The Antisymmetry of Syntax

by Richard S. Kayne

It is standardly assumed that Universal Grammar (UG) allows a given hierarchical representation to be associated with more than one linear order. This book proposes a restrictive theory of word order and phrase structure that denies this assumption. According to this theory, phrase structure always completely determines linear order, so that if two phrases differ in linear order, they must also differ in hierarchical structure. It is standardly assumed that Universal Grammar (UG) allows a given hierarchical representation to be associated with more than one linear order. For example, English and Japanese phrases consisting of a verb and its complement are thought of as symmetrical to one another, differing only in linear order. The Antisymmetry of Syntax proposes a restrictive theory of word order and phrase structure that denies this assumption. According to this theory, phrase structure always completely determines linear order, so that if two phrases differ in linear order, they must also differ in hierarchical structure. More specifically, Richard Kayne shows that asymmetric c-command invariably maps into linear precedence. From this follows, with few further hypotheses, a highly specific theory of word order in that complement positions must always follow their associated head, and that specifiers and adjoined elements must always precede the phrase that they are sister to. A further result is that standard X-bar theory is not a primitive component of UG. Rather, X-bar theory expresses a set of antisymmetric properties of phrase structure. This antisymmetry is inherited from the more basic antisymmetry of linear order. Linguistic Inquiry Monograph No. 25

Genres: Linguistics, Reference

175 pages, Paperback

First published December 14, 1994

Reviews

[Gary Bruff · Rating 3 out of 5]

In principle, any infant could be transported to a foreign clime and would learn to speak Achenese, Chinese, Chamorro, Maori, or whatever language they are nurtured by, and learn to speak it as a native. This seems to say two things: a child is not genetically predisposed to learn any specific language; and all languages are learnable because there is something in our brains formally unifying all languages. Kayne’s The Antisymmetry of Syntax begins with an exploration of how an optimally constrained meta-language might could reduce structural ambiguity to a minimum and thereby maximize a language’s learnability. Kayne argues, for example, that all branching nodes in a syntax must be binary, because this simplifies the representation of a sentence in a way that makes a sentence structure comprehensible to the learner. Allowing both binary and ternary branching would make a sentence’s linear structure indeterminate and therefore not learnable.

Kayne’s work falls under the minimalist adventures of Chomsky and his loyal fellow travelers. Under minimalism, syntactic facts must still be explained entirely within the context of other syntactic facts. This system building enterprise has no interface for social or cultural influences on utterance choice. In fact, the rhetorical dimension of grammar has been excised and thrown to the hounds. But there is a certain satisfaction, a rhetorical satisfaction, in seeing an ersatz mysterious structure through the lens of a particular formal and well-developed theory. Here in the land of minimalism, the seemingly chaotic flux of syntax is domesticated, made to follow a body of meta-rules for syntactic representations and transformations.

These meta-rules serve to abrogate syntactic diversity while at the same time promulgating certain stipulations about how syntax can be represented. Consider Kayne’s LCA (Linear Correspondence Axiom), where asymmetrical c-command maps to linear precedence as all terminal nodes are pairwise compared in such a way that the overall ordering of a sentence can be determined (for the linguist and for the language learner.) This LCA is highly constraining, logically requiring a universal ordering of Specifier-Head-Complement. And Kayne does mean universal. Languages like Japanese, generally assumed to be Complement-Head, are really just like English, French, and Italian, except that the complement is moved over the head. We know that the complement raises, and the head does not lower, because all movement, in every language ever anywhere, is leftward. What appears to be rightward movement of X is really the stranding of X after its neighboring material is moved up the tree. Kayne at one point offers a kind of proof of how, when we hear a sentence as it is being spoken, we begin the parse word by word, affix by affix, clitic by clitic as we move from the left edge downward through the tree as it is being generated in real time. Under standard generativist assumptions, traces are bound by a higher antecedent, and the foot of the movement-chain starts out invariably lower in the tree before moving in local steps up to an argument or non-argument position at the head of the chain. I think this might be a serious problem for localism. If, for example, we can’t interpret a variable’s theta role until we hit the foot of the chain (often the silent gap at the end), then are we not really talking about global information marked in a clause from head to toe and back again? A related problem with Kayne’s approach is that non-configurational languages are legislated out of existence by means of these highly constraining meta-rules. There are some locality conditions in Latin, but not many. PROarb to assume that all languages are configurational, strictu sensu, amounts to buying in whole hog to what I have labeled the UEH (Underlyingly English Hypothesis).

