The two programs were in fact quite complementary, and the tensions between them not only bound each to the other, but also steered them jointly on to a more productive path than either of them individually might otherwise have taken.
—Geoffrey Huck and John Goldsmith (1995:3)
I am still a Chomskyan, as far as linguistics is concerned, after my fashion (and my own interpretation of “Chomskyan”). Indeed, as far as generative grammar is concerned, I might fairly claim to be a more orthodox Chomskyan (in the original sense of the adjective) than Chomsky himself has latterly become.
—John Lyons (1991:154)
The Generative/Interpretive Semantics feud was not an honorable episode in the history of linguistics. A time of “bad vibes and yelling and shrieking and shit like that” (Ross), it left a pleasant taste in no one’s mouth. Chomsky regards the period as one of acute irrationality, seeing the Generative Semantics movement as coterminous in time and motivation, in its rise and its dissolution, “with the big popular movements of the 1960s,” just another “cultish fad.” Not that the irrationality touched him or his students and immediate colleagues in any way, he suggests, but all the same he was glad to see the back of the dispute. Postal, too, was happy to see it end, and like Chomsky also regards the whole chapter as fundamentally misguided, though for very different reasons. “After all,” he says, it was “about the right form of transformational grammar”; not unlike a wrangle about the “right form of phlogiston theory” in his view. “The bad thing is not that Generative Semantics disappeared,” for Postal, “but that the other branch of transformational theory didn’t disappear” (Huck & Goldsmith 1995:138). Fodor has much the same reaction, yet again for different reasons. “A plague on both their houses,” he says, in the Shakespearean phrase not uncommon for the episode: “It was what they shared that turned out to be false [excessively abstract representations], not what distinguished them.”
McCawley lamented the clash. While he did the most to keep the methodological and theoretical flames of Generative Semantics flickering long after everyone else had left the hearth, he complained about the “dismaying readiness of linguists on both sides of the dispute to assent to . . . kindergarten-level caricatures of the facts” (1999:158). Robin Lakoff remembers it more fondly than most, writing a wistful memoir of the period. The division grew inevitably, she relates, toward “the realization that there were finally and irreparably two schools, no more unity, no more us-against-the-world. Camelot had fallen.” When that happened, when it was no longer an in-house debate over formalisms, methods, and the architecture of a shared model, when “GS realized the futility of the battle, it disintegrated on its own. Much of the fun was in the fighting, and without fun, there was no GS” (1989:970).
George Lakoff does not betray the same nostalgia, saying flatly “I’m glad the Generative Semantics days are over” (Huck & Goldsmith 1995:119). He made the remark in the context of what he now sees as the technical errors of the program, but he is surely embarrassed by some of his personal excesses as well. And his recurrent foe, Jackendoff, talks sheepishly of the times he “would try to make a point and end up screaming,” adding a more disciplinary regret as well, perhaps shared by others: that the feuding “had a lot to do with driving psychologists, and philosophers, and computer scientists away from the discipline. . . . I think it did more damage outside of linguistics, to the image of the field, than it did inside.”1
Even Pieter Seuren, who regards Generative Semantics as “by far the most inspired and promising development in theoretical linguistics during the twentieth century”—indeed, who inscribed this belief in the ultimate narcissistic historical textbook of linguistics, five hundred pages tracing a progressive arc from Socratic Athens to Generative Semantics (Seuren 1998)—cannot mention the program without bitterly recounting its fate as the result of a “vicious sociological character assassination, likewise unique of its kind in twentieth century linguistics if not in the entire history of the subject.” Nor does the fault for this epic disgrace fall only on Chomsky and the “large numbers of mediocre young linguists [following] the Master uncritically [in] reject[ing] Generative Semantics merely on the strength of his authority.” Generative Semanticists are equally guilty because (in an excruciatingly yoked pair of heterogeneous images) they “had knees that were . . . weak . . . and ran for it,” instead of standing and fighting.2
It was not honorable, but I’m with Seuren at least part of the way: it was important. The friction was productive. Transformational Generative Grammar dominated twentieth century linguistics. It had broad implications for the manifold overlapping studies of what it means to be human, with major impacts on philosophy, psychology, English studies, and computer science, and ripples of influence in fields like anthropology, sociology, even art history; the term Deep Structure alone can be found in almost every intellectual pursuit practiced in the latter twentieth century, and not a few anti-intellectual pursuits in the bargain. The program may yet have the transformative effects on neuroscience and genetics that Chomsky and like-minded researchers currently predict. It was a Big Deal. A Chomsky-sponsored descendant of that program is still highly influential, and may prove to be yet an Even Bigger Deal.
Linguistics is in a very good place now, but maybe the particular path it has folowed is irrelevant. Maybe we would have ended up pretty much where we are now without the Linguistics Wars. Maybe Chomsky’s personality, or the instability of the framework, or the remarkable ingenuity of linguists, or just the richness of language and the messiness of verbal behavior, would have set off the cascading revisions we have seen in Transformational Generative Grammar, with the attendant innovations, retrenchments, and parallel enterprises that accompanied those revisions. Maybe the Chomskyan trajectory was as ineluctable, without the Deep-Structure-stretching shitstorm of Generative Semantics and its Interpretivist repudiation. Maybe Jackendoff’s interests on their own, nurtured by Chomsky in an environment where the horsefolk were not hammering on the door, would have set off the cascade. Maybe restrictiveness would have become a byword without globality at the gate. Maybe the proliferation of Generative models explicitly positioned against various Chomskyan tools or methods or goals was inevitable without the challenge of Generative Semantics. Maybe some natural corrective to the equation, linguistics = grammatical model building, would have kicked in, triggering the interest in rhetorical, pragmatic, sociolinguistic and other “performative” dimensions of language that flowered in the 1970s and have continued to bloom since. Maybe that dam would have broken without Generative Semantics.
Maybe—if Postal had become a field linguist and spent his career on the intricacies of Mohawk, if George Lakoff had stayed in English studies or mathematics, if Robin had never ventured out of Radcliffe and found herself in Chomsky’s charismatic ambit, if Ross had never returned to MIT after Halle told him he wasn’t the right material for their program, if McCawley had become the world’s most gifted Sino-Scottish chef, if Fillmore and Bach and Zwicky and Sadock and Levi and Binnick and Green and so on had found other scholarly pursuits, or if everybody had just nestled comfortably into the Chomskyan fold—the same fecund churning of data and renovation of theory would have occurred in that alternate universe. Maybe.3
Maybe an Aspects monolith would have determined the field of linguistics for the half century after its arrival, an utter Chomskyan hegemony ruling the field, with, let us say, a natural and uncontroversial drift toward Principles and Parameters and beyond—no dissent, few perturbations—while such developments as Frame Semantics, Cognitive Linguistics, and Construction Grammar arose spontaneously and harmoniously in Berkeley, nurtured by fair-trade chai lattes on Telegraph Avenue, organic produce, and yoga.
Maybe, in other words, history doesn’t matter, people don’t matter, debate doesn’t matter.
What we do know is that the record before us, in the history we are living, gives us a three-stage answer to why that monolithic, hegemonic Chomskyan framework did not come to pass, and why seeds took root that grew into a coalition of approaches defined in significant ways by their opposition to central Chomskyan tenets: (1) the emergence of Generative Semantics, (2) the Interpretivist backlash, and (3) the ensuing Linguistics War.
There are several motivations for proposing a generative semantic theory.
George Lakoff (1976a [1963]:50)
In the summer of 1975, I realized that both transformational grammar and formal logic were hopelessly inadequate and I stopped doing Generative Semantics.
—George Lakoff (1997:39)
In 1965–66, Chomsky was not moving quickly enough to the Promised Land for most Chomskyans. Syntactic Structures made modest promises about semantics, which helped attract a squad of eager recruits. With that influx the promises lost their modesty, though it was not just the recruits making them. Aspects expanded the semantic promises, Cartesian Linguistics expanded them, Language and Mind expanded them: Deep Structure encoded the primary elements of meaning; transformations delivered the variable arrangements of meaning that we see and hear on the surface; those elements were universal, philosophically backed, psychologically real, and they fundamentally implicated human creativity. The neighbors even thought, not without encouragement, that those elements were computationally tractable, pedagogically sound, psychologically testable, and analytically productive. The mediational goal of linguistics—a consistent, coherent, robust linkage between concept and form, meaning and sound—was imminently soluble, and the human sciences stood to benefit profoundly. Some program with a fuller and more explicit semantics, growing out of Aspects, seemed inevitable.
We can see now, from Chomsky’s earliest writings as well as his post-Aspects response, that he was never going to give meaning the controlling vote in his model. It is even clearer, from every word Chomsky has ever penned, that Chomsky was not going to give up. So, there was a new program in Transformation Town and there was obdurate resistance from the sheriff.
Science is a full-blooded activity; Hobbes calls it “a Lust of the mind” paling even the “short vehemence of any carnal Pleasure” (1991 [1651]:42). Science demands an extraordinary level of personal investment. The personalities it both attracts and induces, especially among its most successful practitioners, exhibit intense, passionate commitment, and whenever passion is in the same room with disagreement there will not be peace. The Generative/Interpretive Semantics dispute was no more acrimonious than many scientific clashes. Hurling obscenities at one another in very public fora may seem atrocious behavior, but within its context—in particular, within the 1960s—it is little different than a geological debate among Victorian gentlemen turning “topsy turvy without scruple” (Edward Turner, in Rudwick 1985:99–100), or Huxley publicly ridiculing a bishop, or Cuvier’s eulogy of Lamarck, so vitriolic it wasn’t even published until after Cuvier’s death, or (fill in the blank). Science is often a very agonistic process, and the style of the vituperation is the style of the age.
We might recall what Kenneth Burke said about the genius of the contest, the rhetorical energy that attends opposition: “When two opponents have been arguing, though the initial difference in their position may have been slight, they tend under the ‘dialectical pressure’ of their drama to become eventually at odds in everything” (Burke 1941:139).
And then there is this:
[Reading about the Linguistics Wars] reminded me of the sessions that my female colleagues and I would have in graduate school where we would only half jokingly take each other’s papers and globally replace what we thought of as “girl language” with what we characterized as “boy language.” “Therefore, it seems to me that” would become “Thus I have shown,” “I will give some evidence that” would become “I will prove that,” “supporting data” would become “conclusive empirical evidence” and “my suggestion” would become “my theory.” I always wondered whether those (of whatever gender) who wrote that way really believed their inflated claims, or were just adopting a certain style. (Speas 1998:532)4
It’s not that there were no significant female linguists contributing to the debates. Besides Robin Lakoff, Joan Bresnan, Janet Dean Fodor, Georgia Green, Judith Levi, and Barbara Hall Partee, for a significant sample, were all notable players in the drama—some of them as partisans, some as hopeful mediators—but they were markedly outnumbered, and other than Lakoff they were not really in agenda-setting roles (for reasons that surely implicate systemic sexism), and they weren’t connected with any notable breaches of decorum. “The entire field, especially during the 60s and 70s,” Geoffrey Galt Harpham has observed, “had the structure, and many of the social graces, of a boys’ club” (Harpham (1999b:220).
