Accueil > Projets > Projets passés > Projet ANR "SFA" (Spatial Framing Adverbials)

Description du projet

par webmestre - publié le

Description détaillée en francais


1.1 Global aim

Through a a series of linguistic studies on discourse conducted in the LATTICE team, in cooperation with researchers from Nancy, Toulouse, Caen and Leuven, we have shown that the marks of cohesion contributing to the interpretation of discourse coherence are not limited to connectives and anaphora – unlike what is usually assumed in the literature on that subject ; on the contrary, they include a large family of adverbial markers in sentence-initial position. Studying these adverbials we called « framing adverbials » has led to several review articles (Charolles 1997, Charolles & Prévost eds. 2003, Charolles & Péry-Woodley eds. 2005) and Ph.D.s (E.Terran 2002, Wang 2004, D.Vigier 2004, G.Schrepfer-André 2006).

The project we are presently submitting aims to establish more strongly this hypothesis and widen it through the uncovering of new linguistic and psycholinguistic data, along three lines of research :

- on how framing is done by spatial adverbials, whose role in discourse organization has been at the center of much less linguistic studies than have been temporal adverbials ;

- on factors that can help explaining why the sentence-initial position of adverbial constituents leads to partial grammaticalization (with framing uses) and/or greater grammaticalization (they are then used as connectives) ;

- on the impact the position of concrete and abstract spatial localization adverbs (sentence-initial and framing / sentence-final and non-framing) can have on comprehension.

2.2 Context

In linguistics, taxonomies of extraphrastic coherence or cohesion marks usually distinguish two major groups of relationships, namely referential relationships (anaphora) and discourse relationships (also called semantical, argumentative, rhetorical ; cf. e.g. Reinhart 1981, Sanders, Schilperoord & Spooren eds. 2001). Similarly, studies in computational linguistics (Hobbs 1990), as well as the Rhetorical Structure Theory (RST Mann & Thompson 1988) and the Segmented Discourse Representation Theory (SDRT, Asher & Lascarides 2003) lead to the definition of a series of discourse relationships ; in SDRT, those relationships are calculated at the same time as referential relationships.

There is no doubt that connectives and anaphora contribute to discourse cohesion, in the complex sentence and beyond, and that they help building a coherent representation of what is said. However, other marks work in the same direction.
In :

(1) « En France, oni dénombre, chaque année, une cinquantaine de cas mortels mais oni n’a pas de recensement officiel de la maladiej, parce que les médecins ne sont pas tenus de laj déclarer aux services de santé. » (La Recherche)

« In France, some fifty casualties are reported each year, but there is no official census of all cases, because physicians are not obliged to give notice to health services. »

the prepositional phrase en France calls up a referent, but also establishes a relationship between the different sentences of the extract, since all are understood as referring to facts taking place in France. This relationship of « scope » (cf. Charolles & Vigier 2005, Le Draoulec & Péry-Woodley 2005, Sarda 2005, Schrepfer-André 2005) is different from those marked by connectives (in bold script) and anaphora (in standard script, with subscript) ; it is not reducible to an anaphorical relationship because in France could be replaced by a noun phrase like selon l’OMS, in sentence-initial position, which has the same cohesive power, without being referential.

Thus, some adverbials (i.e. some non-argumental complements), be they referential or not, can exert influence downward from the sentence in which they are located ; however, they only have this capacity in their sentence-initial uses. This constraint is clearly illustrated in (1), since en France placed at the end of the first sentence would not exclude that facts mentioned below take place elsewhere than in France. Besides, the fact that only sentence-initial adverbials can extend their influence beyond the sentence in which they are uttered means globally that because of their position they are already context-activated, and that they can therefore act as some sort of ‘topic’ element (Charolles & Prevost eds. 2003, Charolles 2003, Prévost 2003, Sarda 2003).
As a matter of fact, sequences like (1) most often follow an opening sentence, and function as part of a series :

