In 1959, Eléments de syntaxe structurale, a monograph by French linguist Lucien Tesnière, was published by Klincksieck in Paris (Tesnière Reference Tesnière1959). This book is now hailed as having laid the foundation of dependency grammar (Liu Reference Liu2009). Janez Orešnk holds that ‘[t]he history of modern linguistics shows that no school of linguistic thought can escape using – in one way or another – Tesnière’s ideas, notably his valency theory’ (Orešnk Reference Orešnk, Čop, Orešnk, Skubic and Tekavčić1994: 7). In this almost 700-page work, Tesnière analyzed over 5000 sentences and phrases from more than 60 languages, in an attempt to demonstrate the universality of his syntactic analysis method for human languages. According to Murray (Reference Murray2004: 431), this book ‘offered an original view of syntax applicable to all languages’.
Within 60 years of the original French publication, there have been translations into German (Reference Tesnière1980), Russian (Reference Теньер1988), Spanish (Reference Tesnière1994), Japanese (2007) and Italian (Reference Tesnière2008), referenced below. But until now, there has been no English translation to allow a much wider audience to access Tesnière’s theory. This translation thus fills an important gap. One translator is a native speaker of English, and the other a native speaker of French, but more importantly, both translators are active researchers of dependency grammar. These two factors contribute to the quality of the translation. To help the reader better understand this classic work, the two translators provide a 44-page introduction. In addition to their translation of the original, they also corrected the various minor errors and added indexes of terms, languages, linguists and citations, which help the reader to access relevant information easily.
The volume is divided into three major parts, containing a total of 278 chapters. Part I, ‘The connection’ (1–324), is subdivided into five ‘books’, with 133 chapters. Part II, ‘Junction’ (325–362), has 17 chapters. Part III, ‘Transfer’ (363–680), is subdivided into six ‘books’, with 128 chapters. Due to limits of space, only the key concepts of Tesnière’s structural syntax will be introduced here, such as stemma, connection, valency and transfer.
On opening the book, the reader will see structural diagrams throughout (366 in total), representing sentences of varying complexity. Tesnière called these syntactic trees ‘stemma’. To arrive at the stemma for a sentence, it is necessary to understand some of Tesnière’s basic principles related to structural analysis, which constitute the main content of Part I.
Tesnière held the view that a sentence is an organized set, the constituent elements of which are the words. ‘Each word in a sentence is not isolated as it is in a dictionary. The mind perceives connections [boldface in the original here and in subsequent direct quotations] between a word and its neighbors. The totality of these connections forms the scaffold of the sentence’ (3). The connection is thus the foundation of structural syntax. What Tesnière called ‘connection’ is what linguists now call ‘dependency’, hence why structural syntax is known as dependency grammar today.
How does this system operate? According to Tesnière, ‘[t]o construct a sentence is to breathe life into an amorphous mass of words, establishing a set of connections between them …Conversely, understanding a sentence involves seizing upon the set of connections that unite the various words’ (4).
Connection has these basic features: ‘Structural connections establish dependency relations between words. In principle, each connection unites a superior term and an inferior term’ (5). ‘A word can be both subordinate to a superior word and governor of an inferior word …The set of words of a sentence constitutes a veritable hierarchy …In principle, a subordinate can only depend on a sole governor. A governor, in contrast, can govern multiple subordinates’ (6). This asymmetry of connections makes it possible for the structural syntax analysis to result in a hierarchical tree-like shape. However, to provide a complete sentence structure, it is necessary to decide on the central node of a sentence. ‘A node formed by the governor that governs all the subordinates of a sentence is the node of nodes, or the central node …The node of nodes is generally verbal’ (7).
Given the conventions above, a structural diagram of a sentence can be devised. ‘The set of connection lines constitutes a stemma. The stemma clearly shows the hierarchy of connections; it presents the various nodes that join connections into groups schematically, and therefore, visually manifests the structure of the sentence’ (8). This syntactic analysis method based on a dependency diagram (syntactic tree) not only makes Tesnière the first scholar to use tree diagrams in syntactic research (Marcus Reference Marcus1967: 263), but also makes dependency parsing the focus of current computational linguistics (Kübler, McDonald & Nivre Reference Kübler, McDonald and Nivre2009, Liu Reference Liu2009).
The asymmetry of connections is the prerequisite for creating a syntactic tree. But this makes coordination a particular problem under the framework of structural syntax. Acknowledging the problem himself, Tesnière wrote a separate part, ‘Junction’, specifically to address this. Although this part is relatively short at 38 pages, it is an indispensible part of a theoretical framework intended to offer a thorough analysis for the syntactic structure of human languages. Indeed, coordination remains one of the issues receiving the most discussion and debate among researchers of dependency grammar (Hudson Reference Hudson1988, Liu Reference Liu2009: 105–106).
Tesnière also mentioned the function of stemma in testing students’ understanding of a sentence in language teaching and learning. This indicates that his structural syntax was intended not only as a syntactic theory, but also as a practically oriented (language teaching and learning) sentence analysis method. This practicality is also embodied in the concept ‘metataxis’, introduced in the final section of Part I, where Tesnière used this concept to refer to the change of structures during translation between different languages. Newmark (Reference Newmark2001: 32) describes this section as ‘forty pages of valuable translation theory’.
