2.1. Research questions

In his writings on the principle of anticipation, Vandeloise uses the concept of a ‘prospective’ verbal reference place but unfortunately never defines what he understands by the term ‘prospective’. Despite this lack of clarity, it is clear that reference is being made to a location that is distinct from the one occupied by the Figure at the onset of the motion event. The idea of aspectual polarity (Boons, Reference Boons1987), a concept applied in much other work on the topic (Aurnague, Reference Aurnague2011a, Reference Aurnague, Aurnague and Stosic2019; Borillo, Reference Borillo1998; Zlatev, Reference Zlatev and Brown2006), is relevant here. Three types of polarity are posited: initial, medial and final. Initial polarity involves the assertion, followed by the negation, of a basic locative relationship (Aurnague, Reference Aurnague2011a). For example, partir references the existence of an initial basic locative configuration (e.g. ‘il est à la maison’) which is subsequently negated (i.e. ‘il n’est plus à la maison’, so ‘il est parti de la maison’). Final polarity entails the opposite process: there is initial negation followed by assertion of the locative relationship (ibid.). Only a small amount of motion verbs are of medial polarity (Aurnague, Reference Aurnague2011a): as such, this category will not be of interest in what follows. Clearly, motion verbs that have final polarity (aside from those that reference the actual location of the Figure like arriver) have prospective verbal reference places. This contrasts to motion verbs with initial polarity, which do not because they focus on the initial phase of the motion event. As mentioned earlier, the current study will focus on verbs that do not have a prospective VRP. This naturally leads us to verbs of initial polarity. With this in mind, the following research questions arise:

1) With which static locative prepositions (apart from à, sur and dans) can verbs of initial polarity (e.g. sortir, partir, s’enfuir) collocate to trigger the principle of anticipation? What factors enable this to occur?

2) Are there differences between verbs of initial polarity in terms of the prepositions with which they can collocate in order to trigger the principle of anticipation? If so, why do these differences exist?

2.2. Choice of verbs

It is beyond the scope of a single paper to investigate all verbs of initial polarity and their collocational possibilities with all static locative prepositions: this is particularly true when one aims to provide a qualitative analysis of the phenomena observed. As is the case in Vandeloise (Reference Vandeloise1987), the target verbs in the present study require deeper analysis in order to arrive at a better understanding of their compatibility with the principle of anticipation; it is not possible to undertake this type of longer qualitative analysis on a large set of verbs. Yet how should the verbs be chosen? Aurnague (Reference Aurnague2011a) identifies four types of verbs of initial polarity, which are given below. These distinctions have been made on the basis of semantic and syntactic observations: for example, verbs of extended change of initial relation (e.g. s’enfuir) encode speed and the Figure’s desire to resist the location encoded as the Ground: this contrasts to verbs of independent initial change of relation like partir (ibid.). The justifications leading to this four-fold division, based on semantic and syntactic criteria, will not be reproduced or discussed in depth here. Relevant ideas will be discussed in the later analysis. Please see Aurnague (Reference Aurnague2011a) for detailed information about these categories, which are listed below.

I have selected three verbs of initial polarity as reported in work by Laur (Reference Laur1993), Borillo (Reference Borillo1998) and Aurnague (Reference Aurnague2011a). I have added a fourth, disparaître, as it seems to encode a more punctual event than that encoded by partir, which provides a potentially interesting point of comparison. The category ‘double change of relation with initial saliency’ will not be investigated in what follows as these verbs also encode reference to a Goal location: it therefore follows that they have a prospective VRP. According to the principle of anticipation, this makes it highly likely that they will be compatible with prospective readings of all static locative prepositions (as a quick analysis of déménager seems to confirm).

