A Look at Tesnière’s Éléments through the Lens of Modern Syntactic Theory Timothy Osborne 615 6th Street Apt. 110 Kirkland, WA 98033 USA [email protected] Abstract nent theoretician to have rejected the binary di- vision of the clause into a subject and predicate A recent project to produce a much belated and to have replaced this division with verb cen- English translation of Lucien Tesnière’s trality. The placement of the verb as the root of Éléments de syntaxe structurale has provided all syntactic structure was the all-important nov- the opportunity for an in depth look at elty (and the main act of genius) in his theory. Tesnière’s theory of syntax. This contribu- Given verb centrality, the theory of syntax that tion examines a few aspects of Tesnière’s Tesnière was proposing could not help but be work through the lens of modern syntactic theory. Tesnière’s understandings of constit- construed as a DG. uents and phrases, auxiliary verbs, preposi- Despite the fact that Tesnière is widely tions, gapping, right node raising, proposi- acknowledged as the father of an entire stream tional infinitives, and exocentric structures of syntactic theory, most syntacticians and are all briefly considered. Concerning some grammarians lack exposure to his work. Few of these areas, we see that Tesnière was vi- grammarians have actually read Tesnière’s Élé- sionary with his analysis, whereas in other ments de syntaxe structurale, largely because an areas, modern syntactic theory now rejects English translation of the Éléments is absent his account. Of particular interest is the fact from the world of linguistics. Spanish, Italian, that Tesnière’s theory was not entirely de- and German translations of the Éléments exist, pendency-based. His account of transfer (Fr. translation) acknowledged exocentric struc- but surprisingly, no English translation is yet tures, which means his system was also em- available. With this lacuna in mind, a recent pro- ploying constituency. In this regard, one can, ject to translate the Éléments into English has surprisingly, classify Tesnière’s theory as a been initiated and is continuing at present. This hybrid dependency-constituency grammar. project is providing an in depth look at Tesnière’s theory and has motivated the current 1 Introduction contribution. Tesnière’s Éléments is large in size, 670 pages Lucien Tesnière (1893-1954) is widely consid- with hundreds of tables and tree diagrams ered to be the father of modern dependency (stemmas). Tesnière addresses many aspects and grammars (DGs). While the dependency concept phenomena of syntax, whereby he employs ex- certainly existed in varying forms in the works amples from approximately two dozen lan- of numerous grammarians that preceded him, guages, many of which he actually spoke – Tesnière (1959) was the first to fully utilize the Tesnière was a true polyglot. In this respect, the concept of direct word-word dependencies in a intent of the current contribution is to briefly comprehensive manner and to illustrate these consider only a few important areas of the Élé- dependencies using tree representations (stem- ments, these areas being the ones that stuck out mas) that left no doubt about the analysis of syn- during the translation work. Certain aspects of tactic structure being proposed. In particular, Tesnière’s understanding of constituents and Tesnière appears to have been the first promi- 262 Proceedings of the Second International Conference on Dependency Linguistics (DepLing 2013), pages 262–271, Prague, August 27–30, 2013. c 2013 Charles University in Prague, Matfyzpress, Prague, Czech Republic phrases, auxiliary verbs, prepositions, gapping, fact that the definition is applicable to depend- right node raising, propositional infinitives, and ency structures as well and that it identifies sub- exocentric structures are considered below. trees as constituents. A subtree that consists of a Two highlights can be mentioned here up single node is simply a word, whereas a subtree front. First, Tesnière rejected much of the termi- consisting of more than one word is a phrase. In nology of syntax that preceded him, declaring other words, DGs can and do acknowledge con- that morphologists had imposed their nomencla- stituents and phrases just like constituency ture on the study of syntax and thus confused grammars do, the only difference being that DGs our understanding of syntax (ch. 15). In this re- acknowledge many fewer of both. gard, Tesnière had a penchant for introducing Tesnière certainly saw the need to new terms, many of which have not become es- acknowledge the status of subtrees as particular tablished. One can therefore speculate about the units of syntax, but his use of terminology in the reduced impact of his work due to his unfortu- area was not consistent and this inconsistency nate use of terminology. Second, Tesnière never has probably contributed to the