The partial exception to this pattern, of course, was the fun-loving Robin Tolmach Lakoff, whom Bucholtz and Hall (1995) argue with some justice has been slighted in histories of the period (mine included). She was instrumental in the beginnings of the movement, with an important Abstract-Syntax book on the cusp of Generative Semantics, including early work on the Performative hypothesis (1968), a consistent output of strong, influential papers, a defining role in the growth of pragmatics, and a distinct fostering role for the Midwestern second generation. Nor was she without her moments of distemper (notably, R. Lakoff 1970). She has herself even claimed, in terms consonant with Speas’s remarks, that academic machismo shaped the Linguistics Wars. She sees that ethos on one side alone, however. The “stereotypical feminine properties” of the Generative Semantics framework, she says in her memoir, did not fit easily into “the hierarchical, dichotomizing tendencies . . . characteristic of academic discourse and masculine world-view” that was deeply embedded in the Aspects program. That incompatibility, in turn, contributed to Generative Semantics’ fall from grace, unable to resist the more masculine, authoritarian Interpretivist program (R. Lakoff 1989:973–75). She has argued that the entire formalizing program in linguistics is misbegotten and patriarchal. “Language resists formality,” she says: “formal descriptions—especially of syntax—miss the point, although they make their creators feel both macho and scientific” (Davies 2010:370).
The socio-rhetorical tone of the dispute, in any case, was undeniably set by the guys; indeed, was largely continuous with the (nearly) all-guy rise of Transformational Generative Grammar. Lakoff certainly remembers the initial conditions in such terms. “The MIT Linguistics Dept. in the ‘60s was . . ., as academia was then, a very male place,” she recalls. “The kinds of things that got said and done to and about the few women in the group seem appalling now, but then [they] were just par for the course: you figured you had it coming to you, you had no business being there anyway” (Davies 210:369–70). Maybe when it is all said and done, much of the Linguistics Wars was just a boy’s-club pissing contest.
If so, we have seen that by the late 1970s there was no question whose stream was mightier. Generative Semantics petered out. Most Interpretivists and their sympathizers can tell you why—“It was disconfirmed”—and they are clearly right in at least one very narrow sense. Homogeneous I was disconfirmed. The Katz-Postal Principle did not fit the data churned up in and around the early Generative Semantics framework, and that’s what most Interpretivists point to when they say the program was disconfirmed. But that doesn’t get us very far, as the Interpretivists also recognize, because that’s when they start frothing about globality, a topic that prompts the opposite reaction from Generative Semanticists. Globality is their trigger word for pointing out the bankruptcy of Aspects and its various Chomsky-directed mutations.
And most Generative Semanticists have a very different, more interesting and more complete, explanation for the demise of their framework. In Levi’s terms, “it made promises it couldn’t keep.” Certainly, it had trouble living up to many of its specific promises. Ross, for instance, never took his performative analysis very far, and G. Lakoff pretty much abandoned global grammar as soon as he suggested it. There were numerous promised publications (largely by Ross and Lakoff, though sometimes involving Postal) which were routinely cited to bolster various points, but which never saw the light of the press. As one would expect, even this shows up in Generative Semantic self-mockery. One of Ross’s acknowledgments, for instance, goes “to George Lakoff, who and I ([improbably] forthcoming) are said to be writing a book about abstract syntax” (1974b: 123). Postal (1972a [1969]:168) cites “a no doubt never-to-be-written paper, Lakoff, Postal, and Ross (forthcoming).” McCawley, when he collected his early essays into Grammar and Meaning (1976b), changed many of the citations from the format “Lakoff (to appear a)” to “Lakoff (abortion a).” This inability to follow through on relatively minor promises—a few papers—was symptomatic of an inability to come through on the larger ones.
When Chomsky got himself into trouble in the mid-1960s by promising more semantics than he could deliver, he never actually retracted the promises, but he stopped making them and he stopped talking about semantics except in very narrow terms. And when one looks back at those promises, one can see something rather clearly that most observers at the time read right past. They were all hedged.
When the Generative Semanticists got into similar trouble in the early 1970s, they went the opposite direction. When the water got deep, they kept wading, further and further out, until they were dog-paddling a long way from shore, still shouting back bigger promises. Not only semantics, but now pragmatics, was going to fall into line. Not only would they handle the cognitive aspects of language, but now the social aspects. Not only would they account for discrete phenomena, but now fuzzy phenomena. Not only would linguistics be rehabilitated, but now logic.5 Eventually, their hats were floating.
Looking back, you may have felt, when you first saw those deep, deep, growing deeper, floor-to-ceiling trees of early Generative Semantics a bit like that anxiety-ridden fish in The Cat in the Hat:
But those deep, tall trees came with a whole array of handles, some of them new, some of them familiar—performative handles, logical handles, transformational handles—that linguists could use to get hold of them. By late Generative Semantics, the situation was very different. The deep, tall trees had become rather quaint as the movement grew so loose jointed and unwieldy that handles were hard to find. Even sympathetic linguists were left with nothing but a vast new view of the whole, messy, commingling, sociomental, signifying thingamajig, language. Generative Semantics ultimately failed for the reason most research programs fail. Scientists need something to do. They need handles. They need results. The motive forces that gave rise to Generative Semantics never went away, but two new programs emerged, Montague Grammar and Relational Grammar, more rigorous, with more manageable methods, asking more tractable questions, and soon many other approaches, some generative, some non-generative, now looked much more promising, some of them peeling away from Chomsky yet again (several new variants of Lexicalism, in particular). Too, Chomsky had engineered another motivation for linguists, restricting the grammar, and provided them with a new set of tools for satisfying that motivation. Postal, using a phrase he got from Katz, calls this the “busy hands” phenomenon: “a framework will be attractive independently of its substance insofar as it provides graduate students with a technically clear thing they can keep busy on and get theses out of” (Huck & Goldsmith 1995:138).
Language is a messy, commingling, sociomental, signifying thingamajig. Generative Semantics was right about that. Chomsky’s astonishing success in the decade after Syntactic Structures was linked rhetorically to the promise of mediating form and meaning, but it was anchored in the equally compelling distributional promise that this thingamajig could be modeled by precise, mathematically amenable formulae, in a coherent, systematic framework that approached the elegance and power and quality assurance of theories in the physical sciences. Generative Semantics tried to cash in both of these promises, the promise of meaning and the promise of formal modeling. But the promises proved incompatible, at least with the machinery that was available. Value hierarchies shifted. Priorities changed. Postal chose modeling over meaning (Relational Grammar). Ross and Robin Lakoff chose data fidelity over either. McCawley and George Lakoff tried to maintain them both, Lakoff for the most part giving the edge to meaning, McCawley to modeling.
What looked like a growing monolith in the mid-1960s, rising from the bedrock of Aspects of the Theory of Syntax, split in the late 1960s, then splintered in the 1970s and 1980s into an assortment of partially overlapping frameworks, clearly defining themselves in opposition to aspects of Aspects, all of them bearing significant marks of the Generative/Interpretive dispute and a range of intersecting, overlapping, or conflicting value commitments. Returning to Gazdar’s roll call from the last chapter of new grammatical countries for Generative-Semantic émigrés, but leaving aside Chinese cooking, we have three, Montague Grammar, Radical Pragmatics, and Relational Grammar (Gazdar 1979b:179). They were only the beginning.
By the mid-1980s, a veritable alphabet soup of competing frameworks had come into being.
Montague Grammar was among the first alternate models to arise, with a more rigorous and more credibly pedigreed semantics than either Generative or Interpretive Semantics could achieve and a syntax as good as either had produced (because, well, they had produced it; it was imported directly from Transformational Grammar). Richard Montague was a philosopher of logic and language, tragically killed in a still-unsolved 1971 murder, after a triptych of rich papers. The theoretical force of those papers is summed up in the title of one, “English as a Formal Language” (1970a), and the machinery it deployed had a lineage running through the Semantic Demi-gods, Frege, Carnap, and Tarski (the latter having supervised Montague’s dissertation and founded the approach Montague championed, model-theoretic semantics). Montague’s work was directly inspired by the unfulfilled promise he saw in Chomsky’s work, both for syntax, which had not achieved a sufficient level of “adequacy, mathematical precision, [or] elegance” (Montague 1970a:374) and for semantics, which he, along with many philosophers, saw as pretty bush-league stuff.
“Pissed off by” Chomsky’s work would be more accurate of Montague’s motivation than “inspired by” Chomsky’s work. Chomsky does have that effect. Montague complained about “the great sound and fury that nowadays issues from MIT about a ‘mathematical linguistics’ or ‘the new grammar’—a clamor not, to the best of my knowledge, accompanied by commensurate accomplishments.”6 As David Lewis, Montague’s philosophy hallmate at UCLA, put it, in an arrow at the heart of the Katz-n-Fodorian program, “semantics with no treatment of truth conditions is not semantics” (1970:18).
Montague’s work was crafted into an appealing approach for linguists through developments by his UCLA colleagues, principally Lewis and Barbara Partee. The name Montague Grammar—coined by Partee as a commemorative gesture—largely gave way to Montague Semantics in the 1980s, in part because the semantics proved quite portable, getting incorporated into other frameworks, and in part because it never developed the scope of other frameworks.7 Without a moderately robust syntax, however, Montague Grammar would likely not have survived in the 1970s ecosystem of formal linguistics, and Montague had considerably overestimated the syntactic adequacy of his own work. Partee at their head, a small group of linguists enriched the model. Partee drew together Montague’s and Lewis’s work, and amalgamated it with transformational syntax. Partee, if anything, was an Interpretivist, while Lewis “was quite sympathetic” to Generative Semantics (Partee 2015a:329), so the architects represented a kind of détente.