(2) Les données épidémiologiques varient d’un pays européen à l’autre

En France, on …

En Italie, P1, P2, P3, Pn …

En Espagne, P1, P2, P3, Pn …

Epidemiological data vary from one european country to the next
In France, …

In Italy, S1, S2, S3, Sn …

In Spain, S1, S2, S3, Sn …

Segments governed by a same adverbial constitute blocks (« frames ») which contain all the information indexed by the initial introductive item. The framing adverbials’ contribution to discourse coherence and cohesion derives precisely from this structuring ability. Relationships induced by frame-introducing constructions are fundamentally descending (they are forward-looking) and are globally in opposition to the connexion relationships born out by anaphora and connectives, which are essentially ascending (Charolles, Le Draoulec, Péry-Woodley & Sarda 2005, Charolles 2005).

In psycholinguistics, a major challenge is to be able to describe the nature of the mental representation resulting from the comprehension of a text (Graesser, Millis, & Zwaan, 1997). It is now very widely admitted that readers build not only a representation of the text itself, but also one of the situation it describes. Called mental or situational model, this representation combines the elements extracted from the text with those that have been reactivated in long-term memory (Glenberg, Meyer, & Linden, 1987 ; Kintsch, 1998 ; Zwaan & Radvansky, 1998). The word situational emphasizes the fact that this model is structured according to relationships that are causal, temporal, spatial or other, and that interrelate events and entities mentioned in the text (Rayner & Clifton, 2002).

Among the multiple dimensions represented in the situational model, space is certainly the one which has received the most attention (Bestgen & Dupont, 2003 ; Rinck & Bower, 2000). A series of studies have shown that when someone reads a story, they concentrate mentally on the protagonist’s place in space (Bower & Morrom, 1990 ; Horton & Rapp, 2003 ; Tapiero & Blanc, 2001). The objects near to him are more available for futur reference in the story. The building of such a representation is possible only if the reader gives particular attention to linguistic elements signalling that there is an evolution in the narration along the spatial dimension, as well as along other situational dimensions (Zwaan & Radvnasky, 1998).

We must add however that a series of conditions must be fulfilled in order for the reader to build a precise and rich spatial representation, the main condition being that the spatial dimension has to be relevant for reading objectives (Zwaan & Radvnasky, 1998). However, even when this condition is not fulfilled, the spatial dimension plays an important role in comprehension, as shows Gersnbacher (1990, 1996). She conceives the understanding of a text as an oscillation between two stages : the readers builds a structure and then passes on to another one. The first words of a text are used for the building of the first structure. That structure is then extended by the adjunction of new information, as long as those new elements are consistent with the initial information. A statement bringing new information about what has happened in the same place, at the same time and with a causal link to what has gone on before is easily integrated into the initial structure. On the other hand, when the coherence is no longer present, the reader must build a new structure. This too shows how linguistic devices marking spatial continuity and discontinuity are very useful clues for the reader.


The present project is in keeping with the global hypotheses exposed in 1.1. The linguistic and psycholinguistic studies we designed aim to the finding of new data on how framing adverbials function. For the linguistic part of the project, descriptive and theoretical studies are based on annotated text corpora and statistical research on the markers and constraints under study. The psycholinguistic studies apply methods of investigation enabling us to test in real time the impact of various factors on the understanding of the texts : the place of adverbials, as well as other contextual clues.

The complementarity of those two approaches is obvious, if you keep in mind that linguistic studies already conducted on framing adverbials give credit to the idea that they play a procedural role : they dispatch incoming information in various files, indexed according to the domain of the adverbial (spatial, temporal, enunciative, …). The fact that the linguistics and psycholinguistics teams already have an experience of cooperation on the subject will facilitate their collaboration (Charolles, Colonna, Sarda & J.Pynte, data under observation, Charolles, Bestgen et Piérard, data under way of collection).