Table 1. Table of motion verbs by aspectual polarity

Before discussing my choice of prepositions for this study, it is relevant to show that each of these verbs is compatible with prospective localization, as evidenced by their ability to trigger prospective readings of à, sur and dans. Vandeloise (Reference Vandeloise1987) has already demonstrated this for partir, so this will not be replicated here. The following sentences show that disparaître can also trigger such readings:

These readings of the verb suggest a clear motion event. However, this is not always the case for disparaître. Hence, one could say ‘il a disparu à/dans Paris’ or ‘il a disparu sur la plage de Manly’: this transition from a place of visibility to one of invisibility does not necessarily require a Figure in motion and the reading is non-prospective in nature. This possibility of a dual reading (i.e. prospective/non-prospective) has also been noted for partir à (Vandeloise, Reference Vandeloise1987). As far as sortir is concerned, prospective readings of these prepositions are also evident:

Again, however, there is the possibility of non-prospective readings of à, as attested by a clause like ‘nous sommes sortis à la station Réaumur’. S’enfuir, however, seems to present a more clear-cut case.

2.3. Choice of prepositions

As far as static locative prepositions are concerned, different sub-categories exist and need to be accounted for in the analysis. Vandeloise notes that all verbs that are compatible with the idea of prospective localization can trigger prospective readings of à, sur and dans, all three of which are topological prepositions. This type of preposition encodes a locative configuration in which the Figure is usually in contact with or contained in the Ground (Borillo, Reference Borillo1998), although each individual preposition is also associated with particular functional relationships (Vandeloise, Reference Vandeloise1986). The second broad category of static locative prepositions is projective prepositions. As less is known about how these types of prepositions work with the principle of anticipation, I will focus on these in the current study. Projective prepositions encode spatial relationships in which the Figure and Ground are at clearly separate locations. This encompasses a wide range of prepositions, and one way of subdividing this category is to consider whether or not the preposition encodes directionally-specific information (Tutton, Reference Tutton2016). ‘Directionally-specific information’ involves the encoding of a direction, understood in terms of the end points of the three spatial axes: ‘up’ and ‘down’ for the vertical axis; ‘front’ and ‘back’ for the frontal axis; and ‘left’ and ‘right’ for the lateral axis. Hence, devant and derrière are directionally-specific prepositions because they reference the directional concepts of front and back; the prepositional expressions à gauche de and à droite de are two other examples. In contrast, près de and à côté de do not encode reference to any one particular direction: they are therefore directionally non-specific prepositions. The question then becomes which prepositions to target in the study. Devant and derrière are natural choices for the directionally-specific sub-category: this is because animate entities prioritize motion along a frontal axis extending out from their intrinsic front and back. As far as directionally non-specific prepositions are concerned, I will focus on près de and à côté de. Although both these prepositional expressions encode proximity, près de can be used to reference location along any spatial axis, while à côté de is often used for relationships along the lateral (left-right) axis (ibid.).

Table 2. Typology of selected French projective locative prepositions

2.4. Clarification of core concepts

In my analysis I draw on concepts from the Cognitive Linguistics literature on Motion events. These are explained in relation to example 14 below.

Figure: the entity that is being localized or which is in motion. This is encoded by le cortège des cheminots.

Ground: the reference entity or entities in relation to which the Figure’s motion takes place. This is encoded by la gare and la Préfecture.

Path: this is a difficult concept to define and is understood in many ways by different linguists (Aurnague and Stosic, Reference Aurnague, Stosic, Aurnague and Stosic2019). For the purposes of the current study I will understand Path in a spatial sense, as the connection between two points in a motion event (Evans and Tyler, Reference Evans and Tyler2004). These two points may not be lexicalized, although both appear in example 14 (i.e. la gare and la Préfecture.) Paths have ‘extension and shape’ (Zlatev, Reference Zlatev and Brown2006), the exact nature of which may not be lexically encoded. In example 14, our extralinguistic knowledge of human beings and processions leads us to assume that the Figure’s Path from the station may not be rectilinear: for example, the group might have to weave their way around numerous streets before ending up at the préfecture. This highlights the importance of extralinguistic knowledge to the interpretation of spatial information (Tyler and Evans, Reference Tyler and Evans2003; Evans and Tyler, Reference Evans and Tyler2004; Naidu et al., Reference Naidu, Zlatev, Duggirala, Van De Weijer, Devylder and Blomberg2018).