Chomsky had attracted a number of people with training in mathematics and philosophical logic to linguistics (including Partee), people who could appreciate the power of Montague’s work in ways that previous generations of linguists might not have been able. Montague Semantics grew rather quickly, with notable refugees from Generative Semantics (Lauri Karttunen, David Dowty, and Emmon Bach), along with Computational Linguists, independistes, and various philosophers of language and logic. Interpretivists showed relatively little interest; Chomsky let his disaffection and skepticism be known.8
Radical Pragmatics, getting back to Gazdar’s roll call, was not a model in the sense of Montague Grammar, Relational Grammar, or other frameworks with that surname. Rather, it was the tag for a loosely organized cabinet of investigations into the interfaces among semantics, logic, pragmatics, and syntax. At worst, it was just “a disparate collection of left-over problems that do not fit well anyplace else” (Stalnaker 1980:902); at best, it was a rich continuation of the rapprochement between philosophy of language and linguistics begun with Ross’s and Robin Lakoff’s performative investigations. Where Chomsky’s engagement with philosophers was mostly confrontational by the 1970s, Generative Semanticists enthusiastically borrowed and blended work, first by Austin and Searle, and rapidly thereafter, by such thinkers as H. P. Grice, Robert Stalnaker, and David Kaplan. Their respective motivations overlapped in central ways. Ordinary language philosophy developed in reaction to narrow truth-functional semantics. Generative Semantics reacted to narrow well-formedness syntax. In both cases, the reaction played to context and situation. In linguistics, this research meant exploding the competence/performance division. Methodologically, Generative Semanticists saw their mission here as pulling data out from under the rug of performance into the light of serious scrutiny.
Linguistic Pragmatics—as I will turn to calling it now, since Radical was a short-lived and ill-suited modifier—often intersected with Montague Semantics, and involved some of the same personnel. But it pushed on notions of grammaticality as much as it pushed on notions of meaning. The Lakoffs, both at the University of Michigan as pragmatic interests started to flower, urged this work forward, with Robin doing particularly ground-breaking research, and a clutch of Midwestern second-generation Generative Semanticists created the field.9 The big coming out party was in Austin, Texas, in 1973, known as the Performadillo Conference, followed almost immediately by the Pragmatics Workshop back in Michigan.
Linguistic Pragmatics was nourished by a host of affiliated concerns for discourse, for sociological factors like class and gender, for figural uses of language, for communicative function, politeness, felicity, and other factors implicating context, factors that involved data going well beyond isolated sentences like 1 and 2 to include utterances like 3 and 4:
With Sentence 3, we start to see a creep of rhetorical figures. It implicates the trope, synecdoche—a semantic shortcut in which a word for the whole entity or object (here, yourself and you) stands in for a part of that entity (here, the part identified in Postal’s relevant rule, euphemistic genital deletion). Neither Borkin nor her professors, nor Postal it seems, realized that these data had any relation to rhetoric or figuration.10 But the theoretical movement away from strict, literal denotation is very clear. The range of data was increasing. And context was inescapable in the trajectory of whatever might be the appropriate solution. For the synecdoche of 3 to go through, language users need a kind of encyclopedic account of the domain (anatomy), and of the social acceptability of anatomical references, a dimension of meaning that goes way, way beyond the entries of an Aspects lexicon or any known pronominal rule. Although no one noticed it, this data also calls for another sort of cognitive mechanism; for any synecdoche to go through, speakers and hearers need to execute a whole/part referential convertibility.
Sentence 3 exemplifies the sort of data that preoccupied many Generative Semanticists in the early 1970s. It concerns garden-variety distributional facts. People say such things all the time. Yourself is a meat-and-potatoes deictic, a reflexive pronoun coding only for person and co-referential with another word sharing the same referent (in this case, the preceding you). Yet here it is, doing something very peculiar. Its referential business has less to do with the first you than with some anatomical protrusion connected to the person identified by that you; the second you has the same referential complications. All of these referential linkages are tricky to a degree unheard of in Transformational Grammar, which is of course the whole point of Borkin’s research. She puts the data on display, pokes at it, reveals its complications for transformational mechanisms, and then confesses utter helplessness in dealing with it.
Borkin’s 3 also activates cognitive relations of a sort that had not even begun to occur to linguists, and of a far more general sort than the Chomskyan Universal Grammar program entertains. It depends on an elaborate semantico-pragmatic infrastructure (an infrastructure that one might call, somewhat quaintly, decorum) that had long been banned from linguistics. It involves not just (1) the referential denotation of its three pronouns or (2) their immediate syntactic environment, or even (3) their discourse environment—all three semantic relations are required, since we need to know there is a second-person addressee to understand the sentence, and that the second (yourself) and third (you’ll) pronouns continue that thread of identity; and what the discourse topic is (the addressee touching a part of his body for sexual gratification), yet there is no true “antecedent” within hailing distance of the second and third pronouns. But Sentence 3 also involves (4) the social unacceptability of using a different word altogether from a class of available synonyms (picking the clinical synonym, let’s identify that class by the word penis), and (5) the meronymic relationship between the relevant concept (call it penis) of that class and the addressee.
Whew! I don’t know about you, but my head hurts. Take two metatransderivational constraints and call me in the morning.
Conditions (1)–(5) don’t get us all the way home, of course, since we need additional anatomical knowledge to extract from the word limp that the addressee is male, we need some socio-biological facts about the non-desirability of lifelong penile limpness to infer that the speech act is cautionary in some way, and no doubt much else. But conditions (1)–(5) suffice to describe the tangle of immediate distributional facts about these pronouns that concern Borkin (and Postal before her). No wonder she threw up her hands in a kind of intoxicated despair. “Although I have dealt with these questions in a Generative Semantics framework,” she ends her paper, “the facts I have brought up present problems for any theory of grammar, and I know of no way in which any present theory would handle them” (Borkin 1972:44).
Sentence 4 is an even more fraught kettle of fish, at least on the surface, where it doesn’t even make sense—a weird complex consisting of a sentence tagged by a seemingly unrelated noun phrase. It’s bafflingly elusive, in desperate need of a grammaticality warning sign of some kind from McCawley’s stigmata palette, at least in isolation. But it becomes wholly unexceptional when seen as a response to 5.
“Both the syntactic well-formedness and the semantic interpretation [of 4],” in other words, “must be stated relative to the linguistic context” (Morgan 1973b:748). Context affects acceptability; and grammaticality, the defining notion of Chomskyan syntax, is at least obliquely related to acceptability. With 4 and 5 cheek-by-jowl, of course, one can simply account for 4 in transformational terms with an underlying structure like 6, some deletion operations that remove everything between the brackets, and a the-insertion rule of some kind:
Transformations can do these jobs easily enough; as we keep hearing, they’re powerful rules. But we also know that the only way to motivate the necessary derivation to get from 6 to 4 would be with some additional device (dare we say it?—a transderivational constraint) that has access to at least some aspects of the derivation of 5.
The motive behind the kind of data exemplified by apparent oddities like 3 and 4, and by the related argumentation, is to confront and overturn the Aspects orthodoxy. In Thomas Kuhn’s terms, the word is not oddities but anomalies, his label for data that strains the current paradigm, potentially to its breaking point: facts which “cannot, despite repeated effort, be aligned with professional expectation” (Kuhn 2012 [1962]:6). In retrospect, we can see that there has never been a period of research stable enough to fit Kuhn’s picture of “normal science” within the Chomskyan program, but that’s not how it looked to Transformationalists at the time, who saw Aspects of the Theory of Syntax as a scripture of linguistic orthodoxy. Within its framework, much of Generative Semantics enacted the first two stages of Kuhn’s anomaly-to-crisis-to-new-paradigm revolutionary pattern. Certainly, the lesson that many Generative Semanticists took from such data, leading them toward Linguistic Pragmatics, was that the derivational equipment of Transformational Grammar is hopelessly inadequate to account for language use (as, lest we forget, Chomsky had maintained all along with respect to Aspects).
Robin Lakoff was a major force in the emergence of Linguistic Pragmatics, both as a scholar and as a teacher. She pioneered politeness studies (when George said Chomsky can’t account for the use of the word please back in Chapter 4, it was Robin’s critique he had in mind), and virtually invented feminist sociolinguistics with her Language and Woman’s Place.11 So her view of the role pragmatic anomalies played in the crisis-infused last days of the Wars is particularly instructive:
It was at the time of the introduction of pragmatics into GS that the dispute between the two sides got most heated and bitter. Not because one side wanted to see linguistics as “scientific” and objective, and the other didn’t; but because both did, and neither could reconcile that desire satisfactorily with the data that were now turning up. Both felt frustrated with the inability of their own theories to deal with these facts fully; one turned away completely, and the other attempted to incorporate the problems within the domain of linguistic science. The internal frustrations created irritation—for which each side blamed the other. (R. Lakoff 1989:986)
Neither Radical/Linguistic Pragmatics nor Montague Grammar/Semantics were particularly susceptible to the alphabetic craze Newmeyer notes, but they are exceptions to the rule. From the 1960s through the 1970s the literature bristled with such uppercase coinages as TG, TGG, ST, GS, RG, and EST.12 My personal favorite is a name that Bach seems to have coined for some late EST advances, and which later showed up in serious books on Chomsky, the Revised Extended Standard Theory, as in “Give it a REST.”
There is no RP, and only the rare MG (though, perhaps in compensation, Partee introduced the convention of referring to Montague’s foundational papers alphabetically, EFL, UG, and PTQ13). But the Call of the Majuscules was just beginning. Before the end of the century, there was a riot of letters festooning the journals: LFG, DDG, WG, APG, GB, P&P, MP, just to single out a few of the more interesting theories—oh, and GPSG, which really tipped over the sort drawer, begetting HPSG, JPSG, KPSG, LPSG, ETCPSG, . . .14
These labels are mostly associated with one or two leading researchers, who tend to work only in that one particular framework, but Chomsky is directly affiliated with an astonishingly large brace of them (TG, TGG, ST, EST, and let’s throw in REST, up through the Wars period; thereafter, GB, P&P, and MP, all of which we will hear from in due time).15 He also has a welter of other such alphabetizations for individual rules, principles, criteria, or domains, completely suffusing the literature, but two new technical terms Chomsky proposed in the wake of the Wars calls for specific attention: I-Language and E-Language. They are yet further attempts to define which aspects of language he feels responsibility for, and which are outside his (and, in fact, science’s) scope. Competence became (sort of) I-Language, with the I for Internalized when he introduced it, later expanding to include Individual and Intentional as well (1986:22, 1995:13).
