then extends this analysis to the co-occurrence relations between verbs and their objects. In an active sentence, the verb “frighten” requires a direct object — “sincerity frightens the boy” is grammatical, but “sincerity frightens” is not unless the verb is used intransitively. The passive transformation preserves this co-occurrence relation: in “the boy is frightened by sincerity,” the verb “frightened” still requires the underlying object “the boy” to appear in the surface subject position. A phrase structure grammar would have to state these co-occurrence restrictions twice — once for active structures and once for passive structures — which misses the generalization. The transformational approach states the restriction only once, at the level of the underlying kernel sentence, and then allows the passive transformation to map that kernel into the passive form.
The significance of this analysis is that it explains why certain passive sentences are ungrammatical. Consider the sentence “John was read by the book.” This sentence is derived from the kernel “the book read John,” which is itself ungrammatical because “read” does not take an animate object in that way. The transformational account explains the ungrammaticality of the passive by reference to the ungrammaticality of the underlying kernel. Similarly, “the boy was frightened” (without a “by” phrase) is derived from an underlying kernel where the agent is unspecified, which is also a possible form in active sentences like “something frightens the boy.”
We continue within the same chapter, where Chomsky turns to the interrogative transformation to further demonstrate the explanatory power of transformational grammar.
The problem of forming questions in English presents a similar challenge to phrase structure grammar. The simplest declarative sentence “the boy will hit the ball” becomes the question “will the boy hit the ball?” by inverting the subject and the first auxiliary verb. But this rule is not simple to state in a phrase structure grammar, because the inversion must refer to a structural relation between the subject NP and the Aux constituent. A phrase structure grammar would need to generate both the declarative and interrogative forms independently, duplicating the entire phrase structure for questions.
Chomsky proposes a transformational rule for interrogation that operates on the same kernel sentences as the passive. The rule takes a terminal string of the form NP — Aux — V — (object) and inverts NP and Aux, producing Aux — NP — V — (object). This accounts for yes-no questions like “will the boy hit the ball?” But Chomsky immediately notes a complication: when the verb is a main verb rather than an auxiliary, the interrogative transformation requires the introduction of the auxiliary “do.” For example, “the boy hits the ball” (where “hits” is a main verb) becomes “does the boy hit the ball?” rather than “hits the boy the ball?” Chomsky accounts for this with the do-support transformation which inserts “do” when the interrogative inversion would otherwise leave a main verb in the auxiliary position.
The explanatory power of this analysis becomes clear when Chomsky considers negative questions. A sentence like “why didn’t the boy hit the ball?” involves both the interrogative transformation and the negative transformation. The negative transformation introduces “n’t” after the auxiliary, and then the interrogative transformation inverts the subject with the auxiliary-plus-negative. But notice that “do” appears only when the auxiliary would otherwise be empty. In “why didn’t the boy hit the ball?” the “do” appears because the negative “n’t” requires an auxiliary to attach to, and the main verb cannot move to auxiliary position.
Chomsky then demonstrates how the transformational ordering — the sequence in which transformations apply — explains otherwise puzzling data. The negative transformation must apply before the interrogative transformation in certain cases, because the negative affix must be attached to the auxiliary before the auxiliary is moved to the front of the sentence. If the interrogative transformation applied first, the negative would be left in the original subject position, producing an ungrammatical string. This ordering requirement is not an arbitrary stipulation but follows from the structural descriptions of the transformations themselves.
We continue with the same chapter, where Chomsky introduces the nominalization transformation to show how the same explanatory principles extend to noun phrase structures.
The problem of nominalization — deriving noun phrases like “the boy’s hitting of the ball” from the sentence “the boy hits the ball” — presents another case where phrase structure grammar would require duplication. A phrase structure grammar would need to generate “the boy’s hitting of the ball” as a Noun Phrase directly, specifying the internal structure as Determiner — Noun — Gerund — Preposition — Noun Phrase. But this misses the systematic relation between the nominalization and the underlying sentence.
Chomsky proposes a nominalization transformation that takes a kernel sentence of the form NP1 — Aux — V — NP2 and converts it into the noun phrase NP1 + ‘s — V + ing — of + NP2. This transformation operates on the same kernel sentences that underlie declaratives, passives, and interrogatives. The nominalization preserves the grammatical relations of the underlying sentence: the subject of the underlying sentence becomes the possessive NP (the boy’s), the verb becomes a gerund (hitting), and the object of the underlying sentence becomes the object of the preposition “of” (the ball).