Source: the “starting point or locational source” (Evans and Tyler, Reference Evans and Tyler2004, p.251) of a motion event. This is encoded by la gare.

Goal: this is ‘a highlighted end point’ (Evans and Tyler, Reference Evans and Tyler2004, p.258). As far as clauses involving prospective localization are concerned, the Goal is often the same entity as the Ground. It is la Préfecture in example 14.

As far as motion event type is concerned, I will focus on ‘“autonomous”’ as opposed to ‘“caused”’ motion events (Aurnague and Stosic, Reference Aurnague, Stosic, Aurnague and Stosic2019: 6). My analysis concerns events of physical, as opposed to metaphorical, motion. Some examples, like the following, straddle the limit between these two types. Such cases, in which there is nevertheless the idea of a physical movement from one location to another, are included in the analysis.

In contrast, examples which referred to two-dimensional entities, such as icons on screens, were not included in the analysis.

Disparaître raised further questions, since the idea of motion it references is often very subtle. Examples in which the idea of gradual occlusion was encoded, with little reference to any clear movement, were excluded from the analysis.

2.5. Corpus

In order to identify as many uses as possible of the verb and prepositional combinations identified above, I decided to undertake a corpus-based study. Following Laur (Reference Laur1993), I chose to focus on each verb in the passé composé, a perfective tense, in the third person singular form: for example, est sorti devant. A search in Frantext revealed very few examples of the target structures: for example, the combination est parti devant returned 7 results, and est parti derrière returned one. I therefore decided to construct my own corpus using Google. Each combination was searched for using inverted commas to make sure the syntactic sequence was respected: for example, “est parti devant”. The total number of examples returned across all verb/preposition combinations was 1,385. For each combination I went to the final page of results and worked back to page 1, going through each example and analysing whether or not it constituted an example of prospective localization. Examples that were repeated were counted only once. Those that were ambiguous between prospective or non-prospective interpretations, as well as those that were non-sensical, were discarded. An example of the latter type of example is given below.

Also excluded were examples that lacked context: these included examples from grammar worksheets, decontextualized examples on websites like ReversoContext, or the following single-sentence translation (from an oriental language into French) that appeared in an academic journal.

I also excluded examples that appeared to be produced by automatic translators.

Needless to say, there were many irrelevant examples but this process enabled me to create a corpus of relevant data. Obviously, Google searches draw data from a range of sources and the different types of text from which examples are drawn need to be acknowledged. I therefore devised a list of 5 categories to account for the source of each example. These categories are presented below.

1. 1. Media outlets

This category contains examples from the websites of newspapers, magazines, and media organizations: for example, the newspaper Le Parisien.

1. 2. Organizations (including educational institutions)

Included in this category are examples from business websites and organizations representing collectives. For example, the motorbike website motorsport.com or the website of the Church of Jesus Christ (churchofjesuschrist.org).

1. 3. Literature (including legal documents)

This category includes examples from books and legal documents, such as court transcripts.

1. 4. Blogs and messages on websites

These examples are essentially personal commentaries and are drawn from personal blogs and user comments on websites.

1. 5. Other

This category groups together the minority of examples whose source cannot be ascertained or which do not fit into one of the four categories above.

This coding system allowed me to differentiate between examples that were likely to have undergone an editing process and those that hadn’t. For example, examples from categories 1 and 3 have almost certainly undergone some sort of review process for linguistic acceptability (except for direct-speech quotations); this is probably the case for most examples in category 2 as well. However, examples drawn from category 4 are unlikely to have been scrutinized in this manner and the register of the texts in which they appear tends to be more informal. The breakdown of examples across these categories is given above in Table 3.

