2.1. General issues

HSS was first presented by Zlatev (Reference Zlatev1997), as a framework aiming to unify cross-linguistic research in the expression of space, including static spatial relations, deixis, and motion events. The name of theory stems from the claim that a spatial situation, either static or dynamic, is expressed by a whole utterance, and not by specific form classes like prepositions or verbs. Spatial situations, including motion events, are non-linguistic events, and should be analyzed accordingly (see Section 2.2). Since human knowledge of space and motion is “grounded in pre-linguistic bodily experience” (Zlatev et al., Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021, p. 5), there should be general similarities in how this is expressed in human languages. This provides the basis for a set of general semantic categories, like Path and Manner, which are presumably universally present, but “conventionalized in language specific ways” (ibid) both in terms of how they are expressed and in terms of what values they assume (see Section 2.3). These are mapped onto form classes in three different ways: (1) one-to-one, in a compositional matter, (2) several-to-one, resulting in patterns of conflation, as noted by Talmy (Reference Talmy1991), and (3) one-to-several or distributed, as noted by Sinha and Kuteva (Reference Sinha and Kuteva1995).

The many-to-many relation between form classes and semantic categories results from the “relative meaning-holism predicted by the theory” (Zlatev et al., Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021, p. 8). In line with Goldberg (Reference Goldberg1995, Reference Goldberg2006), Slobin (Reference Slobin, Niemeier and Dirven2000), and Feist (Reference Feist2010), it is assumed that the meaning of individual motion verbs contributes to the meaning of an utterance, but does not predetermine it. For instance, even though most researchers would agree that there is a certain degree of directionality included in the semantics of the verbs corresponding to the English verb run (e.g., Cifuentes Férez, Reference Cifuentes Férez2010; Pedersen, Reference Pedersen, Gries and Yoon2016), it does not follow that every construction having such a verb as a predicate would express the idea of change of location. For example, one can run on a treadmill.

One aspect of such constraining is the phenomenon of covert expression (Blomberg, Reference Blomberg2014; Zlatev, Reference Zlatev1997; Zlatev et al., Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021). It relates to meanings that are not explicitly coded, but stem from the interlocutors’ shared understanding of the situation described or knowledge of the world (see more in Section 3.4). The inclusion of covert expression in the analysis helps to overcome the notorious Pragmatics–Semantics divide. It also facilitates the analysis of how meanings are shared across languages regardless of whether they are linguistically coded or not.

2.2. Taxonomy of motion situations

Three parameters allow us to distinguish between different types of motion situations,Footnote ² irrespective of how they are expressed or construed in language: (1) translocative versus non-translocative, (2) bounded versus unbounded, and (3) uncaused versus caused (Zlatev et al., Reference Zlatev, Blomberg, David, Evans and Chilton2010).

Translocation is “the continuous change of an object’s average position according to a spatial frame of reference” (Zlatev et al., Reference Zlatev, Blomberg, David, Evans and Chilton2010, p. 394), while in non-translocative motion there is no change in this position, with respect to the spatial frame in question.

Three assumptions follow from this definition. First, translocation by default involves the whole Figure and, therefore, motion of some parts of it is not considered as instances of translocation. Consider, for instance, an act of waving in She waved as a situation of motion that only involves the Figure’s hand, or an act of the change of posture, such as sitting down or standing up. Second, motion is considered translocative only if it is continuous. Therefore, instances of full stasis or appearing out of thin air or disappearing exemplified in sci-fi fiction would not be considered as translocative motion and, in fact, lack motion altogether. Third, translocation can only be attested if an explicit frame of reference is given. If one, for instance, is seated on a plane that is taking off, one moves up in reference to the core of the earth and yet remains still in reference to the plane. Thus, depending on the frame of reference invoked, the speaker can construe an event as translocative or non-translocative.

Boundedness involves the presence vs. absence of a definitive state transition involved in the motion situation. For instance, while going to a store has such a point, going upward does not.

Causation concerns whether the figure is perceived to be moving under the influence of an external cause or not. Since the present study focuses on uncaused motion events only, this parameter will not be relevant.

The parameters presented above give grounds to identify four types of uncaused motion situations: bounded translocative, unbounded translocative, bounded non-translocative, and unbounded non-translocative, which are exemplified in Table 1.

