10.01.2014 Views

In press: In: Dimitrova-Vulchanova, M - NTNU

In press: In: Dimitrova-Vulchanova, M - NTNU

In press: In: Dimitrova-Vulchanova, M - NTNU

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

<strong>In</strong> <strong>press</strong>: <strong>In</strong>: <strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. & E. van der Zee (eds.) Motion encoding in Language, Oxford: OUP.<br />

Motion Naming in Three Satellite-framed Languages: an exploratory study 1<br />

Mila <strong>Dimitrova</strong>-<strong>Vulchanova</strong>, Liliana Martinez, Ole Edsberg & Thomas Brox Røst<br />

The Norwegian University of Science & Technology, Trondheim<br />

Abstract<br />

<strong>In</strong> this paper we report the results of an exploratory study whose main purpose was to test how native<br />

speakers in satellite-framed languages categorize and name scenes of biological motion. Our study<br />

demonstrated that speakers of English, Norwegian and Bulgarian use what we here define as basiclevel<br />

motion verbs, such as walk, run, crawl and climb for the naming of a wide range of biological<br />

motion stimuli rather than more concrete and specific manner verbs or more general verbs, like go or<br />

move. Furthermore, our study provided experimental support for the basic features we propose on<br />

theoretical grounds for the decomposition and semantic representation of motion verbs crosslinguistically.<br />

The current study has also provided important insights and adjustments to the original<br />

typology in terms of satellite-framed vs. verb-framed languages proposed by Talmy (1985, 2000). <strong>In</strong><br />

the paper we address the results and discuss the implications for future studies of biological motion<br />

and its encoding across languages.<br />

0. <strong>In</strong>troduction<br />

It is a well-established fact that languages exhibit a variation in the mapping of conceptual<br />

categories to lexical items and grammatical categories, respectively referred to as<br />

lexicalization and grammaticalization (cf. the seminal work by Talmy 1985, 2000, also Slobin<br />

2001, Clark 2001, among others). Various proposals have been put forward to explain the<br />

nature and consequences of this mapping, all focusing on the issue of universality of<br />

conceptual categories and the likelihood of their lexicalization/grammaticalization crosslinguistically.<br />

Among potentially universal categories, the notions of (biological) motion and<br />

path feature prominently.<br />

This paper reports the results of a pilot study of how motion scenes are encoded in three<br />

satellite-framed languages. The languages under investigation are Bulgarian (a Slavic Balkan<br />

Sprachbund language), English and Norwegian (both of them Germanic languages). The<br />

focus is on open-class forms, more specifically – verb roots, and on lexicalization patterns.<br />

The experimental set-up included a free naming task for basic biological motion scenes<br />

featuring running, walking, crawling and climbing. The purpose of the experiment was to map<br />

out the basic lexical inventory available to speakers when naming a range of biological<br />

motion situations and to check for variation in naming preferences when the motion scenes<br />

differed on selected parameters.<br />

1 Research conducted under Nordic Research Council NOS-H Grant # 10088.<br />

1


1. Background<br />

An important characteristic of language is that it imposes structure on the ‘reality’ it<br />

represents (Jackendoff 1996, Talmy 1985, 2000) by selecting for ex<strong>press</strong>ion only specific<br />

concepts and features. As a consequence, the linguistic ex<strong>press</strong>ion of spatial scenes, including<br />

motion, is constrained by a schema consisting of a closed class of elements like Motion, Path,<br />

Figure and Ground. Motion-event typology focuses on how these elements are encoded crosslinguistically.<br />

One of the most widely applied formats recently is Talmy’s (1985, 2000),<br />

which approaches motion encoding in language from two perspectives, that of linguistic form,<br />

and that of semantic content. <strong>In</strong> terms of linguistic form, the verb is viewed as the main<br />

lexical item encoding the motion event. Languages differ as to what elements of the motion<br />

situation they choose to bundle into their most frequently occurring basic motion verbs.<br />

According to these preferences, languages can be divided into several types, the most<br />

common being path-type languages and manner-type languages. From the perspective of<br />

semantic content and its interface with syntax, languages can be divided into two types,<br />

according to how they ex<strong>press</strong> the path of motion. <strong>In</strong> the so-called verb-framed languages<br />

(e.g. Spanish, Turkish, and Japanese), path is encoded in the verb in the main clause (enter,<br />

exit, ascend, etc.), while manner is specified most commonly in a subordinate clause. <strong>In</strong><br />

satellite-framed languages path is encoded in ‘satellites’ (e.g. adpositions in Germanic<br />

languages and verb prefixes in Slavic languages) which accompany the verb. Manner in such<br />

languages is encoded in the main verb of the clause, and typically those languages have a<br />

large number of commonly occurring manner of motion verbs. Although Talmy’s<br />

classification is useful and captures important insights in motion encoding, the rigidity of the<br />

system has engendered much debate. Many recent studies (e.g. Strömqvist & Verhoeven<br />