Table 3. Source of corpus examples

3.1. partir and disparaître

In section 1 it was shown that the principle of anticipation was not activated by partir devant (see example 6). In fact, ‘for this verb, the principle of anticipation applies only to à, sur and dans’ (Vandeloise, Reference Vandeloise2017, p.5), and Vandeloise (Reference Vandeloise1987, p.104) mentions specific static prepositions that don’t readily acquire a prospective reading with partir: namely, derrière, à gauche, à droite and près de. This, he claims, is because partir requires a more distant point of view from which to view the Ground, whereas the use of these prepositions is more feasible when the Figure and the Ground are in the viewer’s visual field (1987, pp.104-105). However, this reasoning overlooks the fact that derrière can also be used to describe spatial configurations in which the Figure is visually inaccessible to an observer (Vandeloise, Reference Vandeloise1986), and thus outside of their visual field. Hence, if I described someone as being derrière la porte, the understanding would be that this person is closer to the side of the door that is currently not in view. Such visual inaccessibility complements the idea of a distal point of view associated with partir, because if an object is far away from a viewer it may well be harder to see. With this in mind, we can hypothesize that derrière may, in fact, trigger the principle of anticipation when used with partir.

The implication in this example is that the coffin leaves from an (unspecified) Source and follows a Path that ends in a Goal location beyond the location associated with the curtain. No single lexical item encodes this Path: rather, it is pragmatically inferred thanks to the collocation of partir and derrière, which are complementary in terms of the notion of visual accessibility to an observer. The frame of reference (understood in the sense of Levinson, Reference Levinson2003), is also an important variable to consider. When speakers encode location using devant or derrière, they appeal to either the intrinsic front/back properties of the Ground (in which case an ‘intrinsic’ frame of reference is applied), or they attribute the Ground such properties via the mirror-like rotation of their own (or another entity’s) front and back onto the entity (in which case a ‘relative’ frame of reference is applied). It is the latter strategy that is at work in example 22 – as is arguably the case in all prospective uses of the combination partir derrière. I will return to this point later.

Partir also collocates with derrière to express motion events in which the Ground does not appear to function as the Goal of the Motion event. In such cases I considered that the principle of anticipation had not been activated. These examples are noteworthy as in almost all cases they encode a sequence in which an animate Figure follows an animate Ground. This reading lends itself well to descriptions of pursuit in different contexts, from criminal justice to car racing.

A comparison of the Ground entities in example 22 on the one hand, and in examples 23 and 24 on the other, reveals a key difference in function. Le rideau triggers the conceptualization of an object that serves as a physical boundary between two separate spaces. This was the type of Ground found in all examples of prospective uses of the preposition. In contrast, le conducteur and moi do not function as boundaries but rather as reference points for the Figure’s ongoing Path; crucially, both Figure and Ground are understood to operate within a single, undivided spatial regionFootnote ⁷ . The majority of Grounds (29 out of 33) that occurred with non-prospective uses of partir derrière were entities in motion such as this one. A pattern therefore emerges: immobile inanimate entities that have clear boundaries tend to occur with prospective uses of partir derrière, whereas entities in motion tend to function as the Ground in non-prospective uses. This suggests that the variable of motion is key to separating prospective and non-prospective uses of partir derrière. In addition, all non-prospective examples of partir derrière conveyed the idea of a Figure starting a motion event (Aurnague, Reference Aurnague2011b), a phenomenon also noted in all non-prospective examples of partir devant.

The use of devant in such examples can also be interpreted as a metaphorical reference to the temporal domain. Hence, as far as the current example is concerned, everyone else departed before the speaker did, meaning that he got off to a slower start. While acknowledging this possible reading, I nevertheless included such examples in the analysis if they also held true in a spatial sense: for instance, in example 25, we understand that tout le monde advanced relative to the front of the speaker.

Devant, as predicted by Vandeloise, is almost exclusively used in examples of non-prospective localization: just one example of prospective localization was found. As far as près de is concerned, almost all examples encoded instances of prospective localization: this contradicts Vandeloise (Reference Vandeloise1987, p.104). In the majority of cases the Ground was a noun referencing a city, a town or a border: that is, an entity that could be considered geographical in nature. This was the case in 21 of the 32 examples of prospective uses.