The explanatory power of this analysis emerges when Chomsky considers the constraints on nominalization. The sentence “the boy hit the ball” nominalizes to “the boy’s hitting of the ball,” but the sentence “the ball was hit by the boy” nominalizes to “the ball’s being hit by the boy.” The distinction between active and passive nominalizations follows directly from the passive transformation applying before nominalization. If the passive transformation applies to the kernel “the boy hit the ball,” it produces the passive string “the ball was hit by the boy,” and then the nominalization transformation applies to that passive string to produce “the ball’s being hit by the boy.”
Chomsky also shows how the nominalization transformation interacts with the passive and interrogative transformations. For example, the nominalization “the boy’s being hit by the ball” is derived from the passive kernel “the boy is hit by the ball,” which is itself derived from the active kernel “the ball hits the boy.” The ordering of transformations must ensure that the passive applies before nominalization but after the kernel is generated. This ordering is not a stipulation but follows from the fact that nominalization is a sentence-to-noun-phrase transformation that must first have a complete sentence as input.
We remain in the same chapter, where Chomsky now turns to the verb phrase structure and the problem of the auxiliary system in English.
The English auxiliary system has long been a subject of confusion in traditional grammar. The sequence of auxiliary verbs in English follows a strict order: modal — perfect (have + en) — progressive (be + ing) — main verb. A sentence like “the boy may have been hitting the ball” shows this order: “may” (modal), then “have” (perfect), then “been” (progressive), then “hitting” (main verb). A phrase structure grammar can generate this sequence with a context-free rule like Aux → Modal (have + en) (be + ing), but this misses the systematic relations between these elements.
Chomsky demonstrates that the auxiliary system is best understood through a transformational analysis that separates the auxiliary elements from their affixes. In the kernel sentence, the auxiliary is represented as a set of abstract morphemes: Aux → Modal + (have + en) + (be + ing). The modal is a free morpheme, but “have” always requires the suffix “en” on the following verb, and “be” always requires the suffix “ing” on the following verb. These affixes are part of the auxiliary string, not part of the main verb. The transformation that attaches these affixes to the following verb is called the affix-hopping transformation.
The affix-hopping transformation moves the suffix “en” from after “have” to the verb that follows, producing “have + en + V” → “have + V + en” (e.g., “have hit” → “have + en + hit” → “have hit”). Similarly, “be + ing + V” → “be + V + ing” (e.g., “be hitting”). This transformation operates independently of the specific verb, applying to any verb that follows the auxiliary affix. The result is that the surface forms of the auxiliary — like “has beaten,” “is beating,” “will have been beating” — are all derived from underlying strings of abstract morphemes.
The explanatory power of this analysis becomes clear when Chomsky considers the interrogative transformation. In a yes-no question, the first auxiliary verb moves to the front of the sentence. But which auxiliary moves? In “the boy may have been hitting the ball,” the modal “may” moves to the front, producing “may the boy have been hitting the ball?” If the modal is absent, “have” moves: “the boy has been hitting the ball” becomes “has the boy been hitting the ball?” If both modal and perfect are absent, “be” moves: “the boy is hitting the ball” becomes “is the boy hitting the ball?” If no auxiliary is present at all, “do” is inserted: “the boy hits the ball” becomes “does the boy hit the ball?” This pattern follows directly from the analysis if the interrogative transformation moves the first element of the auxiliary string, and “do” is inserted when there is no auxiliary element to move.
We continue in the same chapter, where Chomsky turns to a deeper problem: the ambiguity of syntactic structures and how transformations resolve them.
Consider the sentence “the shooting of the hunters was terrible.” This sentence is ambiguous: it can mean either that the hunters were shooting (active meaning) or that the hunters were shot (passive meaning). A phrase structure grammar would generate this sentence with a single constituent structure — Noun Phrase — but would have no way of representing the two different meanings. The transformational analysis, however, derives the noun phrase from two different underlying kernel sentences. In the active reading, the noun phrase is derived from the kernel “the hunters shoot” via the nominalization transformation, producing “the hunters’ shooting” and then “the shooting of the hunters.” In the passive reading, the noun phrase is derived from the passive kernel “the hunters are shot” via the nominalization transformation, producing “the hunters’ being shot” and then “the shooting of the hunters” (where “shooting” is the nominalized form of “shot”).
