第4章
Language as a Window into the Mind
Chomsky's generative grammar represents a revolutionary departure from structuralism by attempting to make explicit the implicit knowledge that native speakers possess about their language. Traditional grammars frequently overlook fundamental patterns that native speakers intuitively understand and apply. For example, French speakers inherently know not to say "La grammaire est tres generative" despite it following the same surface pattern as many acceptable French sentences. Similarly, English speakers automatically recognize that while "The child seems happy" is correct, "The child seems to happy" is not, without ever having been explicitly taught this rule.
Chomsky proposes a sophisticated formal model anchored by a base component containing rewriting rules that generate sequences mechanically. To illustrate this concept, consider a simple three-word language consisting of only "Jules," "Chloe," and "loves." Native speakers instinctively understand which combinations create valid sentences. A grammar with precise rewriting rules can systematically generate acceptable sequences like "Jules loves Chloe" and "Chloe loves Jules" while automatically excluding invalid combinations like "Loves Jules Chloe" or "Chloe Jules loves." This demonstrates how complex linguistic knowledge can be reduced to formal principles.
These linguistic structures can be visualized and analyzed through tree diagrams or phrase markers that illuminate the hierarchical relationships between sentence elements. However, natural language grammar proves far more intricate, necessitating additional theoretical machinery beyond simple rewriting rules. Chomsky convincingly demonstrated the need for at least two other crucial levels: a morphophonological component handling sound patterns and word formation, and a transformational component that converts one syntactic structure into another, explaining phenomena like passive voice or question formation.
The development of generative grammar can be traced through three distinct periods: The first phase (1950s-mid-1960s) focused on establishing linguistics as a rigorous science, introducing formal methods and explicit rule systems. The second phase (1965-1970) addressed the crucial role of semantics and meaning in grammatical structure. The final phase (post-1970) explored universal grammar and the biological foundations of language acquisition.
Chomsky details how his approach evolved from structuralism while addressing its limitations. Initially, he attempted to refine structuralist discovery procedures while privately questioning their "fictionalist" interpretation of linguistic reality. For several years, he worked to overcome the inherent defects in these procedures, aiming to develop grammars with infinite descriptive capacity from finite data sets. This effort revealed fundamental problems with the structuralist approach.
Two critical questions emerged from this investigation: whether these linguistic methods required psychological interpretation, and whether they accurately reflected the biological foundations of language acquisition. Chomsky ultimately concluded that structuralist procedures were fundamentally inadequate, requiring replacement with more abstract principles. In contrast to empiricist approaches that build knowledge through classification and induction from experience, Chomsky's theory posited innate linguistic principles that determine the range of possible grammars. Under this view, language acquisition becomes a process of selecting the optimal grammar compatible with available linguistic data, rather than building knowledge from scratch through exposure alone.
第5章
The Transformational Revolution in Linguistics
Chomsky clarifies the fundamental difference between Harris's transformations and his own, marking a pivotal shift in linguistic theory. In Harris's view, transformations are primarily generalizations about linguistic data that show equal acceptability under systematic substitution. For Harris, grammar served merely as a compact description of observed language patterns, with transformations being tools for organizing and describing these observations. Harris maintained that alternative linguistic descriptions couldn't conflict in the way scientific theories do, viewing them more as different ways of organizing the same data.
In contrast, Chomsky's transformations in LSLT represent a more complex and theoretically rich system. These transformations function as rules within a comprehensive system that assigns structural descriptions to an infinite class of sentences. They operate on abstract representations (deep structures), systematically transforming them through multiple steps into surface structures that represent actual utterances. This process requires sophisticated empirical arguments, not just substitution relations, to establish transformations in generative grammar. Unlike Harris's "incorrigible" transformations, those in generative grammar are treated as scientific hypotheses-never finally established and always subject to revision based on new evidence and theoretical insights.
Chomsky makes an important distinction between formalization and mathematics in linguistics. While the generative program demands precise, formalized principles for describing language structure, this formalization isn't mathematics per se. Mathematical linguistics properly begins when researchers study the abstract properties of these formalizations, proving theorems about these systems and their capabilities. This distinction helped clarify the relationship between linguistic theory and mathematical methods.
The late 1940s and early 1950s saw enthusiastic application of information theory and automata theory models to language, particularly finite state Markov sources, championed by linguists like Charles Hockett. Chomsky's skepticism toward these approaches stemmed from two sources: political concerns about behaviorist manipulation of human behavior, and his deep intuition that these models were fundamentally inadequate for capturing the complexity of natural language. In his groundbreaking work "Syntactic Structures" (1956), he provided formal proof that natural languages possess properties that cannot be expressed by such simple mechanical models.