Tesnière was also known as the founder of the modern valency theory (Ágel Reference Ágel2000, Liu Reference Liu2009). Valency is discussed in Book D of Part I, but in very little detail. It has only 23 sections (Sections 97–119), 46 pages in total (239–284). Tesnière compared the relationship between verbs and actants to an atom’s ‘bonds’, and defined the numbers of ‘bonds’ that a verb has as its valency. The remainder of the book is devoted to the discussion of verbs with zero, one, two, or three actants.
To better understand Tesnière’s conceptualization of ‘valency’, we need to look back at his exposition on verbs: ‘The verbal node, found at the center of the majority of European languages, is a theatrical performance. Like a drama, it obligatorily involves a process and most often actors and circumstance …Transferred from the theatre to structural syntax, the process, the actors, and the circumstances become respectively the verb, the actants, and the circumstants’ (97).
These two analogies reflect Tesnière’s two views on valency: The comparison of a verb to an atom can be seen as emphasizing the syntactic aspect of valency, focusing more on sentence form, while the comparison of a verb to a theatrical performance can be seen as emphasizing semantic properties of valency, because it focuses more on sentence content. Therefore, it may be inferred that Tesnière intended valency to have dual properties: syntactic and semantic. Tesnière believed that from the semantic perspective, the first actant is the agent of the action, i.e. the subject in traditional grammar, and the second actant is the one that bears the action, i.e. the direct complement (object) in traditional grammar. The difference between subject and object lies in their meanings, while in terms of structure, both subject and object serve as complements to complete the governor. In this sense, the subject is not distinct from other complements. The third actant is what benefits or takes detriment from the action, i.e. the indirect object in traditional grammar (102–103). These few lines of exposition not only help to determine the actants, but also deepen our understanding of valency; that is, valency has not only the property of quantity but also differences in quality. Using the analogy of an atom with ‘bonds’ or ‘hooks’ (Fr. crochets), different words may have different numbers of ‘hooks’, and the shape of the ‘hooks’ may also be different.
In his book, Tesnière reiterates the following notion: ‘Speaking a language involves transforming structural order into linear order and conversely, understanding a language involves transforming linear order to structural order’ (12). In other words, the generating and understanding of sentences is a process that shifts back and forth from linear order (one dimension) to structural order (two dimensions). During this process, the function of stemma cannot be neglected, because it is capable of representing the deeper structure of a sentence. In Tesnière’s stemma, only content words can occupy the place of a node. The author also proposes the use of symbols rather than actual words in a stemma, which are named ‘virtual (symbolic) stemma’ as opposed to ‘real stemma’. In a virtual stemma, O stands for noun, A for adjective, I for verb and E for adverb (57). In fact, these four symbols are the final letters of the labels for the four corresponding word types in Esperanto. The virtual stemma forms the basic framework or pattern of a sentence and, as a generalization of a sentence structure, occupies a very important position in Tesnière’s structural syntax, which can thus be described as virtual-stemma-based and parsing-oriented. The virtual stemma derived from valency is the basis for generating analytical results.
Furthermore, the nodes of Tesnière’s stemmas are all content words. In other words, functional words such as prepositions, which are used to denote grammatical relations, cannot be used as nodes in Tesnière’s stemmas. In this sense, Tesnière’s syntactic tree is meaning-oriented. We use the expression ‘meaning-oriented’ here because though these functional words cannot occupy an entire node, they appear in the stemma in some form, and will normally become a constituent part of a transferred node. In this respect, the structural stemmas of a sentence in Tesnière’s theory of structural syntax differ greatly from the dependency trees in contemporary dependency grammar. In a current dependency tree, every word, be it a content word or a function word, occupies a node in the tree (Kübler, McDonald & Nivre Reference Kübler, McDonald and Nivre2009, Liu Reference Liu2009).
To allow a position for every word of a sentence in the virtual stemma consisting of only I, O, A and E, Tesnière proposed an approach called ‘transfer’ (Fr. translation). He defined it as ‘the changing of the grammatical category of a content word to another grammatical category’ (368). It is thus practically feasible to construct a sentence stemma whose nodes are content words through transfer, during which function words play the role of transferring content words from one category to another. Tesnière devoted more than 300 pages to the discussion of transfer, which accounts for almost half of the entire work. Unfortunately, subsequent researchers have shown little interest in ‘transfer’. We argue that ‘transfer’ is an integral part of a syntactic theory based on virtual stemma. Without it, the syntactic system based on virtual stemma could not operate, in that it would not be possible to generalize infinite sentences with finite patterns, to represent recursive operations in the syntactic model, to reasonably limit the number of syntactic rules, or to carry over the rules for the analysis of simple sentences to the analysis of complex sentences.
Through the concepts of virtual stemma and transfer, Tesnière endowed his syntactic theory with the basic functions that any modern syntactic theory should have. From the perspective of syntactic analysis, virtual stemma provides not only a tool of sentence abstraction and generalization, but also a means of describing sentence formation rules.
The ideas about valency and dependency that Tesnière developed in this book laid the foundations for subsequent researchers to construct a universal parsing system for human languages. We believe that the publication of this English translation will not only help disseminate and popularize Tesnière’s original ideas about the syntactic structure of human languages, but also deepen our understanding of those syntactic structures.