Table 1. Taxonomy of motion situations and English sentences (in past tense) that can be seen as expressing them

It is important to note that these parameters have been elaborated on the basis of an analysis of motion as pre-linguistic experience. However, when they need to be expressed in language, there is necessarily the effect of linguistic construal, which could, for example, present an unbounded event using the category Path, and thus as linguistically bounded.

2.3. Semantic categories

In its most recent versions, HSS claims the existence of the following universal semantic categories, hypothesized to be necessary and jointly sufficient for the expression of space in all human languages (e.g., Zlatev et al., Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021).

1. 1. Figure (F): The focal entity that is presented as located or as undergoing the process of translocation.

2. 2. Landmark (LM): One or more physical entities (typically expressed by nominal phrases), in relation to which the translocation of the Figure can be specified.

3. 3. Region (R): An area of space usually defined in relation to a Landmark. This category can take values like INSIDE, OUTSIDE, ABOVE, BELOW, and BESIDE. These values are quite often defined by the characteristics of the LM.Footnote ³

Consider the following examples (3) and (4). In (3) R acquires the value INSIDE since the LM is “container-like” (e.g., there is a potential for the F to be inside it at the end of translocation). In a similar way, in (4) R profiles not only the proximity of F to LM, but also an additional property of the LM: that it has a side easily identifiable as front.

It should be kept in mind, however, that the ways in which this category carves up space are somewhat language specific, with different languages profiling its values more or less frequently. Fig. 2 schematically represents some properties of the LM that can be profiled by Region.

Figure 2. Topological subspaces (adopted from Becker et al., Reference Becker, Carroll, Giacobbe, Perdue and Porquiez1997).

The ways the concepts schematically portrayed in Fig. 2 are profiled in a specific linguistic expression are somewhat language specific. For instance, Becker et al. (Reference Becker, Carroll, Giacobbe, Perdue and Porquiez1997) observe that even the concept of interior subspace (corresponding to Region:INSIDE in HSS terminology) is rather flexible since it can be applicable to LMs of different shapes. Compare, for instance, walking into the forest vs. walking into a building or swimming out of the harbor vs. getting out of a pool. The LM in question can be profiled as two- or three-dimensional as is the case with French and Swedish translations of an English sentence He walks in the street. While the French expression in (5) profiles the LM as three-dimensional and Region assumes the value INSIDE (i.e., included in the interior space of a bounded LM) accordingly, the Swedish expression in (6) profiles the same LM as two-dimensional and the value Region assumes is ON (i.e., in contact with the surface of the LM).

Moreover, the actual boundary, as construed linguistically, may not refer to anything boundary-like in the real world (as is the case with swimming out of the harbor).

In a similar way, the concept of proximity (values NEAR and FAR that Region can assume) relies on the speaker’s understanding of what Becker et al. (Reference Becker, Carroll, Giacobbe, Perdue and Porquiez1997, p. 23) call neighboring subspace – “a sort of ‘sphere of influence’” within which interaction with the LM is possible. They also note that the attribution of a neighboring subspace to a LM presupposes its boundedness since it is conceptualized essentially as the space that adjoins the boundary but does not include it (Becker et al., Reference Becker, Carroll, Giacobbe, Perdue and Porquiez1997, p. 23).

1. 4. Motion (M): Perceivable actual, as opposed to fictive, or non-action motion (Blomberg, Reference Blomberg2014). This is neutral with respect to the parameters described in Section 2.2, so any of the sentences in Table 1 would have a positive value for M.

2. 5. Frame of Reference (FoR): This generalizes the well-known proposal of linguistic frames of references by Levinson (Reference Levinson2003) to dynamic situations in a three-dimensional plane (Zlatev, Reference Zlatev and Cuyckens2007). There are three different kinds of FoRs:

3. (a) Object-centered (OC): Defined by means of one or more Landmarks as in (7), (8):

1. (b) Geocentric (GC): Involving relatively fixed (“absolute”) reference points or axes (such as UP, DOWN, WEST, EAST, etc.), for example, (9) and (10):

1. (c) Viewpoint-centered (VC), which covers motion situations anchored at a viewpoint that could be a deictic center, for example, (11):

but not necessarily, as in: When you reach the crossing, turn to your right.