Six of the examples encoded movement into a spiritual realm, referencing God as the Ground entity.

Vandeloise’s (Reference Vandeloise2017) distinction between material entities and spatial entities is useful at this juncture. Spatial entities are concepts like countries, the world and the universe (ibid.). They possess boundaries that are harder to define than those of material entities (ibid.), perhaps owing to their more abstract nature. While a Ground like son Maître, in reference to God, ostensibly refers to a being, it is also a metonymic reference to another dimension, the boundaries of which are unknown. This is arguably a type of spatial entity itself, perhaps one of the most abstract to be conceptualized by the human mind. It is logical that references to spatial entities, both celestial and otherwise, are introduced by près de but not derrière, because the former does not require the entity to possess (or be attributed) any particular spatial property. As far as à côté (de) is concerned, this prepositional expression “indique la proximité dans l’espace” (Larousse); notably, it is this concept of proximity, as opposed to any reference to a side attribute, that is foregrounded when a spatial entity occupies the role of Ground. Hence, in the following example we understand that the Figure has moved to an area close to Grenoble.

Only six non-prospective examples with à côté de were noted. However, in contrast to the prospective uses, material entities (e.g. chez moi, moi), as opposed to spatial entities, were encoded as the Ground. This is shown by the use of votre table in the following example.

Table 4. prospective/non-prospective reading of selected prepositions with partir

Disparaître

Disparaître encodes an action in which an entity ceases to be perceptible, either visually, audibly, or in terms of smell (Larousse). The pivotal role of the speaker/observer in this process is clear, thus creating a parallel with partir (see above). However, compared to the latter verb, the results for disparaître are far starker: that is, the corpus suggests that collocation with each of the four prepositions leads to a necessarily prospective or non-prospective interpretation, but not both. As such, all examples with devant triggered non-prospective readings of the preposition.

In contrast, interpretations of derrière were uniformly prospective.

The difference between the two prepositions in terms of access to perception may explain these results. Specifically, derrière encodes a viewer’s lack of perceptual access to the Figure (Vandeloise, Reference Vandeloise1986) whereas devant does not. Given that disparaître encodes a process in which the Figure ceases to be perceptually accessible to the viewer, the Ground NP introduced by derrière is logically interpreted as a Goal given the reading of perceptual inaccessibility associated with the preposition. In contrast, the idea of access to perception associated with devant aligns with the initial phase of the motion event encoded by disparaître in which the Figure can be perceived by the viewer: this entails that the Ground introduced by the preposition is interpreted as a Source. Yet these corpus results and accompanying explanations are challenged in the face of the following example, found in a novel outside the corpus.

Leaving aside the fact that disparaître is in the passé simple, the key factor here is the prospective interpretation of devant. The reading of movement beyond the viewer’s visual field is facilitated by the immediately preceding context, in which the Figure’s Path is clearly established in a direction away from the viewer. The use of an intrinsic frame of reference to interpret devant means that the idea of visual inaccessibility is not hampered: that is, it is the intrinsic front of the calandre which is referred to, and this is understood to be oriented away from the viewer. This shows that a prospective use of devant is, in fact, possible with disparaître provided that the context helps to shape a reading in which devant no longer expresses the idea of perceptual accessibility to the viewer.

As far as non-directionally-specific prepositions are concerned, all examples with près de and à côté de had non-prospective readings.

This is unsurprising if the two prepositions are understood to be more felicitously used when the viewer has perceptual access to the Figure and Ground, as has been suggested previously for près de (Vandeloise, Reference Vandeloise1987, pp.104–105). Spatial entities accounted for almost all Grounds introduced by près de, which suggests that the preposition might be used most frequently in descriptions of larger scale space. This stands to reason as when describing location in small scale space it is more communicatively helpful to localize a Figure in relation to specific properties of the Ground (as mediated by prepositions such as devant, derrière, etc), as opposed to using a coarser-grained expression of proximity which fails to narrow the search domain to a specific directional axis.

