Generative Grammar

The contribution of Generative Grammar to Latin and Ancient Greek linguistics


2. New perspectives and new trends

2.1. The consequences of "Principle and Parameters" framework

The starting point of the current theory named Minimalism is Principle and Parameters. The goal of this evolution is to reduce the theory to the maximal simplicity (Economy Principle), with very few rules. In addition to simplicity comes also ‘elegance’ - I’ll quote Boeckx (2006: 120):

  • “Work within minimalism seeks to develop beautiful theories”

This framework has had two consequences in the study of dead languages:

- a reunification of ancient and modern languages, licensing a new perspective in comparison

- explanations on diachronic evolutions


A strong hypothesis is that Latin and Greek have the same type of syntactic arrangement that all the languages, that is to say an asymmetric structure from which the following structures are derived by movement, something like:

Languages differ from each other by some parameters, for example, as we have seen before, some languages are pro-drop and others are not. Parameters can be interconnected : according to Cecchetto and Oniga and Ferraresi and Golbach, the loss of rich infinitive inflection is connected to the demise of infinitive clause.

A new form of comparison is developed starting from P & P assumptions, for example by Longobardi and some colleagues working on Latin, Greek and modern languages. Under the label “Modularized Global Parametrization” they propose an alternative approach to classical typology. One of the goals is to discover how many parameters are relevant and necessary to compare, for example, all the types of NP structures. The method is exhibited in Longobardi (2003) Methods in Parametric Linguistics and Cognitive History, and a good example of the comparison between ancient and modern languages is Crisma & Gianollo (2006) Where did Romance N-raising come from ? A parallel study of parameter resetting in Latin and English.


Diachronic change is taken into account and explained by parametrical modifications. For example, concerning AcI, Calboli (1990) assumes that the change from AcI construction to quod/quia has to be located within WH- movement :

  • «[WH- movement] causes the development of the CP-barrier, linked to IP, and the following series of barriers makes an essential contribution to the transformation of Latin from a non-configurational language to a configurational language» G. Calboli (1990: 223)

However Ferraresi & Golbach’s hypothesis is that the change from this structure to quod/quia subordinate clauses has no syntactic motivation according to Inertia principle, and they assume that syntactic changes do have external origins (Greek influence of course coming from Vulgata) or are consequences from semantic or phonological changes. In this case, the loss of rich infinitive inflection is connected to the demise of infinitive clause.

A rather similar explanation, starting from phonological considerations, is given by Calabrese (1998) concerning the case system reduction from Latin to ancient Romance languages.

Greek offers a better continuum than Latin to the diachronic change study: a complete analysis from Archaïc to modern Greek has been developed by Geoffrey Horrocks (1997 Greek : A history of the language and its speakers, London & New York, Longmans).

2.2. Some examples

I want to focus on two main topics in present research : NP structure and Left Periphery of the clause.

In both topics, generative studies lean on previous observations made in formal semantics and functional grammars.

2.2.1. NP structure

First remark, the internal structure of NP is similar to any Phrase structure, since this structure is supposed to be universal. Oniga (2007 : 167) gives the following general representation :

According to this line of analysis and as observed previously by Bertocchi & Maraldi (1990), the postposition of objective genitive is expected, while the subjective one would be usually pronominal. The data of Giusti & Oniga (2007) shows that the unmarked word order is actually GenSubj – N – GenObj in the rare cases in which both genitives occur :

  • (20) pro ueteribus Heluetiorum iniuriis populi Romani (Caes. Gall.1,30,2)
  • (21) omnium exspectatio uisendi Alcibiadis (Nep. Alc. 7,6,1)

How to explain now unexpected word order like:

  • (22) Gen.Obj. pronominal : locorum descriptione
  • (23) Gen.Subj. postnominal : rapina praedonis

Word order variations have to be explained by movements from heads or complement positions to higher positions in the left side of the tree:

a- looks like a passive NP :

and in support to this analysis Giusti & Oniga notice that the Agent can be expressed by PP ab + abl.:

  • (24) Ad haec Alexander gratiarum actionem ab hoste superuacaneam esse respondit (Iustin. Epit. 11, 12,11)

In Devine & Stephens’ analysis (2006 : 383) the same pre-head objective genitive occupies a Focus position:

b- Since there is no right movement, we must consider that the N moves to a higher position at the left side of Gen.Subj. :

Concerning dislocated NPs, the generative analysis is not so far from Bolkestein’s one. The only difference is that Topic and Focus are syntactic positions with the same structural characteristics as the others. When the N is in the very left side of the entire NP, according to Giusti & Oniga’s and Devine & Stephens’ (2006) analyze, it occupies a Topic or Focus position, as in :

  • (26) [N uerbum] [NP aliquod ardens ti]

Pre-determiner adjectives and genitives also support a focus interpretation, as in :

  • (27) magnai [NP aliqua ti ac nobilis uirtus]

This movement, called scrambling, takes place, cf. Devine & Stephens (98 sqq), at an intermediate level of derivation.