This theoretical breakthrough led to the development of more sophisticated formal systems, particularly "phrase structure grammars" and their important subset, "context-free phrase structure grammars." These became a distinct branch of mathematics, studied extensively by mathematicians like Marcel-Paul Schutzenberger, who collaborated with Chomsky on fundamental results in formal language theory. Through this work, Chomsky demonstrated that empiricist learning theories face fundamental limitations, as shown by the inability of finite state models to capture even basic linguistic knowledge that every native speaker possesses. This argument became central to his critique of behaviorist approaches to language acquisition and his development of the theory of Universal Grammar.
The transformational revolution thus represented not just a new way of analyzing language, but a fundamental reconceptualization of the nature of linguistic knowledge and its acquisition, with far-reaching implications for psychology, philosophy of mind, and cognitive science.
第6章
The Semantic Dimension: Meaning and Structure
Chomsky introduces the evolution of his model from Syntactic Structures to the Standard Theory presented in Aspects of the Theory of Syntax (1965). The new model represents a significant advancement in linguistic theory, consisting of three primary components: a base component (comprising rewriting rules and lexicon) that generates deep structure, a transformational component that converts this to surface structure, and two interpretive components-phonological and semantic. The semantic component, developed through collaborative work by Fodor, Katz, and Postal, aimed to make explicit what speakers intuitively know about intrinsic word and sentence meanings through detailed lexical descriptions and projection rules. This framework revolutionized how linguists approached meaning in language.
Chomsky clarifies his frequently misunderstood position on the autonomy of syntax. He demonstrates this through examples like "Colorless green ideas sleep furiously" - a sentence that is grammatically perfect yet semantically nonsensical. While he maintains that syntactic elements operate independently and aren't established primarily on a semantic basis, he strongly rejects the common misattribution that he excluded semantics from linguistics altogether. In fact, much of his early work focused intensively on semantic interpretation of formal systems, particularly in exploring how meaning emerges from structural relationships.
When Fodor and Katz proposed integrating semantic interpretation rules into the Standard Theory, they introduced an innovative intensional approach with semantic representations based on universal semantic categories, analogous to phonetic features. Their work suggested that meaning could be decomposed into fundamental units, similar to how sounds are broken down into distinctive features. While Chomsky accepts that certain semantic properties like "agent," "instrument," and "goal" may be universal, and that semantic relations between words like "persuade," "intend," and "believe" can be expressed in purely linguistic terms, he remains skeptical about completely separating semantic representation from beliefs about the world. Drawing on philosophical insights from Wittgenstein, Quine, and Leibniz, he argues that concepts inherently involve beliefs about real-world behavior-we wouldn't know how to name or categorize an object that doesn't conform to natural laws or our experiential expectations.
Chomsky introduces and explains his specific use of "logical form" as designating linguistic representation that incorporates semantic properties strictly determined by linguistic rules, distinct from broader semantic representation that involves other cognitive systems. He clarifies this distinction through various examples of co-reference principles, which he argues are well understood and governed by explicit linguistic rules. For instance, in "John sees him," John and him cannot refer to the same person-a purely linguistic constraint independent of context or world knowledge. Similarly, in constructions like "The man who entered wanted to know who he was," the pronoun "he" cannot refer to "the man" - demonstrating how structural relationships constrain possible meanings. These binding principles reveal the intricate interaction between syntax and semantics while maintaining their distinct domains.
第7章
Deep Structure and Universal Grammar
Chomsky explains the evolution of the concept of "deep structure" in generative grammar. Initially, the closest concept was the T-marker representing transformational history, showing how sentences relate to each other. In the Standard Theory, deep structure became "overdetermined"-generated by the base, receiving lexical items, undergoing semantic interpretation, and converting to surface structure. These properties were later dissociated as the theory evolved.
Theoretical terms naturally change meaning as empirical hypotheses develop, though this causes more confusion in linguistics than in natural sciences. To reduce confusion, he later replaced "deep structure" with "initial phrase marker," though he acknowledges that technical terminology is often misinterpreted regardless of the terms used.
Surface structure is actually quite abstract, incorporating properties not visible in physical form. The "structuring effect of silence" through trace theory affects both phonology (changing intonation) and semantics (blocking co-reference). Chomsky rejects accusations of "returning to structuralism," noting that modern surface structure is defined differently and is more abstract than structuralist concepts. He emphasizes that scientific progress requires changing one's mind frequently as better approximations emerge-unlike in humanities where positions become personalized and unchanged. "Anybody who teaches at age fifty what he was teaching at age twenty-five had better find another profession. If in twenty-five years nothing has happened which proves to you that your ideas were wrong, it means that you are not in a living field, or perhaps are part of a religious sect."
Chomsky defines universal grammar as "the system of principles which characterizes the class of possible grammars"-not a universal language but a metatheory specifying how grammars are organized. It represents empirical hypotheses about the biological language faculty that enables children to acquire language despite limited data. Unlike earlier work that permitted numerous rules, Chomsky now focuses on limiting rule expressivity through general principles governing rule application. He emphasizes that universal grammar is not an innate grammar but a theory of grammars-a schematism constraining what counts as possible human language.