Nevertheless, a work like Kayne’s Antisymmetry of Syntax lends shape to some often nebulous domains of syntax. Meta-rules or not, binary branching is now orthodox, as are perhaps universal leftward movements from a specifier-head-complement base. For Kayne is consistent. Under his system, the planet’s many thousands of languages all display a tenor of universality in their syntactic categories and combinations. In consistently narrowing the gamut of syntactic representations through meta-rules, Kayne limits the vocabulary used in the description of syntactic phenomena and their operations. This in essence is what minimalism is all about, as art, architecture, or as syntax. Less is more.

Also congruent with minimalist economy, there should be no construction-specific rules nor language-specific constructions. If you have a syntactic gadget, the rhetoric and aesthetics of the field insist that this gadget do more than one thing and be motivated by cross-linguistic comparisons. This appears to be a necessary step (in part because it creates a common vocabulary for doing syntax) to deriving the minimal representation of Universal Grammar, but it has little empirical justification. Indeed, it is not necessary for an aspect of grammar to be generalizable to other constructions or other languages. Syntax must come to grips with the reality that (a particular) language can be mightily uneconomical and that speakers can in fact introduce baroque adornments of syntax, marking their syntax in often surprising ways, pace minimalism.

Just for yacks, I am going to try and think about three syntactic phenomena and how they might be better (mis-)understood in terms of antisymmetry: (a) Groucho sentences, (b) Mandarin ‘de’ complementizers, and (c) the Latin ‘-que’ conjunction.

(a) There are some sentences attributed to Groucho Marx which demonstrate structural ambiguity. A sentence like...

I shot an elephant in my pajamas.

...is funny because Groucho follows with...

How he got in my pajamas, I’ll never know.

Like the double entendre of a pun, the second, disambiguating sentence forces us to redraw the tree of the first sentence in our heads, shifting ‘in my pajamas’ from the normal reading of adjunct to ‘shot...’ to the comic reading of adjunct to just ‘an elephant’. The difference in meaning is between how the agent shot (shot it while wearing pajamas) and what is attributable to the theme itself (a pajama-wearing elephant). Both readings have the same surface structure, but if vacuous movement is entailed, how do the pajamas get around the elephant in the verb adjunct reading? Is it natural to derive that reading from:

I shot [an elephant] in my pajamas t.

This appears to violate the theta criterion and so makes an impossible chain. Adding a functional head to give the tree some more governors or greater asymmetry might in fact complicate the representation more than clarify the derivation.

(b) Mandarin has a complementizer ‘de’ that heads relative clauses. It occurs after the subordinated clause and before the modified constituent. Kayne sees a unification of sentence final question particles on one hand and clause final relative pronouns on the other. In Mandarin, the toneless clitic ‘de’ stays in place while the clause is rotated around the clitic in leftward movement stranding ‘de’ at the end. Such movement parallels that of +Q ‘ma’ (another toneless clitic), which occurs sentence finally in Mandarin. Underlyingly and initially, ma or de actually occur at the beginning and the entire clause/sentence is revolved around it, stranding it to the right. In Kayne’s system, these four Mandarin sentences:

ni gei tamen de qian.
(You give them de money → the money (that) you gave them)

na shi wo yinggai zuo de.
(That is I ought do de → that is what I ought to do.)

zhe shi ni bu neng wangji de shir.
(This is you not able forget de thing/business →
This is something you cannot forget.)

ni chi le ma?
(you eat ASP ma → Have you eaten?)

...are derived from these deep structures::

*qian de ni gei tamen (= money that you gave them)

*na shi de wo yinggai zuo (= that is what I ought to do)

*zhe shi shir de ni bu neng wangji (= this is business that you not can forget)

*ma ni chi le? ( = +Q you eat ASP)

Not suprisingly, the underlying order is virtually identical to English surface structure (The initial ‘ma’ coinciding with a +Q COMP dominating the matrix clause in English.) But these underlying sentences are barbaric Mandarin, not just wrong. It seems that the meta-rules prohibiting complementizers from being adjoined to the right has a very un-economical reflex of massive and otherwise unmotivated movement leftward. Certain minimalist bylaws are, it would appear, fairly lawless in their application to differing languages. It seems reasonable to me that both the left and the right periphery of the sentence are potential targets for functional material dominating focused constituents. Even in English there is a sentence final intonation contour for questions which functions like the Mandarin +Q ‘ma’. Must we flip the whole English clause around that?