1. 6. Direction (D): Motion along a vector specified by an axis provided by one of the three kinds of FoR, given above. Direction is involved where the motion is linguistically construed as unbounded, meaning that it could, in principle, (though not practically) unfold forever (as in (12) and (13)). Consider (12) which occurs as the first utterance of the first narrative produced by a participant. There can be no transition point or change of state since the relationship between the F and the LM does not change and any other LM is absent. What is expressed here is not a case of translocation in purely topographic terms, but rather that the motion of the F is directed towards the functional space of the speaker. The situation is, therefore, construed as translocative and unbounded, due to the use of the deictic verb. The construal is from the speaker’s viewpoint, which is very different from the previous examples (3) and (4), where LM was functioning as the reference point and where one can see not just an average change of location within a given FoR, but also a definite transition (from outside to inside, or from far to near). We may, of course, infer that, at some point, the motion will come to an end, but this is our knowledge of the world informing the inference, while the semantics of the verb does not allow to imply this.

Note that prepositions vers (French: ‘towards’) and mot (Swedish: ‘towards’) have been considered to be expressions of Direction in the version of HSS in Blomberg (Reference Blomberg2014), but as expressions of Path in Zlatev et al. (Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021). The decision was made by the authors based on these expressions being temporarily bound (in terms of Croft, Reference Croft2012) and on the grounds of the state transition “strongly implied” (Zlatev, personal communication). However, the coding adopted for the present study follows Blomberg (Reference Blomberg2014) rather than Zlatev et al. (Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021). To demonstrate this point with an example, it seems reasonable to suggest that there is indeed a difference between such expressions as to (instantiating Path) and towards (instantiating Direction). For instance, when giving the driver instructions, one can say “Drive towards B, but turn left one block before you reach it”. The same sentence would no longer be acceptable were we to substitute towards with to, since it would contain two contradictory propositions – one cannot reach B and not reach it at the same time. The same logic seems to hold for the distinctions between vers ‘towards’ and jusqu’à ‘all the way until’ in French and mot ‘towards’ and till ‘to’ in Swedish.

1. 7. Path: this is limited to linguistic construals of bounded translocation, where a Landmark and Region are recruited to specify the BEGINNING, MIDDLE (movement of the F towards and then past the LM), and/or END of the translocation, as in (14). Note the change of the FoR (and, therefore, the perspective on the situation) from Viewpoint-centered to Object-centered, where the LM (grocery store) is recruited to specify the location of the F at the end of the translocation.

The Path:END along with the Region change from FAR to NEAR LM is expressed through the preposition till (‘to’).

The final three categories are not strictly speaking necessary, as (trans)location can be fully expressed only using the first seven. They do, however, add to the “color” of the motion event and are very often included.

1. 8. Manner: specifications of various aspects in the way motion is conducted, including (a) bodily locomotion (jump); (b) vehicle (ride); (c) medium (sink); (d) velocity (rush); (e) attitude (mindfully); and possibly other kinds.

2. 9. Shape: the geometrical form of the trajectory followed by the F (e.g., to zigzag). In some of the motion event literature, this is included under Manner, but as the latter is already a very broad category, it is best to distinguish it.

3. 10. Cause: the initiator of the movement and the nature of the causal force (e.g., push vs. slide).

3.1. Participants

Two groups of participants – native speakers of French and native speakers of Swedish – took part in the study (see Table 2 for information about the participants).

Table 2. Information about the participants

The participants were recruited through their respective universities (Lille University for the French participants and Lund University, Stockholm University, Uppsala University for the Swedish participants) as well as via social media (e.g., Facebook groups).

It should be noted that, with the exception of two participants in the Swedish group and two participants in the French group, the speakers recruited for the study spoke English fluently in addition to their native language. Some participants in the Swedish group also spoke German and a few reported knowledge of Spanish at the beginner level. In the French group, some participants also spoke Spanish or Italian. While knowledge of foreign languages, especially on advanced levels, could have influenced the produced description, there was no way to assess the strength of such effects in a principled way. Thus, the topic was left outside the purview of the investigation.

All participants signed (digitally) an informed consent form and were compensated for their participation.

3.2. Materials

The data for the present study has been collected using an elicitation tool developed by the author, based on the considerations described below. The stimuli used were six short (between 0.37 and 0.54 min) cartoons each having a plot and presenting various types of motion situations (see Appendices B and C).

