Ray Jackendoff, Foundations of Language: Brain, meaning, grammar, evolution. Oxford and New York: Oxford University Press. 2002. Pp. xix, 477. Hb $40.00.
Reviewed by Robbins Burling
Anthropology, University of Michigan
Ann Arbor MI, 48109
rburling@umich.edu
In this extraordinary book, Jackendoff proposes nothing less than a new way to understand the architecture of language and a new way to view the relation of language to the brain, to the mind, to behavior, and to the evolution of our species. It is, among many other things, an invitation for cooperation from one of the world's leading formal syntacticians to linguists of diverse orientations and to those from adjacent fields, including sociolinguistics and linguistic anthropology. If we don't want to be left behind, we had better pay close attention.
Consider, first, the architecture. Jackendoff wants to escape what he calls the "syntactocentrism" of all the standard generative models where syntax has always been taken to be the central component of language. Jackendoff points out that phonology (PF, Phonetic Form) and meaning (LF, Logical Form) have always been treated as if they are derived from syntax.
Even in the Minimalist Program, where D- and S-structures (deep and surface structures) are formally eliminated, syntax emerges as movement and merge operations combine lexical items according to their intrinsic lexical constraints (pg. 110). After this, the derivation splits, as in all
earlier versions of generative grammar, leading in one direction to phonology (PF, Phonetic Form) and in another direction to meaning (LF, Logical Form).
Instead of treating phonology and meaning as if they are both derived from syntax, Jackendoff puts conceptual structure at one end, and phonological structure at the other, with syntax in the middle. At one end, conceptual structure interfaces with perception and action. At the other, phonology interfaces with hearing, vocalization, and even with gestures (e.g., hand beats for stress) and with music. Conceptual and phonological structures also interface directly with each other, and both interface with syntactic structure. (Diagrams of these interfaces are found on pp. 125 and 272). A speaker starts with meaning, processes what he wants to say through syntax and phonology (though with multiple feedback loops) until noises emerge. The hearer starts with the noises and processes it in reverse. Of course the classical syntacto-centric model, where everything begins with syntax, was not supposed to be a performance model, but all else being equal, a model that conforms to the way language is really used should be welcomed over one that does not. Those who are not immersed in formal syntax may find my simplified description of Jackendoff's architecture to reflect nothing more than common sense, but coming from a student of Chomsky, it amounts to a radical dethronement of syntax from its ruling position.
Inspired by autosegmental phonology, Jackendoff wants to extend its tier structure to the two other major components of language: syntax and semantics. Although he does not work out the details in this book, the entire assemblage of tiers gives language what Jackendoff calls a "parallel architecture". Like the prosodic, syllabic, segmental, and morphophonological tiers of phonology, each of the tiers of the syntactic and conceptual structures has considerable autonomy, but all are related to one another by "interface rules". These allow one tier to constrain another, but never to fully determine another's form.
Important among the interfaces is the lexicon, and the lexical entry for a word, in Jackendoff's architecture, is "... a small-scale three-way interface rule. It lists a small chunk of phonology, a small chunk of syntax, and a small chunk of semantics, and it shows how to line these chunks up when they appear in parallel phonological, syntactic, and conceptual structures" (131). Words are prototypical lexical items, of course, but many other bits of language can also be usefully looked on as lexical items, and one of the most dazzling aspects of Jackendoff's architecture is a typology of lexical items that includes much more than just words.
Start with affixes and idioms. Affixes are smaller than words, and idioms are larger, but both have both meaning and phonology, and since both are productively embedded in larger syntactic constructions, they also have syntax. Since they interface the same three components of language as words do, both affixes and idioms certainly need to be recognized as lexical items. We also find bits of language that are usefully seen as lexical items even though they are defective in one way or another. Words such as "hello" and "yes" have meaning and phonology but they do not occur as parts of larger constructions so they are defective in syntax. Dummy it (it's raining) and supportive do (what do you want?) have syntax and phonology but no meaning. Neither syntax nor meaning, then, is an essential part of a lexical item.
What about phonology? Consider the resultative construction, exemplified by such sentences as Wilma watered the tulips flat, Clyde cooked the pot black, or Drive your engine clean. This construction consists of an ordered sequence of variables, an NP subject, a Verb, an NP direct object and a Predicate Adjective. In any particular example, these variables are realized by particular lexical items, but the construction itself consists of nothing but variables so it has no phonology of its own. The construction does have a meaning since it reports the result of some action, and it has syntax. It is impossible to list every example of the resultative construction in the lexicon because new examples can always be productively generated. Rather, it is the construction itself that needs to be listed in the lexicon. Even without phonology, its syntax and meaning are enough to let it be used with other lexical items, including words, to form sentences (pg. 175-6). From here, it is a relatively short additional step to recognize phrase structure rules as lexical items that consist of nothing but a sequence of ordered variables. This gives them syntax, but neither phonology nor meaning. If this stark summary is unconvincing, read the book. I think you will be convinced.