2.1. Linguistic studies

2.1.1. Spatial adverbials

We will analyze spatial adverbials introduced by spatial prepositions, static (indicating position : à « at », dans « in, into », en « in, at, into ») and dynamic (initial : de « of, from », depuis « from, since » ; median : par « by, through », à travers « through » ; directional : vers « to, towards » and final : jusque « until, up to »). The approach we adopted is contrastive (comparison between French, English (Carter-Thomas), and Serb (Stosic)) and historical (for French, on the hypotheses presented in section B-2.1.2). Our aim is to classify prepositional phrases formed on the basis of those prepositions according to their capacity to 1) appear in sentence-initial position and 2) introduce discourse frames.

The underlying hypotheses we will test on corpus are that 1) as spatial PPs are allowed more freedom in terms of dislocation, they become more capable of functioning at the discourse level as framing adverbials, and that 2) static PPs are better candidates for these functions than are dynamic PPs, except when the latter appear in sentences whose verb is a perception verb and not a motion verb.

Most studies on the dislocation of spatial complements (e.g. Borillo 1992, 1998, Boons, Guillet & Leclère 1976, Combettes 1996, Fuchs &. Ali 2003) give no attention to dynamic complements, as they represent a particular type of situation. Papahagi hypothesizes in her Phd (2005) that, unlike static complements, dynamic complements are not easily preposed. She argues that they cannot be moved unless a second complement remains in the rhematic part of the sentence. This hypothesis is partially brought into question by the first results of a study on the dislocation of complements in par and à travers (Sarda & Stosic, to appear), showing that those complements, when preposed, are in a great proportion linked to verbs of perception and that they frame the whole scene perceived by the reader. Starting out with this observation, we propose to test on corpus an alternative hypothesis : that the possibility for dynamic adverbials to be placed in sentence-initial position grows as they are no longer used in connection with a motion verb, and as they acquire a discourse function in a domain other than space, in this case in the domain of perception. This hypothesis is in keeping with predictions on category changes made by the theory of grammaticalization (cf. 2.1.2.).
As a complement to this study of spatial PPs capable of playing a role in discourse structure, we will study the changes in unmarked spatial frame (Maeso-Gardeil), especially after verbs of perception (Paul s’approcha de la fenêtre. Des oiseaux chantaient « Paul came near the window. Birds were singing »). Sequences of this kind are well attested in narrative texts, and have not yet been systematically studied with corpus analyses.

2.1.2. Multiple uses and grammaticalization of adverbials

The linguistic studies we have conducted on adverbials globally confirm that :

- as an adverbial is preposed from the predicate, it becomes more capable of playing a role at discourse level ;

- the sentence-initial position is a necessary condition for an adverbial to have a framing potential (i.e. for him to index more than one proposition).

The fact that the sentence-initial position is the position par excellence of connectives likens them to framing adverbials, all the more since we have to deal, in both cases, with adverbial constituents (i.e. constituents that are not arguments).

This similarity raises an important question : why is it that, for constituents that are not well integrated syntactically, the detachement from predication and the movement to sentence-initial position that follows entails different evolutions, namely that some become connectives and others framing adverbials ? These possibilities are respectively illustrated by :

- the adverb justement, which accepts in Modern French integrated uses expressing manner (Paul a très justement répondu que … « Paul very rightly answered that… ») and adverbial sentence-initial uses, where it no longer is gradable and takes on a positive value (A : Tu devrais venir avec nous au cinéma B : j’ai trop de travail A : Justement, cela te ferait du bien « A : You should come with us to the movies B : I have too much work A : precisely, it would do you good ») quite similar to the value expressed by a connective such as mais,

- participle-based phrases like concernant X « concerning X », which presented integrated uses in Old French (Combettes 2003 ; Il a parlé concernant X « He talked concerning = about X ») and no longer have in Modern French any other use than as an adverbial (Concernant X, … ), where they introduce a thematic frame and have a meaning similar to au sujet/à propos de X.