Chomsky extends this analysis to other ambiguous structures. The sentence “visiting relatives can be tiresome” is ambiguous between the reading where “visiting relatives” means “relatives who visit” (the relatives are the subject) and the reading where “visiting relatives” means “visiting one’s relatives” (the relatives are the object). The transformational analysis derives the first reading from the kernel “relatives visit (someone)” and the second reading from the kernel “(someone) visits relatives.” The surface structure is identical, but the underlying kernel sentences are different.
The significance of this analysis is that it shows how transformational grammar accounts for systematic ambiguity that phrase structure grammar cannot handle. The phrase structure grammar can only assign one structural description to a sentence, but a transformational grammar can assign multiple deep structures to the same surface structure. This explains why native speakers can recognize multiple meanings in such sentences, and it predicts that any sentence derived from different underlying structures will be ambiguous.
Chomsky then shows how the same principle applies to sentences with multiple negation or multiple quantification. The sentence “I didn’t see no one” can mean either “I saw someone” (where “didn’t… no one” is a double negative that cancels out) or “I saw no one” (where “didn’t… no one” is a dialectal negative concord). The standard dialect uses the first reading, while some non-standard dialects use the second. The transformational analysis accounts for these differences by different rules for negative placement and negative concord.
We continue in the same chapter, where Chomsky now addresses the motivation for the kernel sentence as the fundamental unit of grammar.
The kernel sentence is defined as the set of simple, declarative, active sentences that are generated by the phrase structure component and serve as input to transformations. Chomsky emphasizes that the kernel is not a psychological or cognitive unit but a formal one — it is the simplest level of representation, from which all other sentence types are derived. The transformations then map these kernel sentences into the full range of surface structures.
Chomsky argues that the choice of kernel sentences is not arbitrary but is constrained by the requirement that the grammar be as simple and general as possible. The kernel sentences must be the minimal set that, when transformations are applied, generates all and only the grammatical sentences of the language. If the kernel is too large, the grammar will be unnecessarily complex; if the kernel is too small, the transformations will have to do too much work and will become impossible to state.
To illustrate this, Chomsky considers what would happen if the passive sentence were generated directly in the phrase structure component rather than derived transformationally. In that case, the grammar would need two sets of phrase structure rules — one for active sentences and one for passive sentences. The co-occurrence relations between verb and object would have to be stated twice: once in the form “the verb takes an object in active sentences” and once in the form “the verb takes a subject in passive sentences.” The grammar would also miss the generalization that the same verb appears in both active and passive forms. By deriving the passive transformationally, the grammar states the co-occurrence relations once and accounts for the passive as a systematic variation.
Chomsky also addresses the status of the phrase structure component itself. The phrase structure rules generate a set of strings that are not intended to be actual sentences of the language but rather underlying representations. The transformations then modify these underlying representations in systematic ways. The phrase structure component is, in effect, a “skeleton” that the transformations flesh out into actual sentences. This division of labor is what gives generative grammar its explanatory power: the phrase structure captures the fundamental relations between constituents, and the transformations capture the systematic variations that languages permit.
We continue in the same chapter, where Chomsky turns to the motivation for the passive transformation in particular, showing how it accounts for a range of data that phrase structure grammar cannot handle.
Chomsky examines the sentence “John was given a book” — a double-object construction where the passive sentence has two possible active counterparts: “someone gave John a book” and “someone gave a book to John.” The traditional grammar would treat these as two separate verbs or two separate constructions, but the transformational analysis shows that they are related through a dative transformation that optionally converts “give something to someone” into “give someone something.” The passive of the dative-shifted form “gave John a book” produces “John was given a book,” where “John” moves to subject position.
The explanatory power of this analysis becomes clear when Chomsky considers the constraints on the passive. The passive transformation applies only when the verb is transitive and takes a direct object. But the direct object can be either the first object in a double-object construction (“John” in “gave John a book”) or the second object (“a book” in “gave a book to John”). The fact that both “John was given a book” and “a book was given to John” are grammatical follows directly from the fact that the dative transformation can operate before the passive.