Others genitives and adjuncts must be taken into consideration. In the so-called “maximal structure theory” (Devine & Stephens 2006 : 380), NPs contain many different positions that host, in a hierarchical order, a great variety of adjectives and perhaps 4 different genitives (Longobardi & Gianollo). For example, Longobardi (2000) gives the following structure for NP (see infra) :

Let us turn now to a second observation : NP became DP in the generative literature : noun is no longer considered the head of the Noun Phrase, but Determiners would be heads and N would be complements. Why ? (cf. Alexiadou, Artemis, Liliane Haegeman & Melita Stavrou, (2007) Noun Phrase in the generative perspective, Berlin, Mouton de Gruyter.)

a- Semantic properties are projected from Specifiers, for example the crucial distinction between definite and indefinite ;

b- adjuncts are expected to be recursive and to occupy any position, like adjectives; determiners occupy a fixed position on the left side of the Noun, so we can reverse the analysis, saying that complement are on the right side, after the head. Therefore Determiners work like Heads.

c- determiners are often very close to pronouns from a morphological point of view. So we may say that pronouns are “intransitive” determiners, in the same way that adverbs are often considered intransitive prepositions. Everybody knows that in Latin grammars there is only one morphological class : the pronouns-determiners.

This analysis appears to be relevant for languages in which there is an article or determiners in a fixed pronominal position. Therefore Greek NP is a candidate to an analysis as DP. That is actually the presentation in Guardiano’s studies.

A difficulty is immediately arising : what happens when there are several determiners? Are they all heads? Obviously no. Giusti (1997, 2001) assumes that only the article occupies the functional head position. The other determiners do not occupy a fixed position. For example in Ancient Greek (Guardiano 2008), demonstratives can be pre-nominal, when deictic, or post-nominal, when anaphoric. And all the other constituents of the DP can move to different positions, including pre-head position, but not pre-nominal position in classical Greek :

  • (28a) ταύτην τὴν τέχνην (Plat. Apol. 20c)
  • (28b) αὕτη ἡ πλημμέλεια ἐκείνην τὴν σοφίαν (Plat. Apol. 22d
  • (28c) τὴν φυγὴν ταύτην (Plat. Apol. 21a) (Guardiano 2008)
  • (28d) τὴν ταύτην φυγὴν

This pre-nominal position is available when there is an adjective (“When one or more adjective is in the DP the demonstrative surfaces immediately after the first adjective (the noun does not cross over any adjective)”):

  • (29) τὰ ἐλεινὰ ταῦτα δράματα (Plat. apol. 35b)

A second difficulty concerns languages that have no article, such as Latin, of course. This objection is well known. The usual answer is that bare-nominals are restricted to particular cases : for example in English, mass nouns, generic uses or indefinite plurals. However Latin offers a more difficult situation : there is no difference between definite and indefinite, neither in the singular nor in the plural. Therefore one of the main functions of the article, to link lexical items to extralinguistic references, is missing. According to the Longobardi’s generalization :

  • “A ‘nominal expression’ is an argument only if it is introduced by a category D. DP can be argument, NP cannot.” (1994 : 628)

So we face a great difficulty. And accordingly Latin word order variations concerning Demonstratives, Genitives and other NP constituents do not necessarily follow the same principles as in Greek. The DP-hypothesis is not adopted in Oniga’s (2007) presentation.

Diachronic evolutions from one structure to another are studied, in Greek by Guardiano and Manolessou, in Latin by Polo and Gianollo.

Most of Gianollo’s studies deal with diachronic change into NP. Her general proposal is that

  • “The changes … from Latin to Romance can be reduced to a series of discrete parametric changes caused by reanalysis of existing constructions on the basis of identifiable local causes.” (2009 “Prepositional genitives in Romance and the issue of parallel development”)

The first crucial change lies in Genitive-Noun to Noun-Genitive change, ruling out the distinction between Objective Genitive and pre-head Genitives. The classical way of describing this type of phenomenon is that a marked construction, competing with an unmarked construction in some specific context, becomes itself unmarked, replacing the first one in all contexts. But we need to explain why. According to Inertia hypothesis this change has to be explained by a parametrical change, whose cause would be external to syntax.

2.2.2. Word order

A great number of studies deal with this issue, directly or indirectly. We have seen that it concerns NP structure. It also concerns, of course, the whole sentence.

Ostafin’s PhD (1986) is the most representative work and a starting point for further investigations. Two other books, from Devine & Stephens, dealing specifically with hyperbaton in Greek (2000) and Latin (2006), is actually about word order in general.