Table 5. prospective/non-prospective reading of selected prepositions with disparaître

3.2. sortir

Like partir, sortir expresses the departure of a Figure from a point of origin but the verbs differ as far as the concept of interiority is concerned. Specifically, partir encodes no reference to the idea, thus suggesting that the underlying locative relationship encoded by the verb is, in most cases, the localising use of the preposition à (Aurnague, Reference Aurnague2011b). In contrast, sortir encodes the Figure in a relationship of containment (Aurnague, Reference Aurnague2011a), thus suggesting that the underlying relationship is aligned with the preposition dans. The movement of the Figure out of a container-like space is expressed by the use of sortir, derrière, and a Ground-encoding NP that references a material entity or a boundary marker (e.g. la ligne de but).

Interestingly, nearly half of these examples came from sport-related writing. These texts represent a variety of sources, including media outlets, professional organizations, blog posts and messages on websites. Just as it was the case for partir, non-prospective uses of sortir and derrière were also attested.

This suggests a similarity between sortir and partir to the extent that both verbs can enable prospective or non-prospective readings of derrière. In addition, the Grounds used with non-prospective uses of sortir derrière were all entities in motion (most of which were human, as in examples 37 and 38); the use of entities in motion as the Ground was also noted for non-prospective uses of partir derrière. However, there is also a fundamental difference between these two verbs. It was mentioned earlier that partir suggests that an observer has a distant viewpoint of the Figure. In contrast, sortir can be used to express movement into the speaker’s field of perception – usually the visual field – thus enabling a closer viewpoint of the Figure. Imagine a scene in which the police have arrived in front of a house to arrest a burglar.

The use of sortir in example 39 does not necessarily mean that the Figure will move closer to the speaker’s location in terms of absolute distance: after all, the robber may walk out a side door and end up further away from the speaker’s location than they are currently. Rather, it calls for the Figure to appear in the speaker’s field of vision. In contrast, partir conveys the idea of movement away from the speaker, thus accounting for the incongruity of the second sentence above within the specified contextFootnote ⁸ . The idea that sortir is compatible with the idea of entry into the speaker’s perceptual field is shown by several examples in which the use of de is elided before derrière. The Ground functions as the Source, rather than the Goal, of the motion event, and the understanding is that the Figure moves into the visual field of the speaker, who constitutes the implicit Goal. These examples were counted in the analysis as examples of prospective localization.

This result provides a contrast to the common use of sortir to encode a movement of the Figure out of a container-like space: instead, the Figure moves out of a non-container-like location and into the perceptual field of a human observer, itself perhaps conceived as a container-like space. The effect may be one of surprise for the observer, who had not anticipated the appearance of the Figure (example 42). This idea of movement into the perceptual field of the observer is also attested in examples with devant. However, the Ground in such examples now encodes the (human) Goal, as opposed to the Source of the motion event.

In all cases apart from two, est sorti devant involved a prospective use of the preposition.

As far as the prepositional expression à côté de is concerned, nine of the ten examples in the corpus were prospective uses. Each of the entities encoded as the Ground in these examples was an immobile material entity.

An analysis of uses of près de did not return any clear results of prospective localization. There were two reasons for this. Firstly, the majority of the examples included uses of près de as a synonym for presque: they were therefore irrelevant to the study. Secondly, the semantic combination of sortir and près de leads to ambiguous interpretations. That is, sortir is quite punctual in terms of the event it encodes: the Figure leaves one space to arrive in an adjacent one, meaning that the non-directional relationships of proximity encoded by à côté de and près de may be valid both before and after the ‘going out’ event has been completed.

In contrast, a spatial relationship encoded by devant or derrière may only be true once the Figure has moved in or out of a container-like Ground (e.g. examples 41–45), thus enabling a stronger association with prospective localization.