Previous studies that used HSS as a theoretical framework tended to use sets of video clips each presenting one situation. In particular, Blomberg (Reference Blomberg2014) and Fagard et al. (Reference Fagard, Zlatev, Kopecka, Cerutti and Blomberg2013) used the Trajectoire tool (Ishibashi et al., Reference Ishibashi, Kopecka and Vuillermet2006), while Zlatev et al. (Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021) have used a novel elicitation tool based on the same principles (a set of individual video clips) but developed in accordance with the taxonomy of motion situations put forward in Zlatev et al. (Reference Zlatev, Blomberg, David, Evans and Chilton2010) (see Section 2.2). In contrast, Naidu et al. (Reference Naidu, Zlatev, Duggirala, Van De Weijer, Devylder and Blomberg2018) used narratives elicited with the help of a picture book “Frog, where are you?” (Mayer, Reference Mayer1969), otherwise known as frog stories. Both techniques allow to control for the topics covered, but have additional advantages and limitations. Sets of individual clips allow more control over the number of utterances produced and the types of descriptions so that each situation is described by all the participants, with some degree of variation on how it is linguistically construed (Zlatev et al., Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021). Pictorial support allows for more creativity and flexibility in verbal descriptions. Further, presenting a narrative is arguably a more natural mode of language use in contrast with verbal descriptions of individual video clips. The resulting samples are narratives, which are the type of data that differs from sets of individual utterances.

The tradition of using frog stories in motion events research goes back to Slobin (Reference Slobin1991, Reference Slobin, Niemeier and Dirven2000, Reference Slobin, Strömqvist and Verhoeven2004)) who suggested that typological differences in event framing “have consequences for the connected discourse” (Slobin, Reference Slobin, Niemeier and Dirven2000, p. 107). He further noted that “languages differ systematically in rhetorical style – that is, in ways in which events are analyzed and described in the discourse” (Slobin, Reference Slobin, Strömqvist and Verhoeven2004, p. 5, emphasis added). Moreover, using the narratives accords with the ideas laid out by Fauconnier and Turner (Reference Fauconnier and Turner2002, p. 97), who suggest that a study of language and cognition should be based on methods that “extend to contextual aspects of language use and non-linguistic cognition. This means studying full discourse, language in context, inferences actually drawn by participants in an exchange, applicable frames, implicit assumptions and construal, to name just a few.” Analyzing narratives allows to do just that.

At the same time, static images are not the best medium to convey dynamicity of motion. In addition, “Frog, where are you?” has a limited number of motion situations that does not allow to thoroughly investigate linguistic descriptions of all situation types (see Section 2.2).

These considerations have paved the way to the development of a novel elicitation tool that would allow to elicit narratives that are rich in different types of motion events. In addition, video stimuli are more suitable for depicting situations of motion than static pictures.

Moreover, special care was taken to ensure that the stimuli were interesting and engaging to motivate the participants to form detailed narratives. Hopefully, this allows to collect data that are as close as possible to the situated and spontaneous character of everyday oral speech production.

The exact description of situations depicted in the video stimuli can be found in Appendix C. Note that, even though sometimes the situations depicted are static rather than dynamic, at least 102 situations are potentially interpretable as motion events. The bulk of the situations were designed to elicit translocative event descriptions. This is because this type of motion events is inherently diverse and adding more scenes of this type allowed to manipulate a few factors that might have had an impact on the type of linguistic construction chosen. In particular, the stimuli were designed to portray different ways of motion (walking, running, crawling, tip-toeing, swimming, hopscotching, etc.), different types of Figures (humanoids, animals, liquids, etc.), different types of Landmarks (vehicles: a car, a boat, a spaceship; buildings: a house, a supermarket; elements of natural landscape: a hill, a cliff, a tree, etc.), different directions of movement (up, down, etc.), different types boundary crossing (entering, exiting, crossing, etc.).

Still, it should be noted, that eliciting free narratives is not without its limitations. This method allows very little control over what situations forming the plot canvas of each story would be described in the narrative, which resulted in some situations being more frequently described than others.

3.3. Procedure

The data were collected between spring 2020 and fall 2021. For the reasons connected to the epidemiologic situation, the elicitations were performed via Zoom.

First, a participant was asked to provide information about their sex, age, and possible knowledge of foreign languages by means of filling out a short survey. This was followed by an instruction to watch a video carefully to be able to present it later. Each video was shown twice. Finally, the participant was given the instruction (in their native language): Please, tell me what you have seen in this video and try to be as precise and detailed as possible.