All these studies start from functionalist or pragmatic considerations and try to give syntactic representation of Topic and Focus interpretation. The first consequence is a complete modification of the (supposed) universal sentence structure. Now, from Rizzi’s (1998) paper, it is assumed that the sentence is composed of three layers :

- CP layer, hosting functional categories, such as Topic, Focus and other operator-like elements (Force = illocutionary adverbs or conjunctions)

- IP layer, hosting functional categories corresponding to morphological specifications on the verb, such as Tense, Aspect and so on

- VP layer hosting lexical categories and whose head is V

Functional categories of the CP layer can host any lexical category moved from VP layer or directly inserted in the so called Left Periphery. There is no definitive agreement among generativists concerning the number and the nature of the categories of the Left Periphery. One usually assumes that there are at least two different positions for Topical elements.

This approach implies that we are able to bring out a basic word order which others are derived from :

  • « In un quadro generativo, in cui alle frasi vienne assignata una struttura dove ogni singolo costituente occupa una posizione strutturale determinata, la classifazione delle frasi … potrà essere effettuata solo dopo che si sarà risposto a domande come : ‘l’oggetto diretto / il verbo / ecc. di questi esempi occupa la sua posizione basica o una posizione derivata ? Come possiamo dimostrarlo ? » (Salvi (2003))

The means of obtaining this kind of information are :

- statistics on extensive corpus

- typological analysis

- diachronic evolution

Generative Grammar, that had first rejected quantification on performances, now systematically uses statistics and accepts typological considerations and usual conclusions about the evolution from Latin to Romance. The main contribution on this last issue is to provide a new insight into synchronic and diachronic variations. Polo’s (2005) paper shows a way of explanation for the development of post-verbal object. In the Petronian corpus she argues that 10% of post-verbal objects cannot be explained by pragmatic functions. According to her, the crucial element in this typological change is a semantic feature : when the object is prototypical, it is preverbal, but when non prototypical, with [+ animate, + human, + definite] features, it is postverbal.

The first goal is to define the fine structure of the Left Periphery. Salvi (2004) is the first to have adapted this hypothesis to Latin and to have used it to explain the appearance of atone versus tonic pronouns in Latin.

Dankaert’s PhD deals especially with the distinction between two types of Left Edge Fronting in subordinate clauses : the first is a Topicalization, the second a focalization. From a diachronic point of view, he observes a decline of focalized LEF from archaic to classical Latin and his hypothesis is that LEF’s decline is connected to VO development, which he argues to be functionally equivalent. In an ongoing research I try to distinguish two types of topical structures : Topicalization and Dislocation, and among dislocated structures, Hanging Topic Left Dislocation and Clitic Left Dislocation.

Topicalization is derived by movement (scrambling), while in HTLD and CILD the dislocated element is generated directly in its functional position. In the absence of prosodic information, a decisive criterion is the presence of an overt anaphoric pronoun :

  • (30) Topicalisation : In cellam oleariam haec opus sunt. (13,2) vs Pars autem fructuaria diuiditur in cellam oleariam, torculariam … (Col. rus. 1,6)
  • (31) HTLD : Mercator Siculus , quoi erant gemini filii, ei … mors optigit (Plaut. Men. 1)
  • (32) CILD : illos qui dant eos derides (Plaut. Asin. 527)

If the theory is consistent, it foretells that NP/DP has a Left Periphery too; actually Giusti & Oniga analyze in this way some pre-head objective genitives or pre-determiner adjectives – see example (26-27).

3. Results and prospect

As a conclusion, and according to my opinion, I would summarize what the main contributions of Generative Grammar. to Latin and Greek linguistics are :

a- Methodology : the methodology is very close to scientific approach ; hypotheses are tested on corpus; predictions are made concerning grammaticality and acceptability of some structure; then we verify if data do confirm or not this hypothesis. This is the mean of making discovery. In reverse direction, complexity of data must be reduced to the simplest set of explicit rules. The theory is not once and for all fixed, so several successive explanations appear (for example concerning AcI).

The theory integrate interfacing modules : morpho-phonology, syntax, Logical Form… so that all the levels of analysis are connected, including pragmatic contents.

b- Generative Grammar provides us a general theory able to explain:

- variations into a particular synchrony

- variations between two different languages ; it supplies us a new base for cross linguistic comparison and for language contact, in particular between Latin and Greek

- diachronic variations

The main stream of the current research on Latin and Greek linguistics deals with diachronic change, reformulating in particular the classical rules concerning Romance. The parametric hypothesis provides a theory able on one hand to fill the gap between Late Latin and Romance languages, and on the other hand to explain dialectal variations. Several papers devoted to these issues take place in Romance studies or general linguistics. It’s the same in Greek studies : most of the papers I have quoted are devoted to change from Ancient Greek to modern Greek.

The most enjoyable conclusion is that classical languages are no longer considered as exotic by the Generative Grammar. If I can make a prediction, considering the new generation of linguists, I’ll say that within ten years the bibliography will be twice as long.

Retour au § 1 ou Aller à la bibliographie

Retour à l'introduction