Performance, or at least the result of performance, or the set of all potential acts of performance, something like that, became E-Language, with the E for Externalized when first proposed, later broadened to include Extensional (1986:20, 1995:14), though perhaps Effluvia-Language might be more appropriate because E-Language is associated with Bloomfieldians, with Quine, and with others on the wrong side of linguistic history, as written by Chomskyans. E-Language “appears to have no status in the theory of language,” Chomsky says (1987:180); “virtually no empirical import,” he says (1987:181); “a marginal and derivative notion at best,” he says (1990:144); he especially uses E-Language as a focus for his rejection of corpus linguistics and the use of statistics. The upshot of this alphabetization subroutine, in any case, was that the correct characterization of the object of linguistic theory was I-Language. Almost any data one might associate with pragmatics, if you’re wondering, is Effluvia.
Alphabetically, though, the most defining letter for Chomsky was T, as in TG. The clearest emblem of separation for all of his programs from the other AGs (Alphabet Grammars), was that, whatever other modifications came along, his programs were all still Transformational Grammars. The transformation was seared in the minds and tripped off the tongues or keyboards of all observers and commentators who talked of Chomsky’s famed technical brilliance. The transformation was the flag that flew over his revolution. But, by the 1980s, no one else wanted it.
In virtually all current theories of generative grammar, much of the descriptive burden shouldered by transformations in Chomsky (1965) and subsequent work has been transferred to other components of the grammar, such as phrase structure rules, lexical rules, and interpretive rules.
—Steven Pinker (1982:665)
In the early 1970s, G. Lakoff imagined “such madness as . . . the elimination of transformations altogether,” but promptly rejected it: “This sounds wrong to me, so far as pure gut reaction is concerned” (G. Lakoff 1973a [1970]:451–52). Within the year, however, his gut had settled down and he pronounced, as one of Generative Semantics’ virtues, that “there are no transformations,” coming up with another innovation. “In their place there are correspondence rules,” he said, “which may have global and/or transderivational constraints associated with them” (G. Lakoff 1975 [1973]:283). In some version or other, correspondence rules became the order of the day in all the new grammars.
The single most important development in this trend, arguably, was actually internal to Chomsky’s program, a thesis by Joseph Emonds in the thick of the Wars, directed by Chomsky: “Root and Structure-Preserving Transformations” (Emonds 1970). Newmeyer puts it cheek-by-jowl alongside Jackendoff’s dissertation of the same vintage, as the two “most impressive pieces of scholarship” coming out of the MIT program in the period: “painstaking works devoted to detailed examination of particular rules in the context of defending the lexicalist alternative to the then predominant generative semantics” (Newmeyer 1996:57).
Mining restrictiveness veins, and taking explicit inspiration from Ross’s dissertation, Emonds provided a taxonomy of transformations in terms of restrictions to their descriptive power. By far the largest and certainly the central category was a class of rules that flipped the Katz-Postal Principle on its head. The Katz-Postal Principle was a meaning-preserving principle; do whatever you want with the structure, so long as you keep the meaning stable. Emonds’s relevant category was the structure-preserving transformations; change the meaning as called for, but keep the structure stable. By the mid-1980s it was widely accepted that “the structure-preserving constraint (‘SPC’) . . .[was] the most general principle available for restricting the derived structures that transformations can produce” (Emonds 1985:138). The SPC didn’t quite mean that transformations couldn’t alter structure, of course. Altering structure is pretty much all there is for transformations. No, the SPC means, rather, that transformations could only alter structure in ways already “allowed” by the phrase structure rules. A transformation could not alter a structure into something incompatible with those base rules. It could not create a substructure, or move anything into a position, that was not generated by the base component. Passive is a classic example: it moves the first Noun Phrase into a Prepositional Phrase in the Verb Phrase. Well, English Phrase Structure Rules define VPs as containing PPs. So, it’s OK.
As masterful as this proposal surely appears to you—in effect, the Phrase Structure Rules become one big filter for allowable surface structures—two flags should be going up now. First, a restriction against altering structures is anathema to Generative Semantics. Its bread and butter, going back to its Abstract Syntax roots, was in transformations that could rearrange remote and arcane structures into surface structures. So, the role of the principle in the Wars is readily apparent. But, second, lurking at the heart of Emonds’s principle, there is a poisoned pill for transformations. If transformations don’t give you anything you don’t already get with Phrase Structure Rules (and if transformations are so damn powerful, and if power is a bad thing), what do transformations buy us? Why not just let the Phrase Structure Rules do their thing? Transformations were so enmeshed with the whole idea of linguistics and generativity in the 1970s that this implication lay in the shadows throughout most of the decade, only emerging with the Alphabet Grammars of 1980s.
As Johnson and Postal describe it, the trajectory that defined linguistics from the late 1960s through the late 1970s was to correct a fatal flaw, a “deficiency built into the most basic assumptions of TG theory, namely, the fact that transformational applications destroy or ‘lose’ information, information which is necessary even after it is ‘lost’ ” (1980:4). The solution to this deep-seated lethal glitch, licensed in part by Emonds’s constraint, comes out of a simple argument: transformations entail derivations; derivations lose information; eliminating transformations eliminates derivations; therefore, eliminate transformations. Voilà, the information-loss problem is solved.
“But, really?” perhaps you’re wondering. “No transformations? What about those definitive rallying-cry arguments back in Chapter 2 that transformations were necessary for any adequate grammatical model?” Chomsky built his case for the necessity of transformations on the alleged weaknesses of (Bloomfieldianesque) Phrase Structure Grammars. They couldn’t handle the syntactic load themselves. They needed something extra. Chomsky’s something extra, inherited from Harris, was the transformation. The new grammars coming online in the 1970s and 1980s had other ideas.16
Relational Grammar—RG to its friends—was at the head of the class, and arguably contributed most to the splintering of the generative enterprise. RG had clear Generative Semantics ancestry—not just in the person of Postal but in work by G. Lakoff and Ross and Morgan, as well as by sympathetic and consonant researchers, like Gruber and Fillmore; and, not insignificantly, RG retained a stance of opposition to Chomsky. It was generative in the original sense of precise formalization and (especially in its abbreviated offspring, APG [Arc-Pair Grammar]; Johnson & Postal 1980), it may have been the most precisely and rigorously formalized generative model in a growing field of generative models); it took the universality mandate very seriously, building its arguments and developing its analyses on a wider and more diverse array of languages than was common at the time; and it had the full-bore formalist commitment to abstraction that fueled American linguistics from at least Bloomfield. Pullum has called RG “the last survivor of . . . abstract syntax” (1991:14), but it took a sharp right turn and adopted a markedly different commitment about the objectives of linguistic theory from recent programs, subordinating both mediational and distributional goals, and discarding psychological interests, in order to pursue a broader empirical question, “In what ways do natural languages differ, and in what ways are they all alike?” (Perlmutter 1980:195).
RG’s name gives away its defining innovation, that grammatical relations like subject and object belong directly among the core theoretical primitives of linguistics, a self-evident fact with respect to all pre-transformational approaches, but strikingly bold within the Chomskyan tradition. In all Transformational-Generative research, grammatical relations were secondary, derivative, held to follow naturally from structural properties. Aspects, in particular (Chomsky 1965 [1964]:71), defines subject as “[NP, S],” effectively, the first NP of a canonical Sentence (Postal in “Postal broke the transformation”); object as “[NP, VP],” the first NP of a canonical VP (the transformation in “Postal broke the transformation”). To define them otherwise, Chomsky says, in his inimitably sweeping way, would “merely confuse the issue . . . is mistaken . . . redundant . . . [a] fundamental error” (Chomsky 1965 [1964]:68–69). In fact, deriving grammatical relations was seen as one of great efficiencies of the Aspects model, one of the New-Grammar advances of precision over the loosey-goosey Old Grammar (Viertel 1964: 79–80).
But Generative Semantics research had begun to undermine these kinds of configurational definitions for grammatical relations and the Universal Base Hypothesis pulled toward “a fixed set of universal grammatical relations such as subject-of, [and] object-of” (Bach 1968:114). Gruber and especially Fillmore’s work on “deep case,” an inventory of semantic relations that cross-cut grammatical relations in evocative ways, was also influential. Important work by G. Lakoff and Ross, among others, increasingly put explanatory weight on grammatical relations. Lakoff later claimed that acknowledging the semantic dimensions of grammatical relations was a hallmark that distinguished Generative Semantics from other theories in the generative tradition, back to Harris (Lakoff 1987:464).17
Relational Grammar emerged from this climate, discarding almost the entire apparatus of transformations: no movement, no derivations, no cascade of representations for a single sentence. No matter how intricate the syntax of a sentence, it only needed one single tree (or “arc,” a related species of representation). But RG remained true to the compelling initial themes of Chomsky’s program—universality, formal precision, and explanatory adequacy, proposing a number of “Laws of Grammar”—while at the same time systematically exposing weaknesses of that program, and eliminating those pesky transformations.
Relational Grammar was not interested in different linguistic concerns from Aspects, as Linguistic Pragmatics largely was. It did not splice a new semantics onto a pre-existing syntax, as Montague Semantics (née Grammar) largely did. It was an arresting new model—Generative Grammar 2.0—and its successes were exhilarating, most dramatically demonstrated on the familiar proving grounds of active-passive relations (Perlmutter & Postal 1977).
Still, RG appears to have been too little and too late for the linguistic ecosystem of the late 1970s. The grammarians, they were restless.
One locus of discontent was to be found, shockingly, in Building 20, down the hall from Chomsky. Joan Bresnan, one of his Interpretivist-era students, who joined the faculty there, began working with a small group of collaborators—most prominently, Ronald Kaplan, a computational psycholinguist across town from her, at Harvard—and brought forth LFG. The L is for Lexical, signaling its allegiance to Chomsky’s Lexicalism. The F is for Functional, drawing this time more from the GS side of the ledger, since it coded for grammatical functions in LFG, what most everyone else called grammatical relations, like subject and object. The G? You don’t need a trig-equipped pocket calculator to figure that one out, the last name of virtually all linguistic models: Grammar.
At its heart, LFG was a purified Aspects model, with the same mediational goal of linking representations of form with representations of meaning—though with more prudence about meaning and greater modesty about overall implications—which raised one particular line of appeal in Aspects to Prime Directive status, what Bresnan called the “grammatical realization problem”: that is, “to specify the relation between the grammar and the model of language use [performance!] into which the grammar is to be incorporated” (Bresnan 1978:1). Chiefly, that meant psychological reality, but psychology in the Transformational-Generative tradition is in large measure conceived computationally, so LFG had a resolute eye on mathematical/computational tractability.
