Book announced at https://linguistlist.org/issues/35.62

AUTHOR: Diego Gabriel Krivochen
TITLE: Syntax on the Edge
SUBTITLE: A Graph-Theoretic Analysis of Sentence Structure
SERIES TITLE: Empirical Approaches to Linguistic Theory
PUBLISHER: Brill
YEAR: 2023

REVIEWER: Andrew Carnie

SUMMARY

This monograph takes a novel approach to the description of syntactic structure by using independently motivated ideas from the mathematical field of Graph Theory (GT) and supplementing them, where necessary, with additional syntactic conditions. Key to the proposal at hand is the idea that one wants to minimize the number of nodes in one’s description, and instead exploit the ability to have multiple connections among those nodes. The book investigates a large number of empirical phenomena and shows how GT provides a superior and explanatory account of them.

Chapter 1, Introduction: Setting the Scene, provides the historical and theoretical context for the book, drawing upon critical ideas from transformational grammar, Lexical-Functional Grammar, Relational Grammar, Arc Pair Grammar, Dependency Grammar, and other perspectives. Some of the key foundational assumptions that are motivated include (1) The idea that the Graph-Theoretic (GT) approach is not grounded in typology, language acquisition, nor learnability, but is instead about description. (2) The book makes no claims about universal grammar. (3) The book entirely divorces the issues of linearization and abstract semantic structure, concentrating almost exclusively on the latter. (4) The approach abandons the single root condition on trees as well as the extension condition. (5) The author takes the programmatic perspective that an ideal theory of syntax minimizes the number of nodes, particularly avoiding abstract nodes, and maximizes the number of connections among these nodes. (6) The connections among nodes are motivated by semantic factors, as in a dependency grammar. (7) The approach is also declarative and constraint-based, rather than procedural and derivational. These assumptions, paired with Graph Theory, allow for greater explanation for such phenomena as discontinuous constituents, apparent instances of mixed computation, and the range of phenomena that have been suggested to fall out from multidomination.

The theoretical meat of the approach is found in Chapter 2, Fundamentals of Graph-Theoretic Syntax. The chapter starts with a primer on the fundamentals and critical definitions from GT (e.g., graphs, vertices (nodes), edges, cyclicity, adjacency, directionality and order, neighborhood sets, traversal, paths, trails, vines, domination, root, connectedness, completeness, etc.). It also sets out some definitions that are specific to syntactic theory (e.g., arbor, elementary and derived graphs, union, intersection, substitution, adjunction, addressing). Critical to the application of GT to syntactic description lies in the content of vertices (nodes) and how they determine the edges in the graph. The approach advocated here makes use of semantic types instead of categories. These determine a relation of predication, which in turn determines hierarchical ordering of nodes in the graph.

Chapter 3, A Proof of Concept: Discontinuous Constituents, demonstrates how GT captures the correct semantico-syntactic relationships between items in situations such as verb-particle constructions, right node raising, across the board phenomena, and relative clause extraposition, where the items exhibit properties of constituents, but are not linearly contiguous. The crux of the matter is that in GT, linear ordering is not a syntactic concern. This combined with the fact that the theory allows multidomination and has abandoned both the non-tangling condition and the requirement that syntactic structures have a single root allows for edges (i.e. lines in the diagram) to cross, thus giving a straightforward description of these kinds of constructions.

The approach advocated in Chapters 2 and3 is then extended and examined in comparison to other approaches in Chapter 4, Some Inter-Theoretical Comparisons. The approach is first compared to Tree Adjoining Grammar (TAG) and Metagraph theory. However, perhaps the most important contribution of this chapter lies in its comparison of GT to Dependency Grammar (DG). At first blush, GT syntax looks suspiciously like a dependency grammar, albeit a version couched in an independently motivated mathematical formalism. But deeper inspection shows that it differs from the standard implementations of DG in some critical ways: It allows multiply rooted graphs; it imposes a stricter correspondence between predicates and their arguments; and there is no strict one-word/one node correspondence.

One of the critical features of the GT approach is its relatively flat non-configurationality, where the structure is determined by semantic dependencies rather than by constituency. The challenge then comes when it’s important to identify different grammatical functions (e.g. subject, object, etc.), which are traditionally defined, at least in the Chomskyan tradition, in terms of hierarchically organized constituency. If you have a set of semantic domination relations {(v1, v2), (v1, v3)}, where v1 is a predicate and v2 and v3 are arguments of v1, there is no way to distinguish v2 and v3 for relations like subject and object. The solution to this problem is discussed in Chapter 5, Ordered Relations and Grammatical Functions. Krivochen adopts insights from Relational Grammar and LFG that there is a hierarchy of grammatical functions (roughly: subject < object < indirect object, etc.). This is encoded by ordering the sets of dependencies: <(v1, v2), (v1, v3)>. There is then an interpretive principle that identifies the first pair in the order as the subject, the next as the object, etc. There are some challenges to such an approach. For example, expletives in unaccusative constructions such as “there arrived a package at the office” are not actually treated as subjects since they are not semantic arguments of the predicate.