Table 6. prospective/non-prospective reading of selected prepositions with sortir

3.3. s’enfuir

S’enfuir (‘flee’) encodes the motion of an entity away from a Source along with the concomitant desire of this entity to evade control. The Figure thus attempts to move out of a zone that equates to the search domain (“domaine de recherche”; Vandeloise, Reference Vandeloise1986) of the pursuing entity that seeks to assert control over the Figure. The speaker adopts the perspective of an observer within this search domain, with the understanding that the Figure moves away from the observer’s location in its bid to avoid detection.

In this example the observer is not necessarily facing the front of the shops such that the Figure escapes around to the intrinsic ‘back’ of these buildings. In fact, the observer could be looking at a small row of shops side-on, thus deictically attributing a front and back, the latter of which would be the furthest side away from the observer’s location. This suggests that there may be a deictic component in the activation of the prospective interpretation of the sequence s’enfuir derrière: specifically, the front/back properties of the Ground are attributed via a relative frame of reference in which the oriented observer’s front and back properties are rotated onto the Ground. However, examples like the following challenge this hypothesis.

From a shopper’s point of view, the ‘front’ of a counter is the side we face when we go to make a purchase. While this ‘front’ may initially have been attributed via a relative frame of reference, it has since become a standard property that is easily identified thanks to multiple indicators: the location of the counter on the shop floor which allows less space behind than in front of it, the register facing away from the customer, the products for sale facing towards them, etc. Example 50 might describe a perspective in which the observer faces this conventional, customer-focused front. However, it could also plausibly describe a situation in which an observer, standing in a room immediately beyond the counter, looks through a doorway onto the ‘back’ of the counter, this ‘back’ being understood within the context of the conventional customer-focused interaction just mentioned. The pharmacist might be chased by their assailant through this back room and out into the pharmacy, where they end up behind the counter. In this case, derrière would not refer to a back property attributed by an oriented observer through a relative frame of reference: rather, it would appeal to the conventional front/back distinction discussed above. In addition, the Figure and the Ground might still be visible to the observer even once the Figure has reached the location associated with the Ground. This suggests that the prospective reading of s’enfuir derrière expresses a motion event in which the Figure moves beyond a search domain that is associated with the observer, and thus out of their control. The reading of a lack of visibility, which is complementary to the notion of an absence of control, may also entail. In such cases it will be prompted for by extralinguistic knowledge relating to the objects functioning as the Figure and the Ground.

Our knowledge of cliffs as large, static opaque entities and bandits as animate, mobile beings prompts for the reading of a lack of visibility in example 51. The layperson’s understanding of physics which views the sun as a mobile entity that rises in the East and sets in the West informs our understanding of the sun’s path behind the hills in example 52. Again, our knowledge of the latter as opaque entities capable of occluding the sun contributes to the reading of visual inaccessibility. In all prospective uses of s’est enfui derrière with the exception of one, the Ground is a material entity that is understood as a boundary separating the Figure from a Goal location, the latter of which is inaccessible to the viewer. The use of this type of entity as the Ground is logical as it lends itself well to the role of a stable shield that protects the Figure – often by hiding it. However, prospective readings are not noted with devant: the idea of accessibility to perception encoded by the preposition means that the Ground is more logically considered the Source, as opposed to the Goal, of the motion event.

As far as directionally non-specific prepositions are concerned, uses of s’enfuir are harder to find. Just one prospective reading of the collocation s’est enfui près de was found (example 54), while eight non-prospective uses were recorded, one of which is presented as example 55.

The action of fleeing implies a desire to be out of reach of a pursuer. As such, it makes more sense to take refuge behind an entity – as opposed to near or next to it – due to the additional layer of protection provided by the entity in its separation of the pursuer and the pursued. In addition to this, there is the additional benefit of possible invisibility. It is therefore logical that far more uses of s’est enfui derrière were found than s’est enfui à côté de or près de.

Table 7. prospective/non-prospective reading of selected prepositions with s’enfuir