LFG also took the lead on a movement Newmeyer terms Superlexicalism (1986a: 187–91, et passim), a ramping up of Chomsky’s and Jackendoff’s shifting of transformational duties to the lexicon.18
The most impressive innovation of LFG, for most of its customers, was the way it accounted for the sorts of semantic and syntactic generalizations that Chomsky and Harris had treated as largely discrete operations that permuted algebraic symbols, into rule-governed attributes of word classes (for instance, that only transitive verbs can participate in passives; again, active-passive relations are the parade example). By directly implicating words, not just abstract category-variables, the approach immediately becomes more concrete, therefore more “realistic,” and by moving all the action “into” the lexicon, it responds to the restrictiveness issues that plagued syntactic transformations.
And there was another base-enrichment movement afoot, which we might call Superconstituentism.
It featured direct contributions by Geoffrey Pullum, Ewan Klein, Ivan Sag, and cognate or overlapping or otherwise influential work by Edward Keenan, Carl Pollard, Henry Thompson, and the ubiquitous Emmon Bach, among others (Gazdar & Pullum 1982:ii–iii), most of them affiliated at some degree of separation or another with Generative Semantics. But Gerald Gazdar was widely regarded as the lead engineer of the specific model that coalesced out this research, GPSG. Gazdar was so instrumental in the development of the model, and so prominent in its advancement, that he and Pullum christened the model with the unwieldy label, Generalized Phrase Structure Grammar, largely to cut short the “growing tendency for people to employ the unpleasantly alliterative (and attributionally inaccurate) name ‘Gazdar grammar’” (Gazdar & Pullum 1982:iii).
Gazdar was a late Generative-Semantics sympathizer and pragmatics pioneer in the 1970s who realized the rational conclusion of Chomsky’s arguments (and the always looming Peters-Ritchie results) against the disastrous expressive power of transformations: get rid of them, every last one. For Gazdar and associates, that meant going back to the demonstrably less powerful Phrase Structure Rules. GPSG was simultaneously regressive and progressive, a return and a rejuvenation, marking a sort of natural culmination of developments that followed the pre-Aspects trajectory of work by Lees, Klima, Katz, Postal, and Chomsky, who moved transformational powers into the Phrase Structure component.19 GPSG, like LFG, did the same basic lexicosyntactic work of Transformational Grammar, but with less machinery. And the semantics? GPSG happily incorporated Montague Semantics, adopting Bach’s (1976:2) rule-to-rule policy of pairing up all syntactic rules with a semantic rule; a policy, in fact, that we might call “the Montague mandate.”
The appeal of LFG and GPSG was tinged with nostalgia. Their “attempts to preserve certain attractive features of the earlier phases of generative grammar” (Wasow 1985:197) harkened back to the mathematical precision and formal rigor of early Generative grammar, to the “Three Models” and Logical Structures era. LFG additionally harkened back to the psychological promise of Chomsky’s late-1950s/early-1960s work with George Miller. Both grammars also manifest a clear rejection of Chomsky’s REST trajectory of maintaining transformations while keeping them in line with an increasingly elaborate system of regulatory mechanisms.
But let’s not miss that, even more fully, by rejecting transformations these models absolutely rejected the last clear formulation of Generative Semantics, a grammatical model that ran everything—semantic, syntactic, lexical—through an all-purpose transformational engine.
The main work of transformations was handled by these new models through different kinds of correspondence rules.
Chomsky’s arguments for transformations were largely arguments against an impoverished grammar, consisting solely of Phrase Structure Rules, that couldn’t handle crucial syntactic phenomena. It needed something extra. When that something extra was added, transformations, the impoverished “grammar” became a “base component” and transformations went to work hopping affixes around to account for discontinuous constituents, moving other constituents from spot to spot to account for phenomena like active/passive relations, and so on. But the 1980s base component was not your parents’ base component. All of the work building toward, and then building upon, Aspects of the Theory of Syntax, had resulted in a powerful lexicon, beefed up and tricked out Phrase Structure Rules, and various interface tools.
LFG makes especially elegant use of the lexicon, reimagining the old Passive transformation, for instance, into a lexical rule. GPSG makes good use of the lexical properties, too, but it developed ingenious “meta-rules.” Meta-rules don’t relate structures, as in a derivation. They relate the rules that generate those structures. To offer a crude reduction: one Phrase Structure Rule makes actives (call it PSR1), another makes passives (PSR2), and a meta-rule pairs them up (PSR1 ⇒ PSR2). The difference may seem subtle, but the generalizations come at a higher level, there are no derivations, and the “the gravest defect of the theory of transformational grammar” is avoided; namely, “its enormous latitude and descriptive power” (Chomsky 1972b [1969]:124–25).
There was ingenuity galore, a renewed sense of collective purpose, and lots of shared mechanisms drawn from both Interpretive and Generative Semantics legacies: primitive grammatical relations and Ross’s auxiliary analysis, from the one side, and lexical redundancy rules from the other, with x-syntax becoming the lingua franca of phrase structure.
Whatever direction Chomsky’s developments took in this period, and there were many, his models always flew the transformational banner. He was unmoved by all this activity to eliminate his signature device, at least outwardly, and his bellicosity was dialed relatively low. He took only a few scattered shots in public at the alternatives. GPSG, in a tired echo of his early Generative Semantics attacks, he dismissed as notationally equivalent to Transformational Grammar, except for its “needless complexity”—which, of course, made it inferior. GPSG was the same, but worse (Longuet-Higgins et al. (1981:[64f]). Meanwhile, in a virtual déjà vu of Ross’s experiences, Bresnan was one of Chomsky’s most talented students, who moved into a faculty position alongside him at MIT, resulting in some unpleasant episodes as their positions diverged.20
On the broader stage, levels of hostility among most practitioners of the new models on the linguistic landscape were decidedly lower than anything the Wars had seen. Frequently, they were downright congenial. The late 1970s, through the 1980s, was largely a period of healthy trans-border generative economies of methods and instruments. Even McCawley, who maintained his commitment to transformations and wholly eschewed the marketing of theories (“when you hear a linguist use the word ‘theory,’ ” he said, “you should put your hand on your wallet”—McCawley 1989a [1978]:83), cheerfully revised and refitted his approach on the basis of these developments. When he published his archly entitled Thirty Million Theories of Grammar, for instance, he renounced his arguments for deep Verb-Subject-Object order on the basis of the RG and GPSG dissociations of linear order and hierarchical structure, and, in consonance with RG and LFG, he embraced grammatical relations:
I now regard the case that I offered [1970b] for deep VSO word order in English as very weak because of my gratuitous assumption that there is a deep constituent order, . . . I now consider the VSO order that appears in structures I propose to be simply a makeshift way of indicating the grammatical relations between predicates and arguments. (McCawley 1982b:7)
For exactly the chance to fix bugs like these in his own linguistics, McCawley welcomed the alphabetic fecundity of the 1970s and 80s. “I regard the multiplicity of frameworks as highly desirable,” he said. “Any framework is going to lead its adherents to come up with facts that quite likely you wouldn’t come up with so easily in your own framework” (Cheng and Sybesma 1998:1).
The multiplicity of frameworks was remarkable, signaled most clearly by the vibrant 1979 Conference on Current Approaches to Syntax, sometimes just called The Milwaukee Conference but often referred to at the time as the Syntax Sweepstakes (Kac 1980, Moravcsik & Wirth 1980). GPSG and LFG apparently did not feel quite ready for prime time and went unrepresented. But DDG was there, and RG, and Montague Grammar (in point of fact, the presented model was labeled Montague Syntax), along with RRG, CORG, and EG, and a half dozen or so more.21 Nor was this the end of it. Yngve identified forty-two grammatical models by name in the mid-1980s, terminating the list with “among others” (Yngve 1986:108)—so, not quite McCawley’s thirty million. But a lot. The linguists plying these models largely appreciated one another’s work, cheerfully sharing axioms, borrowing innovations, and coöperating harmoniously. The only steadily visible resentment in the period was directed at Chomsky.
The rhetorical focus of this resentment was on what appeared to be growing inconsistencies in Chomsky’s work, especially in his adoption of positions he had previously denounced, and on Chomsky’s growing indifference to the technical machinery required by his own proposals. We enter a period in which “trashing Chomsky was the only reputable occupation for a self-respecting linguist” (Harpham 1999b:211). Michael Brame is a case in point. Not long removed from his (1976) EST attack-dog role against G. Lakoff and the Generative Semanticists, he railed not only about Chomsky’s stealthy adoption of many features from that program (a charge to which we will return), but about an even more egregious defect, Chomsky’s assumption of Generative Semantics’ nonchalant imprecision. He complained that the post-bellum Chomsky now regularly failed to formulate explicitly the rules sketchily evoked in his arguments. “Without explicit rules,” Brame says, targeting one particular claim, “it is difficult to envision what such an analysis would look like, and whether, in fact, it would even be formulable” (1978:210). Of another: “the formal account of the actual content of the words is taken to be a mere detail, a trait of Generative Semantics in its heyday” (1978:12). Of an entire subtheory:
As time now runs out on trace theory, one sees ever more far-fetched devices proposed to accommodate counterexamples that genuinely follow from more realistic approaches. Just as Generative Semanticists were inspired to propose global rules and other prophylactic devices to immunize their theory against refutation, so also trace theorists have begun to follow suit by adopting theoretical constructs which are seldom made explicit. (1978:13).