The hypothesis we propose to explore in order to describe these phenomena consists in supposing that the evolution from integrated adverbial to framing adverbial can be described as a grammaticalization process (Hopper & Traugott, 1993 ; Traugott & Heine eds, 1991 ;

Traugott & Dasher, 2002 ; Heine, Claudi & Hunnemeyer, 1991), and that this process is less complete than that leading to a connective (an indication of this lesser degree of grammaticalization being the fact that connectives are more lexicalized than framing adverbials).

We will add to this global hypothesis the idea that the loss of predicative power for adverbials tends to facilitate their moving on to uses as connectives rather than as framing adverbials. This loss is well attested for adverbs such as semblablement, which can be used in constructions such as semblablement à Paul, Robert a félicité Marie « similarly to Paul, Robert congratulated Marie » as well as semblablement, Robert a félicité Marie « similarly, Robert congratulated Marie ». However, when it appears isolated, semblablement needs to lean on a previous statement (Paul a offert des fleurs à Marie. Sembablement Robert l’a félicitée « Paul offered Marie some flowers. Similarly, Robert congratulated her ») and marks a relationship of similarity. With framing adverbials, such alternatives are less common. For instance, even in a context as favourable as Paul est au courant de tout. Selon lui, le départ de Marie est imminent « Paul always knows everything. According to him, Marie will leave soon », it is absolutely impossible to delete the pronoun following selon. Structuring adverbials (such as premièrement « first, firstly ») as well as correlatives (like d’un côté/de l’autre « on the one hand, on the other ») seem to have lost all their predicative power, but actually they are linked to an implicit phrase (je dis premièrement/d’un côté que … « I say first(ly) / on the one hand that… »).

The studies we designed for this project in order to test these hypotheses will bear on :

a) the adverb parallèlement (literaly « in a parallel way »), which has in Modern French various uses : integrated spatial uses (Les bateaux avançaient parallèlement « The ships sailed forward ‘in a parallel way’ »), extrapredicative spatial uses where it is usually followed by a prepositional complement (Parallèlement à la côte, les vagues dessinaient une barre d’écume « Along the coast, the waves formed a line of foam »), temporal prepositional uses (Parallèlement à ce projet, le gouvernement … « As well as this project, the government… »), non-prepositional temporal uses (Parallèlement, le gouvernement … « At the same time, the government… »), and at last non-prepositional ‘additive’ uses where it could be translated by moreover.

b) prepositional phrase adverbials in en N « in N » which can be framing adverbials, be they spatial (En Egypte, … « In Egypt, … »), temporal (En 1930, … « In 1930, … ») or praxeological (En biologie, … « In biology, … ») as well as grammaticalize into connectives such as en effet « indeed », en conséquence « consequently ».

We will analyze the uses of these expressions in Modern French (written and spoken) as well as in Old and Middle French ; for these periods, we will take into account other forms of various origins, in order to shed light on the early stages of their grammaticalization. We will also conduct contrastive studies (Modern French, Modern English and Modern Spanish) in order to see whether we find multifunctional adverbials in those languages as well, or whether on the contrary the values they take in French are expressed differently in English and Spanish.

2.2. Psycholinguistic studies

2.2.1. Incidence /impact of spatial adverbials on text understanding

The studies reviewed in section B-1.2 show how important it is to study in depth linguistic constituents participating in the construction of the situational model. However, one has to admit that very little studies in psycholinguistics have measured the impact of such elements on comprehension. Besides, to our knowledge, only temporal and spatial expressions have been investigated in that perspective (see e.g. Bestgen & Vonk, 2000 ; Caron, Micko & Thuring, 1988 ; Degand, Lefèvre & Bestgen, 1999 ; Millis, Graesser & Haberlandt, 1993 ; Zwaan, 1996).

Moreover, these studies have mostly emphasized the discourse-segmentation marker function (Bestgen & Vonk, 1995) of adverbial expressions, and have mostly neglected the cohesive function that enables them to index several sentences following the one in which they are uttered.