Chomsky then introduces a further constraint: the passive of the dative-shifted form “John was given a book” does not allow the preposition “by” to be omitted in the same way as ordinary passives. The sentence “John was given a book” implies an unspecified giver, but “John was given a book by Mary” explicitly states the agent. This follows from the structure of the underlying dative construction, where the agent is always present in the underlying kernel.
The analysis of the passive also explains why certain verbs do not have passive forms. For example, the verb “weigh” in “the suitcase weighs twenty pounds” does not have a passive form — “twenty pounds is weighed by the suitcase” is ungrammatical. This is because “weigh” in this sense is a stative verb that does not have an agent argument. The passive transformation requires an agent argument (the “by” phrase), so stative verbs like “weigh,” “cost,” and “contain” do not permit passivization.
We enter the final chapter of the work, where Chomsky turns from demonstrating the explanatory power of the theory to considering its limitations and the problems that remain unsolved. This chapter is titled “The Boundaries of Transformational Grammar,” and it serves as a honest assessment of what the new theory can and cannot yet account for.
Chomsky begins by acknowledging that the transformational analysis presented in the preceding chapters is far from complete. He writes:
“We have not attempted to present a complete grammar of English. We have been concerned only with a limited range of phenomena, and even within that range our formulations are tentative and incomplete. Nevertheless, the analysis we have given seems to indicate that the theory of transformational grammar is on the right track.”
The first boundary Chomsky discusses is the problem of lexical selection. The phrase structure component generates a string of grammatical categories — Noun Phrase, Verb Phrase, etc. — but it does not specify which lexical items (nouns, verbs, adjectives) can appear in those positions. The grammar needs a way to state that a particular verb can only take a certain class of nouns as subject or object. Chomsky acknowledges that this is a serious problem but suggests that it can be handled by adding selectional restrictions to the grammar without altering the basic framework.
The second boundary is the problem of compound sentences. The analysis in the preceding chapters has focused on simple sentences — kernel sentences and their transformations. But English also has compound sentences like “the boy hit the ball and the girl kicked the book” where two simple sentences are conjoined. Chomsky suggests that these can be handled by a conjunction transformation that joins two or more kernel sentences under certain conditions, but he notes that the conditions are complex and not fully understood.
Chomsky then turns to the problem of embedding — sentences like “the boy who hit the ball is tall” where one sentence is embedded inside another. The embedding involves relative clause formation, which is itself a transformation that converts “the boy is tall” and “the boy hit the ball” into “the boy who hit the ball is tall.” Chomsky proposes that the relative clause transformation takes a kernel sentence containing a Noun Phrase that matches a Noun Phrase in the matrix sentence and converts it into a relative clause attached to that Noun Phrase.
Chapter 12: The Boundaries of Transformational Grammar (Continued)
We continue in the final chapter, where Chomsky examines the problem of morphological and phonological manifestations of syntactic structure.
Chomsky acknowledges that the transformational grammar as developed so far has focused entirely on syntax, but the theory must ultimately connect to phonology — the sound system of the language. The transformations produce terminal strings that consist of abstract morphemes — like “en” for the passive participle, “ing” for the progressive, and “s” for the plural. These morphemes must then be converted into actual phonological forms by a system of morphophonemic rules. For example, the abstract morpheme “en” becomes “-ed” after regular verbs (“hit + en” → “hit + ed” → “hit”), “-t” after certain verbs (“bend + en” → “bent”), or “-en” after others (“write + en” → “written”).
Chomsky notes that the boundary between syntax and phonology is not sharp. Some phenomena that appear to be phonological — like the contraction of “will” to “‘ll” or “have” to “‘ve” — are actually sensitive to syntactic structure. The contraction is possible only when the auxiliary verb is in certain syntactic positions. For example, “I will go” can contract to “I’ll go,” but “I will not go” cannot contract to “I’ll not go” (though it can contract to “I won’t go”). This suggests that the syntactic structure must be available to the phonological component.
The problem of semantic interpretation is also addressed. Chomsky acknowledges that his grammar does not yet provide a semantics — a way to determine the meaning of a sentence from its syntactic structure. He suggests that the deep structure — the underlying kernel sentence — is the level at which semantic interpretation takes place. The transformations preserve meaning in the sense that the deep structure determines the grammatical relations (subject, object, etc.) which are crucial for meaning. But he concedes that “the relation between syntax and semantics is extremely complex and by no means well understood.”