Gazdar and his GPSG collaborators went further, ritualistically excommunicating Chomsky from the church of which he was (or, rather, an earlier version of him was) still revered as the originating prophet. One would be hard pressed to find a more unequivocal charge of scientific impiety than the one they lay against Chomsky, setting his founding scripture against his contemporary blasphemies:
The term “generative grammar” is sometimes used [in the linguistic literature] as if it referred to (even solely to) contemporary work in Chomsky’s “Revised Extended Standard Theory” (REST) such as “Government-Binding” (GB; Chomsky 1981[a [1979]]) but not, for example, GPSG, Lexical-Functional Grammar (LFG; Bresnan 1982b), or Arc-Pair Grammar (APG; Johnson & Postal 1980). . . . It will be clear that our use of the term ‘generative grammar’ covers GPSG, LFG, APG, Montague Grammar in all its varieties, the work presented in Syntactic Structures (Chomsky 1957), Stockwell et al. 1973, Lasnik & Kupin 1977, and other work, but includes little of the research done under the rubric of the ‘Government Binding’ framework, since there are few signs of any commitment to the explicit specification of grammars or theoretical principles in this genre of linguistics. (Gazdar et al. 1985:6)
From the hostile outside—and the contingent of linguists who regarded the entire generative enterprise as a sham was growing—all generativists were contaminated:
The radical nature of variants of Chomsky-like linguistics is reflected by a particular attitude of their proponents: they tend to see their variant of Chomsky-like linguistics as the only “true” or “pure” representation of The Master’s views, often using the term “generative grammar” in a monopolistic fashion to denote this, and no other, variant. “I am not the one to fault,” they often claim, “it is The Master himself who has deviated from the straight and narrow road to linguistic salvation.” [Such representations only amount to “a Deceptive Detour” leading to a] crazily confused . . . corner of [the] detractor’s own pitiful little private property. (Botha 1987:8)
But—you’ve been with me a while now, so this will not shock you, I’m sure—it turns out that Chomsky didn’t need Brame’s blessing, or Gazdar’s, or Bresnan’s, or Botha’s, or [insert name here]. He lost no sleep over their assaults, and his framework lost no unrecoverable ground to the various alternatives. His program was sputtering, but it was still the benchmark; indeed, as his program alone was becoming plural, it might be more appropriate to say they were the benchmarks. His constraints-based framework of the 1970s, even with a much greater field of competitors than in the Wars, was the common measure for all of these models. Every theory defined itself first against Aspects, their shared stock, and then against assorted Chomskyan developments in the wake of Aspects (usually identified as EST, REST, or, sometimes Trace Theory). Even with several pre-1957 and alternate-tradition frameworks represented at the Milwaukee Conference, for instance, Giorgio Graffi rightly notes how thoroughly Chomskyan that event was, in terms of the themes addressed and even the suite of common sample sentences they referenced (2001:369–70).
With Lectures on Government and Binding (Chomsky 1981a [1979]), the game changed again.
While American linguistics was still somewhat fractious—not bitter, not particularly quarrelsome, but alternate models were jostling each other for market share in the linguistics bazaar—Chomsky’s (1973a [1971]) paper, “Conditions on Transformations,” caught some eyes in Europe. “Conditions” offered the first set of clearly articulated positive proposals since Aspects, and it entered circulation just as the looser and more tentative proposals of “Remarks” (Lexicalism and x-syntax) were flowering under the careful tending of Jackendoff and others. The ideas resonated in Paris and Pisa, and most strongly in the Netherlands, where Chomsky presented the work in 1971 (followed up by a compelling lecture by Emonds the next year). As we have seen, Generative Semantics had established itself widely in Europe, including the Netherlands. Pieter Seuren was a commanding theorist, functionalist Simon Dik was supportive, and young researchers like Henk van Riemsdijk and Jan Koster were enthused. But not for long:
[The Dutch] consensus about generative semantics broke down because students . . . became more and more skeptical about the ever wilder extensions of grammar, especially in the hands of George Lakoff (Broekhius et al. 2006:xix)
It didn’t hurt, either, that Brame had come through Utrecht with his anti-generative semantics harangue, with Chomsky’s and Emonds’s persuasive account of tractable alternatives tipping the balance decisively.
Yet again, revolution crackled in the Chomskyan air. There was a feeling of “a real crisis in linguistics” (Koster 1990:306). Van Riemskijk and Koster both spent sabbaticals in Mecca—Building 20, MIT—and returned to spearhead the formation of a new European group, GLOW (Generative Linguistics of the Old World). Their manifesto (written with Chomsky’s student, Jean-Roger Vergnaud) spoke in the now-familiar terms of a rising new epoch. In the light of later Chomskyan developments, the manifesto now looks to be “in retrospect a document of touching orthodoxy” for at least one of its authors (Broekhuis et al. 2006b:xix), but it carried the revolutionary banner to new territories:
The Chomskyan revolution has been a revolution of goals and interests as much as a revolution in method or content . . . In our opinion, generative linguistics acquired a new momentum in Europe after Chomsky’s “Conditions on transformations” (1973). This epoch making paper shifted the interest of linguistics from rather arbitrary rules to simple well-constrained rules operating under general common ground. (Koster, van Riemsdijk, & Vernaud 1978:5)
The first GLOW conference was in Amsterdam; at the second, in Pisa, Chomsky delivered Lectures, and with it, he once again pulled ahead in the Syntax Sweepstakes, shortly thereafter lapping the field. The Pisa lectures advocate a recognizably new program in familiarly paradoxical terms: reclaiming a revolutionary stance while simultaneously maintaining continuity. To this day, Chomsky can be seen tying himself in rhetorical knots by portraying the GB phase as an inevitable progression from his previous work and a radical shift in direction (e.g., 1982a:67, 2005:8, 2013:38). Its impact was even greater than the rapid growth of his Transformational-Generative program the first decade of Chomsky’s ascendancy, given how vastly the field of linguistics had expanded in the twenty-five years since he broke onto the intellectual scene. Lectures has now gone through seven editions; Syntactic Structures and Aspects of the Theory of Syntax, legendary pillars of Chomsky’s legacy, have gone through only two each.
A plausible assumption today is that the principles of language are fixed and innate, and that variation is restricted. . . . Each language, then, is (virtually) determined by a choice of values for lexical parameters: with one array of choices, we should be able to deduce Hungarian; with another, Yoruba. This principles-and-parameters approach offers a way to resolve a fundamental tension that arose at the very outset of generative grammar.
—Noam Chomsky (2000a:122)
He was speaking somewhat from the margins, advancing a quite specific grammatical model as other models were challenging or emerging, but Chomsky begins Lectures on Government and Binding in the well-earned tones of someone representing the entire field of generative grammar, with a chapter that chronicles its development as the natural emergence of an Aspects-era rule system, identified in the following way:
Subcomponents (i) and (iia) are effectively the Aspects base, generating “D-structures (deep structures) through insertion of lexical items into structures generated by (iia),” which are then turned by (iib) into “S-structures [Chomsky provides no gloss; give it your best shot] which are assigned PF- and LF-representations by components (iii) and (iv)” (1980a [1979]:5). The rules of component (iv), that is, are updates of the Semantic Interpretation Rules. This is the shape of Universal Grammar. The story continues.
After the inevitable emergence of this view of language, generative grammar came to the obvious realization that multiple systems of principles, each with its own “theory,” are responsible for the articulation of the grammar in any language:
The architecture of 7 and the specific machinery of 8 represent the culmination of generative grammar, we learn. But also, now that the field has reached this achievement, we approach a tipping point into a radical new paradigm.
The defining machinery in this conception is not, of course, transformational. It is the devices that restrict transformations—bound them, govern them, bind and control them. There is only a single transformation in the (iia) syntactic subcomponent, a highly general one, Move, a rule that does double and triple duty by also operating on phonological (PF) and semantic (LF) representations. It works indiscriminately to shift anything anywhere. Move is, on its own, just permutation run amok. But the elaborate system of shackles and constraints built up from the early 1960s prevents ill-formed sequences by precluding extraction sites, landing sites, or in-transit sites under the conditions stipulated in these various subtheories (8(i)–(vi)).
Bounding theory, for instance, is effectively a reformulation of the island constraints by way of movement traces. θ-theory (pronounced theta-theory) concerns the assignment of θ-roles, the GB version of Gruber’s thematic roles/Fillmore’s deep cases; things like, agent, theme, goal, and so on. It ensures the appropriate who-is-doing-what-to-whom semantic readings, but it also filters off ill-formed structures by precluding the ones that don’t match the specifications of the relevant verbs. Case theory assigns something called abstract Case (always capitalized), which may or may not be realized as morphological case. Many languages have widespread morphological case systems (German, Russian, Latin), but English isn’t one of them. We just have a little bit in our pronouns, but “case” is what makes 9 and 10 grammatical, in contrast with the ungrammatical 11 and 12.
GB has a Case Filter which sifts off 11 and 12, since Move would otherwise happily serve them up, but it also separates 13 and 14, branding the latter ungrammatical because with the Bordeaux blocks verb-to-NP Case assignment (present-to-Sonia):
It doesn’t matter that we can’t see the Case. It’s there—invisible and inaudible in the best abstract morpheme tradition, but there—and the Case Filter is on the job to exclude 14 from English.
Syntactic Structures-era grammar, in short, gets utterly flipped on its head in this model. There is no Passive Rule, no Affix-hopping, no construction-specific rules at all. Rather, all of the possible passives and all of the raised constructions would be generated, helter-skelter, by the promiscuous Move, along with every other conceivable permutation of words, without this elaborate assemblage of subsystems getting in the way.
Chomsky maintains almost total indifference to competing frameworks in Lectures. Aside from a few passing allusions to other models, his grammar—the natural, inevitable, descendant of Aspects—is just the only game in town. He notes, for instance, that “it would be unreasonable to incorporate . . . such notions as ‘subject of a sentence’ or other grammatical relations within the class of primitive notions” (Chomsky 1993 [1979]:10) without mention of Relational Grammar or its Generative Semantics antecedents, or the emerging Superlexicalist work—without even a clear sign as to why he might want to caution his readers about such unreasonableness—and then he gets back to his own proposals.
But, Dear Reader, there’s more. Not just a product of natural scientific development, his grammar is also radical. There “is something relatively new and quite important” overlaid on all this inevitability Chomsky says (1993 [1979]:4), representing “a substantial break from earlier generative grammar, or from the traditional grammar on which it was in part modeled” (Chomsky 1993 [1979]:7). This is where the notion of Principles and Parameters comes in.
Chomsky was not merely waving new banners in an old revolutionary parade. Lectures highlighted an electrifying coalescence in his program that rapidly charged the discipline. Yes. Again.
The stacking and nesting of labels around his work and Chomsky’s congenital shifting of focal range in his arguments blurs the picture somewhat. But, very roughly, the two main labels, Government-Binding (GB) and Principles and Parameters (P&P) take their names from mechanisms that drive in opposite directions; the first centrifugally, toward modeling individual languages, the second centripetally, toward the constrained variety of all languages; ultimately, toward the universal core of all languages. Chomsky’s program had seemed like a bit of a moving target through the 1970s, but the restrictiveness mandate had been consistent. Still hewing to that mandate, Lectures coupled a renewed sense of descriptive elegance (GB) with a distinct research agenda for charting Universal Grammar, along with intimations of direct genetic compatibility (P&P).