The book has three parts. In the first, "Psychological and Biological Foundations" Jackendoff argues for the vast complexity of language, places it firmly in the mind, and argues that we need to recognize a Universal Grammar (UG) that allows each individual to learn a language. Part II, Architectural Foundations, presents his ideas about the architecture of language. Part III, "Semantic and Conceptual Foundations", should be the part that articulates most closely with the interests of readers of this journal, but I found it the most challenging section. I had not been familiar with Jackendoff's extensive earlier work in this area so many of the ideas were new to me. It required close reading and lots of thought.
The most difficult part of Jackendoff's program for many readers to accept is likely to come when he deals with conceptual structures and advocates what he calls "pushing 'the world' into the mind" (303-306). What this means is that language does not refer directly to the world, but rather to the world as conceptualized by the speaker. Our conceptual structure, after all, includes plenty of things that are not in the world at all, from Sherlock Holmes to a perfect triangle, and we certainly refer to such things. Conceptualization is there in the head along with phonology and syntax. Reference becomes the relation of syntax and phonology, via all those interface rules, to our conceptualization, rather than the relation of language directly to the world outside. Having pushed reference firmly into the mind, Jackendoff finishes the book with surveys of lexical semantics (Chapter 11) and phrasal semantics (Chapter 12) that should delight linguistic anthropologists, for they bring meaning firmly back into linguistics. It is impossible to summarize these rich chapters in a few sentences, but perhaps the tone of his message can be conveyed by one quotation: "On the other hand, these difficulties [in the study of lexical semantics] in themselves point out one of the fundamental messages of generative linguistics: We language users know so much. And hence as children we learned so much--starting with some innate conceptual basis of unknown richness. Next to lexical semantics, the acquisition problem for grammar pales by comparison" (377, emphasis in the original). This is a point that some of us have suspected for a long time, but it amounts to a quiet revolution from a linguist whose teacher has done his best to persuade us that what really matters about language is what is built in rather than what is learned.
One of the most appealing aspects of the book is Jackendoff's moderation, indeed his common sense. He believes that there is something special about the human brain, a UG that makes language learning possible, but he does not want to wall off language in an isolated "language box," and he recognizes the enormous contribution of learning to the full achievement of language competence. He sees some modularity in the mind and in language, but would rather ask to what degree and in what ways they are modular than to argue about whether they are or are not modular (229). He feels that we need a competence--performance distinction, but he does not want to forget performance and he leaves a place for it in his architecture (Ch. 7). He is skeptical about principles and parameters (190). In both ontogeny and phylogeny, he sees syntax and phonology as developing in the context of an earlier and more complex conceptual structure. Apes, like human beings, after all, need to think about the world and about each other, and language presumably began to emerge in an animal whose conceptual structure was not so different from that of apes. He welcomes contributions from neighboring fields He is forthright about what he finds useful in other's work and what he does not, but he avoids the aggressive polemics that have marred so much of the linguistics of the last few decades. One can hope that he will help to set the tone for a more civil linguistics of the 21st century.
I believe this book has the potential to reorient linguistics more decisively than any book since Syntactic Structures shook the discipline almost half a century ago. It offers a vision of a post-Chomskian linguistics that is true to the brilliant insights that have come from Chomsky and from his many follows and antagonists, but leaves out some of the unnecessary baggage that has encumbered so much of formal linguistics. In rejecting syntactocentrism, in leaving a place for performance, in his opening to neuroscience, cognitive science, and language evolution, and in his serious concern for meaning, Jackendoff has offered a kind of linguistics to which scholars from neighboring fields should be able to relate more enthusiastically than they have been able to relate to the orthodox generative linguistics of recent decades.
The book, alas, will seem too big and too technical for the tastes of most nonlinguists, but everyone who calls herself or himself a sociolinguist or a linguistic anthropologist had better absorb its many messages. It is an invitation from a leading generative linguist to cooperate in the enterprise of understanding language, not only for itself, but for its role in evolution, biology, physiology, ontogeny, society, and culture. The final sentence of the book says it well: "Above all, my hope for the present work is that it can help encourage the necessary culture of collaboration" (429). It is my hope too.
(Received 10 September, 2002)