Chomsky then discusses the problem of grammaticality judgments. A grammar must distinguish between grammatical and ungrammatical sentences, but native speakers often disagree about grammaticality. The sentence “the boy who hit the ball is tall” is clearly grammatical, but “the boy who hit is tall” (where the relative clause verb has no object) is questionable for some speakers. Chomsky suggests that these borderline cases are inevitable in any empirical science, and they do not undermine the theory.
We continue in the final chapter, where Chomsky turns to the problem of universality — whether the theory applies to all human languages or only to English.
Chomsky emphasizes that the theory of transformational grammar is intended to be universal — applicable to all natural languages. The specific phrase structure rules and transformations vary from language to language, but the general architecture — phrase structure component plus transformational component — is proposed as a universal property of human language. He writes, “If the theory of transformational grammar is correct, then each language consists of a phrase structure component and a transformational component, and the sentences of the language are those generated by the application of the transformations to the strings generated by the phrase structure component.”
Chomsky acknowledges that the evidence for universality is limited, because the theory has been applied only to English and a few other languages. But he suggests that the typological diversity of languages can be captured by differences in the phrase structure rules and the inventory of transformations. For example, a language like Latin, which has a relatively free word order, would have a different set of transformations from English, which has a rigid word order. But both languages would have the same underlying architecture.
The question of language acquisition is also raised. Chomsky argues that the theory provides a framework for understanding how children acquire language. A child learning English must infer the phrase structure rules and transformations of English from the sentences they hear. The theory provides a hypothesis space — a set of possible grammars — from which the child selects the one that best fits the linguistic data. This is consistent with the view that children are born with an innate capacity for language, which Chomsky would later develop into the theory of universal grammar.
Chomsky also considers the problem of economy — whether the theory can be simplified. He suggests that the phrase structure component and the transformational component might be reducible to a single component, but the evidence from the preceding chapters suggests that the division is necessary. The phrase structure captures the fundamental relations between constituents, and the transformations capture the systematic variations. Any attempt to eliminate the transformational component would result in an unacceptably complex phrase structure grammar.
We continue in the final chapter, where Chomsky concludes with a discussion of the empirical consequences of the theory and the challenges that remain.
Chomsky acknowledges that the theory of transformational grammar is still in its infancy and that many problems remain unsolved. He lists several specific problems that future research must address:
First, the problem of ordering transformations — the sequence in which transformations apply — is not fully understood. In the preceding chapters, Chomsky has assumed a fixed ordering, but he recognizes that the ordering might be determined by general principles rather than arbitrary stipulation. The question of whether transformations can apply to their own output (recursive application) is also open.
Second, the problem of the kernel sentence — whether all sentences can be derived from a single, simple kernel — is not fully resolved. Some constructions, like the passive, clearly derive from kernel sentences. But others, like the double-object construction, might require transformations that apply to non-kernel strings.
Third, the problem of lexical categories — whether the traditional parts of speech (Noun, Verb, Adjective, etc.) are adequate — is raised. Chomsky notes that his phrase structure rules use these categories, but the categories might need to be refined or expanded to account for all syntactic phenomena.
Chomsky also addresses the problem of the limits of transformation — whether any syntactic phenomenon can be captured by a transformation or whether some phenomena resist this treatment. He suggests that the limits of transformation are determined by the structure-dependence of transformations — they can only operate on structural descriptions that are defined by the phrase structure component. This means that transformations cannot refer to individual words or to non-syntactic properties like meaning.
The final problem Chomsky discusses is the problem of completeness — whether the grammar can generate all and only the grammatical sentences of a language. He acknowledges that no complete grammar of English exists and that the task of constructing one is enormous. But he argues that the theory of transformational grammar provides the right framework for this task, and he expresses confidence that future research will fill in the details.
We have now reached the end of the work. The final chapter concludes with a call for further research and a restatement of the generative hypothesis: that language is a system of rules that generates the set of grammatical sentences, and that the theory of transformational grammar provides the most promising framework for discovering and stating these rules. No summary or closing statement is needed beyond this point.