Some linguists were jaded, some were happy working with the new generative models, and a few were setting off with very different ideas of cognition. But some (Howard Lasnik, Joseph Emonds, Richard Kayne, Luigi Rizzi) had helped lay the groundwork for Lectures, and many were ready to be inspired. “In a tightly integrated theory with fairly rich internal structure,” Chomsky told them, riffing on an insight from Rizzi (1978) and sounding very much like his old Syntactic Structures, a-grammar-is-a-scientific-theory self,
change in a single parameter may have complex effects, with proliferating consequences in various parts of the grammar. Ideally, we hope to find that complexes of properties differentiating otherwise similar languages are reducible to a single parameter fixed in one way or another. (Chomsky 1993 [1979]:6)
A single parameter fixed one way could give you Italian, say, or Spanish; fixed the other way, English or French. Humans are born with a basic grammatical capacity, a “faculty of language” in a richly prepared but incomplete “initial state” that we can think of as
a fixed network connected to a switch box; the network is constituted of the principles of language, while the switches are the options to be determined by experience. When the switches are set one way, we have Swahili; when they are set another way, we have Japanese. Each possible human language is identified as a particular setting of the switches . . . we should be able literally to deduce Swahili from one choice of settings, Japanese from another, and so on through the languages that humans can acquire. (Chomsky 2000a:8)
The picture is not quite as tidy as this suggests, since the switch box image suggests one binary setting per “circuit” in the network. But the circuits of GB correspond to the various sub-theories—Bounding Theory, Government Theory, θ-Theory, and so on (that is, the components 8(i)–(vi) above)—which are not solitary principles, but systems of principles. So, we end up with something more like an array of dials rather than a single, master switchbox. But the metaphor, like many of Chomsky’s at key moments of theory shifting, is arresting.
The principles of P&P are universal properties, present in all languages, hallmarks of Chomskyan linguistics back to Ross’s islands of transformational resistance and beyond. But the parameters, the switches, were new. They determine the shape of specific languages. Together the Principles and the Parameters breathtakingly determine the categories of possible languages, explaining the snake bed of differences among all human languages that exist, have ever existed, ever will exist.
These mutually defining concepts are responsible for changes in the history of languages, for the differences between and within language families, responsible for the specifics of language acquisition, responsible for all the matters of stability and variation that characterize human language. It’s all just differences in settings. In acquisition, for instance, the merest environmental exposure is all that is required to trip the parametric switches, with a cascade of subtly intertwined linguistic properties falling into place. In diachronic change, pressures of some kind—isolation, contact, tribal power struggles—trip parameters, and we enter a new stage in the life of that language.
The opportunities were invigorating. Hands could really get busy. New vistas of research were open. New possibilities for scientific advancement were available, not just at home but in related disciplines as well, with linguistics taking the lead. Chomsky had moved the needle in psychology in the 1960s, over innateness, cognitive processes, and Deep Structure. Would it be molecular biology this time? The Government-Binding end of the program was making descriptive headway, giving credibility and cachet to the whole Chomskyan-generative enterprise. But the universalist Principles-and-Parameters was the revolutionary end.
The P&P parade example was Rizzi’s null subject parameter (sometimes known as the pro-drop parameter; pro-drop, for short), which, operating on a principle of Case Theory, supplies a clean dividing line for languages. Flip the switch up and you get all the null-subject languages, like Spanish and Italian. Flip it down, you get all the non-null-subject languages, like French and English. One principle + one parameter × two settings = one neat division. Null-subject languages on the left, non-null-subject languages on the right. The basic data looks like this (those es are empty categories, ‘dropped pronouns’ or unpronounced null subjects, the asterisks telling us they can’t be ‘dropped’ in English):
The a-examples are well-formed, grammatical sentences in Italian (the null-subject switch is up). Their direct correspondents in English, the b-examples, are ill-formed, ungrammatical non-sentences (switch down). (See Chomsky 1993 [1979]:253–54.)
Other parameters included the configurationality parameter, which divided fixed word-order languages (switch up) from free word-order languages (switch down) (Chomsky 1981 [1979]:133; Hale 1982 [1978]), the head-directionality parameter, to distinguish head-initial languages (in which head words, like nouns, occur before their complements; switch up) from head-final languages (heads after complements; switch down) (Smith 1989:69–71; Chomsky & Lasnik 1993:518), and the verb-second parameter for separating languages in which finite verbs are the second constituents in clauses (switch up) from languages in which they show up elsewhere (switch down).22
Much of the attraction was the elegance of a simple binary, on/off notion that linguists have always found compelling (think of those ± semantic and syntactic features, for instance; phonology has them too). Part of the attraction was the scope of explanation that followed from that simplicity. But a goodly part of the attraction was the Gyro-Gearloose joy and challenge of problem-solving science. The GB/P&P system is ramified and intricate. The work is hard, but the results are promising, and the reward potentially immense. The genetic implications were stunning: the great apparent surface variation of languages resulting from small blueprint changes at some deep assembly level, like eye-color, or blood type, or having a widow’s peak. The switch-box metaphor was especially appealing. One textbook even provided an illustration of the switch settings for various linear order patterns in terms of x-syntax parameters (Figure 8.1).
Figure 8.1 
-syntax parameter settings rendered as electrical switches: up gives one order, down gives another. Adapted from Carnie 2002:128.
Both the mechanisms and the goals of GB/P&P were rapidly influential, avidly championed by GLOW. For the first time in a decade, Chomsky was generationally persuasive, not just for his students but across the linguistic landscape. The timing could not have been worse for the alternate generative models that had developed in the wake of the Linguistics Wars. On the sidelines, the weather-beaten Fred Householder seemed rather pleased at the fractioning of generative grammar. “Any observer of the American linguistic science during the past few years,” he wrote in 1978, “cannot help noticing that the natives are restless. . . . So many people are looking for a Moses, at least, to lead them out of bondage, if not a true Messiah” (Householder 1978:170, 171). What he failed to notice was that the old Messiah was still in business, and new disciples were gathering. A reviewer of two far-reaching Relational Grammar volumes that appeared in the mid 1980s, for instance, was compelled to observe that
had these volumes appeared, say, five years ago, then I think they would have aroused great interest. My suspicion is that in the present climate they will go largely (and unjustifiably) ignored. In part, this is due to recent changes in orientation in syntactic theory: government and binding syntax, following on from the publication of Chomsky (1981) [Lectures on Government and Binding], seems to have caught on as the dominant formal syntactic theory (dominant in the sense that it is more dominant than any other single theory), and certainly it is government and binding, rather than any other model (including relational grammar), that seems to be attracting the best minds among the upcoming generation of formal syntacticians. (Comrie 1986:774)
The reviewer spells out the implications for Relational Grammar somewhat reluctantly, given the lack of fresh blood (“it could turn out that relational grammar has already provided us with the major insights into language that it is capable of providing,” 788), and history has largely proven him right on that score.23 Work in RG had tailed off; by the 1990s, it was effectively extinct. But reviews of the major LFG and GPSG books which came out at roughly the same time as these volumes (Bresnan 1982a, Gazdar et al. 1985), might have offered similar conclusions, at least about the growing ascendancy of GB.
It’s not that other frameworks from this period no longer have any adherents. One of the offspring of GPSG, for instance, HPSG, is quite successful, and for a time was the most widely deployed grammar in computational linguistics. LFG, too, has a significant presence in computational linguistics (and for much the same reason, careful formalization). It also has several major publications in this century (Bresnan 2001, Dalrymple 2001, Kroeger 2004, Dalyrimple et al. 2019), a broad contingent of practitioners working on numerous languages, and a big annual conference. And the best half of Montague Grammar lives on as Montague Semantics.
But there was something else coming online at the same time as Government-Binding Theory that would have a much broader impact than any of these frameworks. It had a minor airing at the Milwaukee conference, where it was voted Theory Least Likely to Succeed.
The proponent of that approach could not show up. He was ill. The scheduled discussant, a friend of the proponent’s, had to present the model in his stead, with some notes but no manuscript, and the discussant’s role was taken up by someone else, a colleague drafted at the last minute. The best days of proponent and his friend were thought by some to be behind them. No manuscript was in fact ever produced, so the approach had no presence in the resulting volume (Moravcsik & Wirth 1980). The proponent was George Lakoff, the friend was Haj Ross, the last-minute replacement for Ross was John Lawler—Generative Semanticists all, past tense—and the framework carried the label Cognitive Grammar.24 The tangled presentation circumstances led to one of the weirder titles in the annals of discussion papers: John Lawler’s “Remarks on [J. Ross on [G. Lakoff on Cognitive Grammar [and Metaphors]]]” (Lawler 1980). It was Lawler who voted this inauspiciously debuted model Least Likely to Succeed.
In 1978, I discovered that metaphor was not a minor kind of trope used in poetry, but rather a fundamental mechanism of mind.
An innocuous-seeming little monograph showed up the year that Lakoff could not give his talk in Milwaukee. The book, by Lakoff and a young philosopher, Mark Johnson, has a distinctly late-Generative-Semantics flavor. It is chock-full of data and redolent with Lakovian grand pronouncements. But it plots a very different course from anything predictable from the beginnings of that movement; from, say, Irregularity in Syntax. Chomsky had this work in mind when he told me, with his familiar casual dismissiveness, “Lakoff simply abandoned the study of language.”
Quite the contrary. Metaphors We Live By has expanded the study of language in richly traditional but also innovative ways, bringing a new clarity to figurative dimensions of language informed by distributionalist argumentation. The three traditional language sciences, the classic and medieval “trivium”—grammar, logic, and rhetoric—have severed and recombined in various ways over the millennia. Zellig Harris brought grammar back together with logic, with very notable downstream effects we have looked at. Lakoff and Johnson’s program brings linguistics and rhetoric back into alignment for the first time in over a century (while also continuing to implicate logic).
Curiously, however, the alignment of linguistics and rhetoric is not on their agenda at all, and their presentation is actively hostile to the rhetorical tradition. Chomsky’s via negativa argumentation style is also Lakoff’s.
Nor does Lakoff and Johnson’s book even concern metaphors, properly understood. It focuses rather on analogic lexical clusters, like the cluster illustrated in this group of examples:
Or, take this group of examples:
Sentences 17–21 manifest the specific cluster Lakoff and Johnson label argument is war (1980 [1979]:4); 22–26 manifest time is money (1980 [1979]:8). (The small-cap naming convention for such clusters is theirs.) They call these phenomena, influentially but erroneously, Conceptual Metaphors.
The clusterings in 17–26 exhibit vocabularies (italicized) that are referentially “out of place,” revealing a subsurface merger of the semantic domains of arguing and warfare on the one hand (17–21), time and economics on the other (22–26). When people talk about argumentation, they often draw on words of warfare, employing a domain-merger “we are hardly ever conscious of” (Lakoff & Johnson 1980 [1979]:5); Lakoff, no doubt, had heard and used lots of such vocabulary over the course of the Linguistics Wars himself. The 22–26 clustering exhibits the same kind of subterranean lexical assimilation, from the management of capital to the management of time. People talk about time with an economic vocabulary.