confusion about actually employed the term dependency gram- whether dependency grammars acknowledge mar (Fr. grammaire de la dépandence). In fact it constituents and phrases. seems likely that he was not aware of the differ- Tesnière defined the node (Fr. nœud) as fol- ence between dependency and constituency, lows: since that distinction would be established later “We will define the node as a group con- during the reception of his work.1 In this respect, sisting of a governor and all the subordi- he did not shy away from employing constituen- nates that are to some degree either direct- cy in his theory of transfer (Fr. translation), a ly or indirectly dependent on that gover- fact that may have been overlooked until now. nor. The governor joins these nodes into a To conclude this introduction, a note concern- single cluster.” (ch. 3, §3) ing the citation practice employed below for Tesnière’s book is necessary. The Éléments is It should be apparent from this definition that split into 278 chapters, whereby each paragraph Tesnière saw any subtree of a tree as a node, in a chapter is numbered. When citing specific which in turn means that he was acknowledging passages, the chapter (ch.) and paragraph (§) are constituents and phrases, although the terminol- given (e.g. ch. 3, §3) instead of the page number. ogy he was using to denote these units (nœud) This practice avoids confusion that might arise if was different from modern usage (constituent, page numbers were cited due to the various edi- phrase). tions of the Éléments in various languages In fact Tesnière’s use of terminology was, as (French, German, Spanish, Italian, and soon stated, inconsistent in this area.2 While his origi- English as well). nal definition suggested that his node was to be understood as a subtree, his later (and preferred) 2 Constituents and phrases use of the term points to the meaning ‘vertex’. In other words, Tesnière usually meant just ‘vertex’ The constituent is the basic unit of syntactic when he wrote nœud despite the fact that he had analysis assumed by most constituency gram- defined the node to be a subtree, i.e. a constitu- mars. A constituent is typically defined as a ent. The contradiction in his use of terminology node plus all the nodes that that node dominates is seen most vividly in the passage where he is (for similar definitions, see Napoli 1993:167; comparing the node to the nucleus: Jacobson 1996:55; Haegeman and Guéron 1999:51; Carnie 2008:37). Given such a defini- “The node is nothing more than a geomet- tion, the number of constituents in a given tree ric point, whereas the nucleus is a collec- structure matches the number of nodes. In the tion of multiple points,…” (ch. 22, §12) past, many DGs seem to have overlooked the 2 This statement may be unfair. The Éléments was 1 According to Jurafsky and Martin (2000:489), Da- published posthumously. The inconsistency in the use vid Hays (1964) may have been the first to employ of the term nœud may have arisen as the manuscript the term dependency grammar. was being prepared for publication by others. 263 Comparing this passage with the previous one distinction that can be interpreted as accommo- where Tesnière initially defines the node, the dating the modern analysis. He distinguished contradiction should be apparent. between constitutive and subsidiary words inside The pertinent question now concerns the ex- nuclei (ch. 29). A constitutive word guarantees tent to which Tesnière’s inconsistent use of ter- the syntactic integrity of the nucleus, whereas minology has contributed to the fallacious per- the subsidiary word is a satellite of the constitu- ception that DGs do not acknowledge constitu- tive word. He also states (ch. 38, §13) that in a ents and phrases. They of course can and do split nucleus consisting of an auxiliary verb and acknowledge such units, although they have not a full verb, the auxiliary verb is constitutive. been clear about their use of the associated ter- Further, he explains that from an etymological minology. point of view, the constitutive word once gov- erned the subsidiary word (ch. 29, §18) and that 3 Auxiliary verbs this fact can be shown inside a nucleus by posi- tioning the constitutive word above the subsidi- Most modern DGs assume that auxiliary verbs ary word. This practice would result in tree rep- dominate main verbs, and in this respect, they resentations like the following one (my rendi- are consistent with most constituency grammars. tion, not Tesnière’s): In the Government and Binding framework (Chomsky 1981), for instance, a finite verb re- (2) has sides in I, which projects up to IP, the root node gone of the clause, and in the Head-Driven Phrase Structure Grammar framework (Pollard and Sag She home 1994), a finite auxiliary verb is the head daugh- The step from this tree to the modern analysis is ter in the clause, which means it passes its fea- not so great.