第8章
Knowledge, Cognition, and Human Nature
Knowledge should be understood as abstract cognitive structure rather than mere capacity or ability. While knowledge enters into our capacities to act, it shouldn't be identified with them-one might possess the cognitive structure "knowledge of English" without the capacity to use it, as in cases of aphasia where understanding remains intact but production is impaired. Similarly, certain intellectual activities may involve only dispositions and habits with no underlying cognitive structures, such as trained reflexes or memorized procedures that don't require deep understanding.
Modern analytic philosophy's tendency to employ "disposition" or "capacity" where "cognitive structure" is more appropriate represents an unfortunate residue of empiricism. These behavior-oriented concepts discourage inquiry into the mind's structure by focusing solely on observable outcomes rather than internal mechanisms. For instance, describing mathematical ability purely in terms of problem-solving behavior overlooks the rich cognitive architectures that enable mathematical thinking. The proper approach is to determine the structure of the mind and its products-there's nothing mysterious about abstract cognitive structures created by innate faculties, represented in the brain, and entering into our capacities to act and interpret.
For problems outside an organism's cognitive capacity, it must resort to trial and error, association, and simple induction. Consider how a rat navigates a maze versus how humans solve complex physics problems - the rat's learning is limited to associative processes, while humans can construct abstract models. Humans might construct conscious scientific theories in such domains, but our "science-forming" capacities are themselves special and restrictive. The same limitations that enable scientific understanding in certain domains also sharply constrain the class of humanly accessible sciences. This explains why we excel at mechanics and chemistry but struggle with quantum phenomena that defy our intuitive understanding. As biological organisms, it's merely lucky when our cognitive capacity matches scientific truth in any area, explaining why there are so few sciences and why much human inquiry lacks intellectual depth.
Children never make certain types of errors in language learning, revealing innate constraints on their cognitive architecture. When forming questions from sentences like "the man who is tall is in the room," they unerringly move the correct "is" (the one following the noun phrase), never the one embedded in the relative clause. Even more striking, they avoid similar errors with more complex sentences like "the man who is happy is the one who is tall." This demonstrates they're using structure-dependent rules rather than simpler structure-independent ones based on linear order. The principle of structure-dependence isn't learned but forms part of the conditions for language learning-a fundamental component of Universal Grammar. This innate knowledge guides language acquisition across all human languages, despite vast surface differences between them.
第9章
Freedom, Creativity, and the Human Future
The empiricist blank slate view carries complex historical associations that belie its progressive reputation. While it emerged partly as opposition to religious dogma and hereditary privilege, this view paradoxically serves as a foundation for manipulative social control. As historian Ellen Wood astutely notes, the debate over mind isn't merely epistemological but fundamentally concerns human freedom - whether we're "responsive cogs" in a deterministic machine or "creative, determinative" beings capable of authentic self-direction. This distinction has profound implications for how we structure society and understand human potential.
The rationalist tradition, stemming from Cartesian thought, presents a rich intellectual lineage affirming human agency. This includes Rousseau's passionate opposition to tyranny and defense of natural rights, Kant's sophisticated defense of freedom and moral autonomy, Humboldt's vision of precapitalist liberalism emphasizing individual development, and Marx's penetrating critique of alienated labor and human essence. Together, these thinkers affirm our "species character" of free conscious activity - our fundamental nature as beings who create, reflect, and act with purpose. Kant powerfully described "man's inclination and duty to think freely" as nature's most carefully developed "germ," suggesting this capacity for autonomous thought lies at the core of human nature.
While human instinct theories are often associated with conservative political views, the "empty organism" doctrine actually provides a more insidious foundation for authoritarianism. If humans lack any essential psychological nature or inherent capacities, what prevents their wholesale manipulation by those claiming special knowledge or expertise? This doctrine conveniently serves both vanguard parties claiming to lead the masses and technocratic managers who monopolize decision-making in state capitalist systems, justifying their control through appeals to scientific management of human resources.
True freedom, paradoxically, requires recognizing intrinsic constraints. As anarchist philosopher Bakunin insightfully observed, "the laws of our own nature... constitute the very basis of our being" and provide "the real condition and effective cause of our liberty." Even creativity itself depends on rule-systems partly determined by human cognitive and emotional capacities. Any viable libertarian social theory must therefore identify these natural laws to establish meaningful parameters for social change and human development.
The rationalist tradition suggests our "species character" provides the essential framework for moral consciousness, cultural achievement, and participation in just communities. While it's a significant leap from cognitive development to specific social conclusions, this tradition offers compelling insights about human nature and potential. If successful, such investigation might definitively refute Russell's pessimism about human nature, revealing instead that our passions and instincts, when freed from competitive and authoritarian structures, could establish a new civilization where human nature truly flourishes through cooperation, creativity, and conscious self-direction. This vision suggests the possibility of social arrangements that align with and nurture our deepest capacities rather than suppress or distort them.