“Argument is war” is a metaphor, albeit a flaccid one. “Time is money” is a metaphor, so flaccid as to have become a cliché. The names of these clusters are metaphors. But the clusters themselves are not metaphors, and calling them so only confuses the cognitive affinity we have for analogy and comparison with the linguistic reflexes of that affinity (that is, analogic clusters of this sort, but also similes, conceits, allegories, and the whole welter of cross-domain mappings that populate human language; yes, centrally including genuine metaphors).
“The most important claim we have made so far,” they proclaim at the crescendo of chapter one,
is that metaphor is not just a matter of language, that is, of mere words. We shall argue that, on the contrary, human thought processes are largely metaphorical. This is what we mean when we say that the human conceptual system is metaphorically structured and defined. Metaphors as linguistic expressions are possible precisely because there are metaphors in a person’s conceptual system. (1980 [1979]:6)
At what or at whom their “on the contrary” is aimed is difficult to know. Association psychology had long insisted that a “comparative” or “analogic” or “similitudinous” affinity shapes human thought and language, back at least as far as Aristotle (Of Memory 451b), and this affinity has been consistently linked to metaphors and other figures throughout history, as well as genres and argumentation styles.
Lakoff and Johnson’s scholarship on this matter is disgracefully negligent (with the greatest guilt clearly falling on Lakoff, by far the senior scholar and the lead author). They are blitheringly indifferent to millennia of prior research, and feign ignorance of the more recent interdisciplinary obsession over metaphor that was cresting in the 1970s.
Beyond some dismissive remarks about the pervasive errors of “most people” for thinking metaphor is merely “a device of the poetic imagination and the rhetorical flourish” (Lakoff & Johnson 1980 [1979]:3), they do not so much as nod at generations upon generations of scholarship in rhetoric and grammar (what Lakoff would later collapse under the label, “the old theory,” 2008:252), a tradition which prominently attended to metaphorical and metonymical extensions. One particularly relevant voice, for instance, is Quintilian—as closely associated with “the old theory” as any one might be—who cites the following set of expressions used about argumentation: “to fight firm; to aim at the throat, and, to draw blood” (Institutio 8.6.51). Quintilian was not alone. The history of noting analogic lexical clusters in ordinary language is long, including such prominent thinkers as Cicero, Vico, Peacham, and Dumarsais. Add to that “the whole bonfire of metaphor [studies]” in lexical semantics from the nineteenth and early twentieth centuries, including Gustav Gerber, whom Nerlich and Clarke call “the Lakoff of his generation”! (2001:41, 43). Lakoff and Johnson also somehow miss the burgeoning metaphor research in philosophy, psychology, and English studies that was going on all around them, for the two or three decades preceding publication of Metaphors We Live By. One observer was willing to “wager a good deal that the year 1977 produced more titles [on metaphor] than the entire history of thought before 1940” (Booth 1978:49).
In fact, even a mainstream Chomskyan textbook of the period notes the pervasiveness and productivity of analogic thought in ordinary language, serving up data like this (using the same italicizing typography that Lakoff and Johnson would shortly adopt):
Packaged up with five more similar sentences, the authors remark that these examples illustrate “an interesting case [of] the metaphorical extension of words from the physical realm of food and digestion into the mental realm of ideas” (Akmajian, Demers, & Harnish 1979:114–15).25
Figurative language had little presence in American linguistics, especially generative linguistics, until it started to seep in—largely unrecognized by the linguists themselves, so long had the discipline been estranged—via the work of Postal, Borkin, Morgan, and others. But that these two guys in particular would utterly neglect such an extensive body of recent research—Johnson, a young philosopher fresh off a dissertation on the intersection of cognition and metaphor, directed by Paul Ricoeur;26 Lakoff, who majored in English studies and whose first linguistics course was with Roman Jakobson;27 both of them having keen interests in psychology—is extraordinary, except for the shabbily obvious; namely, that it contributes to their self-portrait as unanticipated prophets of metaphor (“We have found, on the contrary, . . .”).
What they have “found,” on the contrary, is not metaphor. Quintilian calls this kind of cross-domain lexical merger allegory—using his data, as Lakoff and Johnson use theirs, to exemplify how figuration operates in “[our] most common understandings, and our daily conversation” (Institutio 8.6.51). Lakoff and Johnson’s data, like Quintilian’s, illustrate submerged analogic frames, activities from one domain expressed through a vocabulary entrenched (“literally”) in another domain. The individual lexical reflexes (defend, attack, fight firm, draw blood, and so on, in the argumentation:battle frame; waste, invest, and so on, in the temporal:economic frame) are metaphorisms, as Michael Reddy called them in the paper that seems to have sparked Lakoff and Johnson (Reddy 1979:299)—small, vaguely metaphorical thingies.
Lakoff and Johnson exaggerate their own accomplishment and massively denigrate the accomplishments of others, of whom they are either irresponsibly ignorant or unethically ignoring. As scholarship, it is disreputable, but as academic imperialism, it is brilliant. It also has more in common with some of Chomsky’s practices than Lakoff would ever comfortably admit.
But—and I do not want to obscure this point by justified complaints (if you want to call them sanctimonious I would not object) of error and omission—Lakoff and Johnson’s work is both important and deservedly influential, initiating the most significant research program into the cognitive and noetic entanglements of thought and language in the current century. It is a substantial advance over previous scholarship going well beyond earlier research into analogic lexical clusters. In particular, they provide a principled taxonomy of such clusters and thoroughly chart their respective and collective implications. The examples above (17–26), for instance, are Structural Analogic Frames because the correspondences impose structural relations in target domains on the basis of relations from a source domain (time in terms of commodity transactions, for instance). The other two taxa are Orientational Analogic Frames like 30–35, where the vocabulary of space and directionality is used for concepts that “literally” have neither, and Ontological Analogic Frames like 36–41, where a vocabulary of solidity and substance is used for abstract concepts (traditionally, these are called reifications).
The extent of such analogical terminology is remarkable, but correlational frames (aka “conceptual metonymy”28) are equally extensive. Metonymy depends not on correspondences but on correlation—taking a hoary example, “The pen is mightier than the sword,” conveys its meaning because pens correlate with language (therefore, argumentation, diplomacy), swords with hostility (militarism, warfare)—and there are systematic lexical patterns that frame one class of concepts in terms of a correlated class. Most immediately, for instance, the citational style of this book, which takes the form of Chomsky (1957) and Lakoff and Johnson (1980), identifies products by the names of their producers (books by the names of their authors), and further differentiates those products with the time of a significantly related event (namely, the year of publication). Here are some further examples of these correlational frames:
Many aspects of Lakoff and Johnson’s work had been long understood, especially the utter pervasiveness of figurative vestiges in language, the use of concrete terminology for abstractions, and the tendency to favor certain domains in that terminology (humans, bodies, locations). But, by bringing the standards of evidence, methodological systematicity, and conceptual tools of contemporary linguistics to the task, Lakoff and Johnson—and the many linguists and cognitive scientists they have inspired—have raised the level of understanding for these phenomena and, in the bargain, raised a central pillar of a new framework that was coalescing at the time.
On the basis of that pillar-raising, Metaphors We Live By signals the clearest transmutation of Generative Semantics into a new and vibrant framework. That book certainly has earmarks of the movement. “It would not be far off the mark to call [Metaphors We Live By] a Generative Semantics book,” John Lawler noted:
Aside from the fact that Lakoff was a leading figure in GS, the book bases its argumentation firmly on the generative potential of metaphor, which is clearly semantic in nature. The wealth of examples provided is also typical of GS at its height, as is the rejection of traditional limitations on linguistic analysis. (Lawler 1983:203n5)
And then there was this, overlooked or set aside by Lawler: Metaphors We Live By is utterly contra-Chomskyan, pitching itself wholly against the line of thought “epitomized by the linguistics of Noam Chomsky, who has steadfastly maintained that grammar is a matter of pure form, independent of meaning or human understanding” (Lakoff & Johnson 1980 [1979]:205).
In the end, though, Lawler did allow for one deviation in Metaphors from the Generative Semantics program, “There are, however,” he noted sadly, “no puns.”
Pun-attrition aside, Lakoff is the horseman who has always been most preoccupied with paradigm-formation, and he had been widely test-marketing a variety of cognitivist ideas to support a new framework in linguistics (prototypes, amalgams, fuzziness, gestalts). None of them really stuck until his work with Johnson, and, as Lakoff later noted, the implications of what he calls metaphor marked a crucial turning point in the late 1970s:
In 1975, I became acquainted with certain basic results from the various cognitive sciences pointing toward an embodied theory of mind—the neurophysiology of color vision, prototypes and basic-level categories, Talmy’s work on spatial relations concepts, and Fillmore’s frame semantics. These results convinced me that the entire thrust of research in generative linguistics and formal logic was hopeless. I set about, along with Len Talmy, Ron Langacker, and Gilles Fauconnier, to form a new linguistics—one compatible with research in cognitive science and neuroscience. It is called Cognitive Linguistics, and it’s a thriving scientific enterprise. (G. Lakoff 1999)
Returning to Lawler’s comments on Lakoff’s non-paper at the Milwaukee Conference, Lawler did not hold out very much hope for Lakoff’s latest approach. He called Cognitive Grammar “dead” and “stillborn” (Lawler 1980:54), “only a steppingstone in a history of theoretical research which proceeds through ‘Linguistic Gestalts’ (Lakoff 1977[a]) [to experiential linguistics]” (Lawler 1980:51). Lawler’s asides on Generative Semantics were not hopeful either; it was a near-extinct ancestor of Cognitive Grammar, he suggested, with all the utility for the linguistics of 1980 that Etruscan had for dinnertime conversation.
That characterization was largely true through the 1980s and into the 1990s, but by the turn of the twenty-first century, the general amnesia about Generative Semantics, punctuated occasionally by a casual dismissal, had begun to lift somewhat, giving way to a range of acknowledgments that Generative Semantics was an experiment of considerable worth—especially as many of its proposals began to surface in work championed by Chomsky; as the Minimalist Program emerged, with an architecture eerily reminiscent of Homogeneous I; and as Cognitive Linguistics, a constellation of non- and anti-Chomskyan developments with Generative Semantics pedigrees coalesced into a formidable framework, formidable in terms both of coherence and adherence. If the appropriate image for Brame’s (1976) assault was shoveling dirt over the remains of Generative Semantics, the one to mark the twenty-first century shift in perspective might be the iconic horror-movie poster where a hand rises from the grave. And . . . wait a minute! . . . that’s not just any hand. That’s George Lakoff’s hand!