The next nine chapters concern the implementation and refinement of the theory to account for a variety of syntactic phenomena. In Chapter 6, Krivochen looks at classic examples of raising and control. Chapter 7 looks at parentheticals and clitic climbing in Spanish. Krivochen proposes a theory of licensing, with exceptions being made when a graph is self-contained. Chapter 8 provides an account of binding theory in terms of parallel arcs and locally defined graphs. Chapter 9 compares the properties of non-restrictive relative clauses from restrictive ones, the former having paratactic properties and the latter allowing hypotactic structure. Wh-constructions are the topic of Chapter 10, where filler gap dependencies are a consequence of the fact that the sets of dominance relations are in an ordered set where the wh-element is visited twice via the edge set. Chapter 11 concerns Bach-Peters sentences. Chapter 12 looks at two different kinds of coordination. In Chapter 13, there is an investigation of relation-changing constructions such as passive and dative shift, which amount to lexical alternations rather than syntactic ones. Chapter 14 surveys a number of phenomena that are more challenging for the approach, including restrictions on rightward extraction; null and implied arguments, resumptive pronouns, and issues surrounding reflexive interpretation when the antecedent is quantified.

The key components of the GT approach are surveyed in the concluding Chapter 15, which is followed by a short appendix that compares the theory to other graph theoretic approaches that have been proposed in Minimalism and Dependency Grammar.

EVALUATION

This book is an important contribution to syntactic theorizing, quite independent of whether or not a researcher adopts the particular approach advocated by the author. First, the work demonstrates that by peeling back many of our linguistics-specific assumptions we can often find insightful analyses that follow from the mathematical foundations at the heart of the formalisms we use. Second, it highlights a variety of problematic phenomena that have escaped explanatory analyses based in constituency based approaches such as Transformational Grammar and Minimalism. Finally, it provides a rigorous formal language perspective on the kinds of formalisms and graphical devices we can use for explicating phenomena.

This is a rich and thorough treatment of the connections that lie among syntactic objects, and because it takes a novel approach, there is a lot of ground to cover. Befitting such a work, Krivochen touches on a remarkable range of theoretical, meta-theoretical, philosophical and empirical issues. So, this is definitely not light reading for the beach. I have to confess I had trouble following the organization and logic of the first two chapters, in particular. Krivochen skips from topic to topic in Chapter 1, often without obvious connections – at least for me – among the items being discussed. One finds everything from questions of the ontological status of representations, to formal language theory, to particular empirical challenges to mainstream thinking in syntax all bundled together. The flow of argument in Chapter 1 was challenging for me. That said, as I mentioned, when taking an entirely new approach, one has a lot of ground to cover, and I appreciated the thoroughness with which Krivochen carefully motivated the myriads of issues that are foundational to his approach.

As mentioned above, two of the most compelling parts of the work are its formal rigor and empirical breadth. That said, there is an area that I wish could have seen a little more ink: the link, or lack thereof, between GT representations and linear order. Krivochen excludes linear order from his consideration, concentrating only on hierarchical relations. He justifies this methodological decision in Chapter 3. It’s important to note that Krivochen is not operating in a vacuum here. There is a long tradition of divorcing linear order facts from hierarchical structure dating back to at least Relational Grammar but also found extensively in recent works in the Chomskyan tradition, where for example the effects of so-called head-movement are just reduced to morpho-phonological conditions (see for example the discussion in Chomsky 2000, Matushansky 2006). So, in this regard, the absence of more than a passing discussion of linear order is not unusual. That said, I was often left wondering how the graphs translated into the surface orders we see, and I think the book would have benefited from at least a little speculation about how this might work. For example, in Krivochen’s system, the NPs that appear before raising verbs such as ‘seem’ are not arguments of the predicate so are not connected to it directly in the GT representation; and as such they are literally not considered subjects of the raising predicate. But then there are numerous unsolved mysteries about these NPs, all of which would follow from them actually being the subjects of the raising verb: Why do they take nominative case? Why can they bind reflexive experiencer arguments of the raising verb (e.g. Marie seems to herself to be right)? And, of course, why do they appear linearly before the main clause verb like other subjects? I assume these are all solvable problems, but in cases like this and others (e.g. the rightward position of the shared constituent in RNR constructions, the leftward position of wh-elements in wh-questions in English, the differences among SVO and SOV and VSO languages, etc.), I was left a little unsatisfied. This doesn’t take away from the impressive empirical results of the GT treatment and the book is already a fairly lengthy tome, but I would have wished for at least a little speculation on these kinds of issues. This however is a minor complaint.

Scholars from a wide variety of theoretical perspectives will find this volume of interest. The specific use of an independently motivated branch of mathematics to capture the properties of syntactic structures is an important part of understanding the deeper properties of syntax and helps to refine our thinking about how to approach syntactic analysis and description.

REFERENCES

Chomsky, Noam (2000) Minimalist Inquiries; The framework. In Roger Martin, David Michaels and Juan Uriagereka (eds), Step by Step: Essays on minimalist syntax in honor of Howard Lasnik. MIT Press. Pp 89-155.

Matushansky, Ora (2006) Head Movement in Linguistic Theory. Linguistic Inquiry 37.1, pp 69-109

ABOUT THE REVIEWER

Andrew Carnie is Professor of Syntactic Theory and Dean Emeritus at the University of Arizona. He specializes in the syntax, morphology and phonology of the Celtic languages, with a particular emphasis on Irish and Scottish Gaelic. He has theoretical interests in constituency and dependency, case, and VSO languages. He’s the author or editor of 13 books, including the best-selling textbook “Syntax: A Generative Introduction” from Wiley and “Constituent Structure” from OUP.