After the procedure, the participant had an opportunity to ask questions and obtain detailed information about the study. It was also an opportunity for the researcher to ask questions of clarification (see, for instance, the discussion of the Swedish hoppa/gå i ‘jump/go inside’ construction in Section 4.2.3). The comments left by the participants were used to inform further research.

3.4. Analysis parameters

The narratives produced by participants were video recorded. The recordings were transcribed using Elan 6.0 software (Sloetjes & Wittenburg, Reference Sloetjes and Wittenburg2008) by transcribers native in the respective language and annotated using NVivo qualitative data analysis software (QSR International Pty Ltd. Version 12, 2018). The narratives produced by the participants constitute two corpora – the French corpus and the Swedish corpus (see Table 3 for detailed information).Footnote ⁴

Table 3. Information about the corpora ¹

¹ Note that throughout the paper the term corpus refers to a total compilation of narratives in the respective language while the term data set refers exclusively to a set of clauses containing motion events.

The annotation was performed by the author. Each specific expression was cross-referenced with the expressions in the coding scheme used in Zlatev et al. (Reference Zlatev, Blomberg, Devylder, Naidu and van de Weijer2021) in order to ensure homogeneity of the analyses performed, with the exception of prepositions mot and vers ‘toward’, coded as Path rather than as Direction, as mentioned in Section 2.3.

The unit of analysis was a clause, defined as “any unit that contains a unified predicate” (Berman & Slobin, Reference Berman and Slobin1994, p. 660) describing a single situation. As a rule, each unit included a single verbal element, with the exception of infinitives and participles functioning as complements of modal or aspectual verbs (e.g., commencer à voler ‘to start flying’).

The three levels of annotation for linguistic data included:

1. 1. The type of motion situation (with regard to the parameters of boundedness and translocation as described in Section 2.2).

2. 2. A semantic category (based on the ones outlined in Section 2.3.: Path, Manner, Direction, etc.) and the mapping onto one or more form classes. Expressions conflating more than one semantic category were annotated respective to each of the categories and additionally put into a list corresponding to the categories conflated (e.g., Path and Region, Manner and Direction, etc.). The only exception was made for Motion. Since Manner and, in the present data (though not in general, see Blomberg, Reference Blomberg2014), Path and Direction always co-occur with Motion, this would have been redundant. In contrast, verbs that express Motion without conflating it with other semantic categories were coded.

3. 3. Value a specific category acquires, for instance, BEGINNING, MIDDLE, or END with regard to Path (see a full coding scheme in Appendix D).

An example of coding is presented in Fig. 3 Footnote ⁵.

Figure 3. An example of coding.

Systematic coding for covert expression was, as mentioned in Section 2, one of the novel aspects of the analysis. In order to be coded as including a covert expression, a clause had to satisfy two criteria: (1) the covertly expressed meaning clearly is not expressed by any of the clause constituents; (2) the meaning nonetheless is present since its absence would result in inconsistency and contradiction when considered within the context of the narrative. Consider (15), for example:

Neither the verb sauter ‘jump’ nor the preposition dans ‘inside’ express the idea of change of location. The narrow semantic meaning that stems from the semantics of individual lexical units would be the cat jumping while already being inside the trolley. However, this gives inconsistency when checked against the previous utterance (‘The cat approached the woman and her trolley’). Since we assume that the speaker intends to be informative, consistent, and cooperative (following Grice, Reference Grice and Cole1975),Footnote ⁶ we can therefore also assume that the meaning conveyed is referring to an instance of translocation (movement from outside into the trolley) and thus the clause is coded as containing a covert expression of Path.

The number of clauses in general, clauses presenting the events of each of four types (translocative bounded, non-translocative unbounded, translocative unbounded, and non-translocative bounded, see Section 2.2) as well as linguistic expressions instantiating each semantic category (e.g., Path) and the form class (e.g., Path – verb) were calculated.

The statistical tests were performed in JASP software (version 0.16.4) and included (for each set) normality tests and Mann–Whitney U test (since the data was not normally distributed).

The full corpus for each language as well as the lists of expressions for each semantic category and the results of statistical tests can be accessed at OSF | Motion events in Swedish and French.