Consequently, we plan to test in the spatial domain the hypothesis according to which a sentence-initial preposed adverbial functions as a kind of indicator which the reader keeps in mind for the processing of its host sentence and beyond, until the appearance of other indicators signalling the end of its scope. According to this hypothesis, spatial framing adverbials should affect processes during reading, and the mental representation resulting from comprehension. We describe below a series of experiments which should enable us to test this hypothesis.
If a framing adverbial guides at least partially the processing of following sentences, then we should observe a slowing down in the reading speed when an action mentioned in a sentence indexed by the adverbial is not or little compatible with that adverbial. What we will do is ask participants to read texts in which three-sentence groups will be manipulated as presented in the following table :

Adverbial Compatible Incompatible
Sentence-initial Dans la cuisine, Luc a allumé la radio.

Il a parcouru le journal.

Il a fait la vaisselle.

In the kitchen, Luc turned the radio on.

He read the newspaper.

He washed the dishes.
Dans le salon, Luc a allumé la radio.

Il a parcouru le journal.

Il a fait la vaisselle.

In the living-room, Luc turned the radio on.

He read the newspaper.

He washed the dishes.
Sentence-final Luc a allumé la radio dans la cuisine.

Il a parcouru le journal.

Il a fait la vaisselle.

Luc turned the radio on in the kitchen.

He read the newspaper.

He washed the dishes.
Luc a allumé la radio dans le salon.

Il a parcouru le journal.

Il a fait la vaisselle.

Luc turned the radio on in the living-room.

He read the newspaper.

He washed the dishes.

The experimental manipulations will be done on the first sentence of a group of three (elements underlined in the table), whereas the reading speed will be measured on the third sentence (the target sentence "Il a fait la vaisselle"), identical in all four experimental conditions.
The first factor we manipulate is the position of the framing adverbial : sentence-initial (and therefore potentially framing) or sentence-final (normally non framing). The second factor we manipulate bears on the compatibility between the spatial adverbial and the action mentioned in the target sentence. The adverbial and the action can be compatible (faire la vaisselle dans la cuisine « do the dishes in the kitchen ») or incompatible (faire la vaisselle dans le salon « do the dishes in the living-room »).

When there is such an incompatibility, the reader must infer that the spatial frame introduced by the adverbial has ended. We therefore hypothesize that it will take longer to read the sentence "Il a fait la vaisselle " after having read "Dans le salon, Pierre..." than after having read "Dans la cuisine, Pierre...". When the adverbial is not introduced in sentence-initial position, but at the end of the sentence, it should not be a framing adverbial. Its scope should therefore be limited to the sentence in which it is uttered. Under these conditions, there should be no incompatibility, and the reading of the target sentence should take the same amount of time, whatever the adverbial might be ("dans la cuisine" or "dans le salon").

Such results would confirm not only the potential forward-labeling role of spatial adverbials, but also the necessity for them to be preposed in sentence-initial position in order to be framing adverbials. If we have such results, we will experiment further along this line. First of all, we will introduce several sentences between the one in which figures the framing adverbial and the target sentence, in order to determine the scope of the framing adverbial. The development of this material will require much care, since it is most important to be sure that the additional sentences contain no clue indicating the end of the framing adverbial’s scope. A second variant will aim precisely to test the function of indicators marking the end of the scope, for various types of framing elements, such as sequence markers like the connective puis « then » (Bestgen & Costermans, 1994 ; Segal, Duchan & Scott, 1991). We think, concretely, that the introduction of puis at the beginning of the target sentence (Puis, il a fait la vaisselle « then he did the dishes ») would be enough to close the spatial frame.

