This is my forward to Postal 2010 "Edge-Based Clausal Syntax" (MIT Press)
Chomsky (1965, 70) defines a direct object structurally as [NP, VP]. Similarly, in introductory linguistics courses, the direct object is simply defined as the DP sister of V. In this book, Postal argues conclusively that this simple picture cannot be right. Rather, there are three kinds of object, each of which has different syntactic properties, as shown by how they behave in passives, middles, nominalizations, -able forms, tough movement, wh-movement, heavy NP shift, right node raising, re-prefixation, and many other tests. The definition of the direct object as the sister of a verb does not easily allow for such a three-way distinction.
In Postal’s framework (Metagraph Grammar, formerly Arc Pair Grammar), objects head 2 arcs, 3 arcs, and 4 arcs (henceforth, 2, 3, and 4 objects). Roughly, 2 and 3 objects correspond to the traditional notions of direct and indirect object, respectively. The distinctive properties of 4 objects have not been recognized before in any grammatical framework. These three types of objects have different syntactic properties, as illustrated in (1)–(3).
(1) a. John wrote the book.
b. The book was written by John.
c. The books that John wrote . . .
(2) a. The affair involved foreign banks.
b. *Foreign banks were involved by the affair.
c. The banks that the affair involved . . .
(3) a. Her name escapes me.
b. *I was escaped by her name.
c. *The students who her name escaped . . .
In (2a), the DP following the verb is a 4 object, which cannot undergo passivization but can be relativized. In (3a), the DP following the verb is a 3 object, which can neither undergo passivization nor be relativized.
As Postal states (sec. 2.8), ‘‘All this appears puzzling if one assumes . . .that the external syntactic properties of nominal expressions are basically determined by the sorts of phrase structure configurations underlying Chomsky’s (1965) definition of direct object. Absent such ideas, though, one can assume . . . that there are primitive grammatical relations and that different nominal expressions represent different relations without dependence on phrase structure configurations, via the possibility of corresponding to the heads of arcs with distinct edge labels.’’
In Postal’s framework, sentence structures are complex graph structures built on nodes (vertices) and edges (arcs). The labels of the edges are the names of grammatical relations (such as 2, 3, or 4). The node that heads a particular edge represents a constituent that bears the grammatical relation named by the edge label to its tail node. Such an approach allows two DPs that have very different grammatical properties (e.g., one heads a 2 arc and one heads a 3 arc) to occupy what look like identical structural positions in the tree (from the point of view of nongraph-based theories).
Certainly, many of the facts discussed in this book have never been addressed in minimalist syntax (while at the same time the generalizations integrate many well-known facts from the generative literature). In the minimalist framework, the operation Merge generates binary-branching constituent structures. In a theory with binary Merge, the most natural strategy would be to try to match 2, 3, and 4 objects up to different positions. If there were only one position in the VP (sister of V), this strategy would not work, since then it would not be possible to account for the three classes of objects. However, work in minimalist syntax argues for more complicated structures involving VP shells (see Larson 1988), where more than one position is available for objects. The challenge, then, is to match up the 2, 3, and 4 objects to positions within the system of VP shells and to show how their syntactic behavior (with respect to passivization, relativization, etc.) can be explained on the basis of this matching.
My personal experience has been that successful minimalist analyses (of the relevant phenomena) often mirror metagraph analyses to a surprising extent. As a simple example, consider the passive sentence The book was written by John (corresponding to the active John wrote the book). In Postal’s analysis, John is an initial 1 (subject) in the passive, but not a final 1, since it is demoted. The assumption is that John heads an (initial) 1 arc in both the active and the passive, thus accounting for the indisputable fact that John obeys the same selectional restrictions in each (see Chomsky 1957). The natural way to translate the metagraph analysis into minimalist terms is to say that the DP in the by phrase of a passive clause shares some position with the DP subject of an active clause. In fact, the minimalist analysis outlined in Collins 2005 claims that John is merged into Spec,vP (position of initial 1) in the passive, in exactly the same way that John is merged into Spec,vP in the active. The difference is that in the passive, John never raises to Spec,IP (position of final 1). Such correspondences between Metagraph Grammar and the Minimalist Program are not always simple (and even the seemingly simple correspondences are sometimes subtle and easy to misinterpret). For another example, see Collins, Moody, and Postal 2008 on camouflage constructions in African American English.
The treatment of anaphora in Metagraph Grammar (see in particular chapters 1 and 8 of this book) deserves special mention. In principles and parameters work, coreference (and bound variable anaphora) is indicated with coindexing: John1 said that he1 would be late. In this sentence, John and the pronoun he both bear the index 1; hence, they are coreferential. The coindexing relationship between a pronoun and its antecedent is the syntactic representation that is also assumed in the formal semantics literature. In Metagraph Grammar, there is no syntactic coindexing. Rather, to capture the fact that a pronoun and its antecedent corefer, the notion of overlapping arc is employed. Briefly, in the example given, the DP John heads a 1 arc in the matrix clause and simultaneously heads a 1 arc in the embedded clause (so the two 1 arcs overlap since they are both headed by John). In other words, the same DP is the subject of the matrix clause and of the embedded clause. The 1 arc in the embedded clause is ultimately replaced by an arc headed by the pronoun. In fact, the notion of coindexing (and its related semantic notion of coreference) is replaced by the syntactic notion of overlapping arc. Nowadays, minimalist syntacticians are also wrestling with the representation of antecedence, especially given Chomsky’s (1995) Inclusiveness Condition, which rules out the use of indices in the representation of binding relations. They could benefit from consulting Metagraph Grammar, which is a deep source of ideas on how to represent antecedence. For more on the notion of antecedence and some preliminary remarks on translating the metagraph treatment into a constituent-structure-based approach, see Collins and Postal 2010.
From the perspective outlined above, this book becomes a rich source to mine for data and theoretical ideas. Therefore, I disagree with Postal’s claim that the book argues in some fundamental way for a graph-theoretical/relational approach to syntax, over a Merge-based approach. That said, one can ask whether relational approaches to syntax are in some sense more conducive than other approaches to the discovery of generalizations like those found in this book. Without a doubt, there is some truth to this. The central object of study in Metagraph Grammar is the primitive arcs: 1, 2, 3, 4, and so on. Hence, all statements in the theory are statements about these arcs and the relations between them. Such a constraint acts to sharply focus theorizing and to bring out the relevant generalizations.
Overall, the book is a tour de force empirically. It illustrates a fact often lost sight of in theoretical discussions: we have barely begun to scratch the surface of the vast empirical domain of English grammar, and English is by far the best-studied language on earth. It is reasonable to believe that understanding the grammars of Ewe, Ju|’hoansi, N|uu, and thousands of other languages to the same depth that we understand the grammar of English would have great implications for Universal Grammar. The empirical achievements of the last 50 years of generative syntax should not lead us into complacency. Far from approaching the end of syntax, we are only now at the beginning of syntax.
Chomsky, Noam. 1957. Syntactic Structures. The Hague: Mouton.
Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT
Chomsky, Noam. 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Collins, Chris. 2005. A Smuggling Approach to the Passive in English. Syntax 8,
Collins, Chris, Simanique Moody, and Paul M. Postal. 2008. An AAE Camouflage
Construction. Language 84, 29–68.
Collins, Chris, and Paul M. Postal. 2010. Imposters. Ms., New York, New York
Larson, Richard. 1988. On the Double Object Construction. Linguistic Inquiry 19,