(c) As a final example, consider the Latin ‘-que’ clitic. Like the free standing ‘et’ it is a conjunction in some ways equivalent to ‘and.’ Here are some examples of ‘-que’:

arma virumque cano.
(Weapons.acc man.acc-que sing-1s → I sing of weapons and a man.)

senatus populusque romanus.
(senate.nom people.nom-que roman.nom →
The Roman Senate and people.)

terra marique
(land.ablative sea.ablitive-que → On land or sea.)

According to Oniga, a latter-day grammarian of Latin, the -que clitic is the head of the conjoined expression. The clitic is stranded on the right after the complement of the conjunction moves over the clitic-head, giving the clitic its host. Oniga says this conjunction only joins together NPs which form a unified concept or which form a pair of contrasting ideas. Leftward movement of the second element does conform to Kayne’s expectations of the conjunction’s complement moving over the head to its surface position before the conjunction. What’s lost is the semantic flavor of the conjunction as an alternative to ‘et’, ‘et’ being more like the ‘and’ found in possibly all languages. Also lost in antisymmetry is the strongly symmetrical intuition that conjoined NPs are sisters with the conjunction, violating binary branching.

Despite the often impressive work that went into its execution, there are a number of things I didn’t like about Kayne’s The Antisymmetry of Syntax. For one, the book was made very hard to read and comprehend because of the sheer absence of trees. Sure, there are lots of trees in the beginning, but they are all abstracted away from the syntactic content of actual sentences and are intended to justify the LCA and other meta-rules in purely formal and structural terms. Lacking in any clear lexical or functional specification of the nodes, these trees are a waste. Far better would the book be if example sentences in later chapters were given trees instead of [bracketed [notations]]. Another problem with Kayne’s book is the narrow focus on English and its European neighbors and friends. Casting a broader net could capture insights into humanity’s cornucopia of functional categories and their oftentimes idiosyncratic realizations. But this is just what minimalism is likely to dread. Finally, I offer without comment the final sentence of the book:

“To a significant extent, the LCA-based theory of syntax proposed here allows us to have the all too infrequent pleasure of seeing the theory choose the analysis.” (p. 132)

As I said, Kayne is consistent if nothing else. His way of doing syntax might not be the best way (only time will tell on that score), but he definitely provides a formal template together with a lingua franca which will enable syntacticians to communicate their findings to one another and perhaps one day to a potentially broader public. But I am not optimistic that the narrowing of diversity which minimalism longs for is in the offing, or is even really something to be desired. Regrettably, one consequence of reducing diversity in one place is the inevitable complication of other parts of the same grammar. Across the board simplifications are often a zero-sum game, I’m afraid, as economy and realism part ways.

In conclusion, I feel tempted to think of the agents of the antisymmetry sub-program of minimalism as being like the Chinese speakers caricatured in Humboldt's work. For Humboldt, Chinese is a less advanced language because it contains less morphology than the paragon cases of Latin, Greek, (and German!). Yet since the Chinese are clearly smart and fully human, Humboldt concluded that the Chinese can nevertheless succeed intellectually by following (unwaveringly) their language’s core grammatical principles (fixed word order, isolating and analytical word-forms). I feel that Kayne subscribes to an overly constrained and under-generating theory of syntax, but like Humboldt’s fictitious Chinese speaker, he succeeds through brilliance admixed with a maddening consistency. To his credit or detriment, Kayne follows his principles to the bitter end.

I recommend that anyone interested in minimalism should bypass this book and instead learn from the more piecemeal but wider ranging (typologically and areally) Annotated Syntax Reader, edited by Kayne, Leu, and Zanuttini. This Reader assumes less from the reader and provides a wider array of ideas to apply to whatever languages or constructions float you boat.

[Bernard · Rating 4 out of 5]

A seminal, and essential, monograph for understanding contemporary Chomskyan linguistics and one of the pillars of modern syntactic research. Elegant and intuitive throughout, though obviously its philosophical repercussions for the structure of language remain an open (but inviting) question.