2.2.2. The impact of abstract localization adverbials on text comprehension : enunciative localization adverbials

The situations mentioned in a text can be localized in a 3D space. However, they can also be localized more abstractly in a practical domain, with the help of praxeological adverbials (En chimie / En anglais, ... « in chemistry / in English, ... »). Besides, they can be mentioned in relation to a mediative criterion (means of accessing to the knowledge of the situation), including enunciative criteria (which indicate what speaker is considered as the source of the information). In her Ph.D. on speech-framing adverbials in selon NP, G.Schrepfer-André (2006) has brought to light a whole group of clues indicating that a speech frame that has been opened must be closed. According to her, one of the most reliable frame-closing clues is the use of a pronoun corefering with the NP depending from selon NP ; she presents a detailed study of this phenomenon (G.Schrepfer-André 2005). We propose to study in the same way the impact of enunciative NPs’ position (selon/d’après/pour NP « according to NP ») on text comprehension. Underlying this experiment is the assumption that, following on whether the NPs are sentence-initial or integrated, their impact will not be the same. This is because sentence-initial positioning signals their use as framing adverbials, with structuring power, which is not the case when they are postverbal or sentence-final. This point is confirmed by a corpus study of the way temporal localization adverbials in un jour « some day » function (Charolles to appear), as well as by the results of a psycholinguistic experiment on the sentence-initial vs. sentence-final positioning of spatial adverbials (Charolles, Sarda, Colonna, Pynte, additional data under way of processing).

The experiment we designed bears on enunciative adverbials. We manipulate two elements : their place in the sentence (sentence-initial or integrated), and whether the target sentence contains a pronoun which agrees in gender with the NP that is subject of the sentence indexed by selon/d’après/pour NP, or with the NP which depends* from selon/d’après/pour.

Conditions Selon/d’après/pour NP

+ : sentence-initial

- : integrated
Pronoun coreferring* with :

+ : the NP depending from selon/d’après/pour

- : the NP subject of S1
1 + -
2 + +
3 - -
4 - +

1.Selon Paul, Marie a pris plus de quinze jours de congés dans les Alpes près de Megève. Elle en profite pour s’amuser.

« According to Paul, Marie has gone on vacation for two weeks in the Alps, near Megève. She’s having fun »

2.Selon Paul, Marie a pris plus de quinze jours de congés dans les Alpes près de Megève. Il en profite pour s’amuser.

« According to Paul, Marie has gone on vacation for two weeks in the Alps, near Megève. He’s having fun »

3. Marie a pris, selon Paul, plus de quinze jours de congés dans les Alpes près de Megève. Elle en profite pour s’amuser.

« Marie, according to Paul, has gone on vacation for two weeks in the Alps, near Megève. She’s having fun »

4. Marie a pris, selon Paul, plus de quinze jours de congés dans les Alpes près de Megève. Il en profite pour s’amuser.

« Marie, according to Paul, has gone on vacation for two weeks in the Alps, near Megève. He’s having fun »

We predict that the reading time of the target sentence will be longer under conditions 2 and 4, where the pronoun selects the NP argument of the speech PP, than under conditions 1 and 3 where it selects the subject of the first sentence. This outcome is well attested in linguistic studies (see among others Ariel 1992, Gundel, Heldberg & Zacharski 1993, Lambrecht 1994) as well as in psycholinguistic studies (see among others Vonk, Hustinx & Simons 1992 ; Garrod, Freudenthal & Boyle, 1994 ; Hudson-D’Zmura & Tanenhaus, 1998 ; Nicol & Swinney, 2003) for pronominal anaphora.
It can be explained by the fact that subjects have a higher degree of accessibility than adverbials, and of course than NPs under a PP. We predict that this factor is strongest whatever the position of the prepositional phrase. Concerning conditions 2 and 4, we expect on the contrary that the reading time be longest under the condition 2. This expected outcome, if it is indeed observed, would show that when the speech PP is preposed it opens a frame whose extension must be controlled and whose closing takes time. In conditions 3 and 4, we chose to insert the PP rather than to postpose it in order to avoid, in condition 4, a recency effect on the interpretation of the pronoun.

We will first test the material presented under B-2.2 ; if the excepted results are verified, the stimuli will be modified in order to test the impact of other factors which could influence the observed results.