2004, Ameka & Essegbey 2000) have demonstrated that the data are more complex. Slobin<br />

(2004) argues that the preferential encoding of path vs. manner in verb roots is not a matter of<br />

belonging to one of two poles, but rather a cline along which manner salience can vary.<br />

Furthermore, there appears to be at least one more distinct type of language, where manner<br />

and path are ex<strong>press</strong>ed by equivalent semantic forms (e.g. verbs in serial-verb languages, cf.<br />

Ameka & Essegbey 2000, Zlatev & Yangklang 2004, among others).<br />

Typological research and Language Acquisition research (cf. Slobin 2001) indicate that<br />

the most frequent verbs of motion cross-linguistically and the ones that tend to be acquired<br />

early on in the process of language acquisition are generalized items denoting default motion<br />

scenes (e.g. scenes which are underspecified for global or local path shape, or parameters,<br />

2


such as figure orientation). Commonly these are the equivalents of come, go, move across<br />

languages. <strong>In</strong> adult language use general motion verbs continue to be the most frequently used<br />

to denote motion scenes generally, and are supplanted by more specific manner verbs only in<br />

cases where the manner of motion itself is at issue (Strömqvist & Verhoeven 2004).<br />

Quite often manner of motion verbs have been treated as pure manner verbs, and not as<br />

motion verbs per se. This line of reasoning has followed the assumption that such verbs<br />

denote an activity which usually accompanies motion, but do not encode motion as such. This<br />

claim, however, cannot be universally substantiated by the data. Thus, cross-linguistically we<br />

find “manner” verbs (e.g. Bulgarian katerja se, Norwegian klatre – ‘climb/clamber’, German<br />

kletteren, Bulgarian turkaljam se, Italian rotollare – ‘roll’) which uncontroversially include<br />

both a Manner and a Path element (cf. also Weisgerber, this volume for a similar proposal). <strong>In</strong><br />

addition, recent research in cognitive interfaces and the cross-linguistic encoding of motion<br />

(cf. Nikanne & van der Zee 2005, this volume, <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2004a, 2004b) has<br />

demonstrated that both Path and Manner are too crude to account for the linguistic evidence,<br />

and, as such, lead to controversial representations and accounts. The notion of Path, for<br />

instance, subsumes parameters as diverse as spatial grids and frames of reference, Regions<br />

and Grounds (van der Zee and Nikanne 2000, van der Zee and Slack 2003). Also, the<br />

encoding of path in languages has been shown to vary in the way a motion event is<br />

segmented, and different path elements may be related to Grounds/ Landmarks and the<br />

regions defined by them (Slobin 2004). Path shape has also been shown to be an important<br />

factor which can be encoded cross-linguistically with different degrees of granularity<br />

(Nikanne 2004, Nikanne & van der Zee 2005). <strong>In</strong> the proposal in van der Zee (2000), and<br />

Nikanne & van der Zee (this volume), a verb like zig-zag encodes a local path shape<br />

specification, as each path segment of the zig-zag scene involves a change in figure<br />

orientation and thereby path orientation at a certain (sharp) angle. Observe that quite often<br />

these verbs have been labelled ‘manner’ verbs, while specifying exclusively features of the<br />

path.<br />

<strong>In</strong> a similar fashion, the notion of Manner can be felicitously decomposed into a number<br />

of independent parameters pertaining to various aspects of the motion scene, e.g. (change in)<br />

figure orientation or curvature (wriggle, shake), rate of displacement (speed) (e.g. the contrast<br />

between run and sprint), psychological state of the figure (e.g. English strut, the Luganda<br />

verb gubagguba – ‘trudge for a long distance with a sad event ahead’, Ndiwalana 2003).<br />

Thus, neither Path nor Manner can be viewed as primitives in the theory. At best, they are<br />

useful ‘short-hands’ for a range of independent and well-defined features.<br />

3


Particularly interesting with respect to the encoding of these features are what we here<br />

label ‘verbs of biological motion’. The term ‘biological motion’ denotes the patterns of body<br />

and limb motion which animate creatures employ to achieve translational motion. Biological<br />

motion differs from non-biological motion in several ways (however, see Bejan & Marden<br />

2006 for a unified perspective on all motion). Its recognition and ex<strong>press</strong>ion in language is of<br />

paramount cognitive and social importance to human beings (Mandler 1996, Troje 2002). It is<br />

by definition self-agentive and involves the cyclic iteration of events of greater complexity<br />

(the cycles/patterns of internal motion of the body and limbs), the function of which is to<br />

cause translational motion (<strong>Dimitrova</strong>-<strong>Vulchanova</strong> 1996/99).<br />

A viable hypothesis of the factors that play a role in biological motion categorization<br />

can draw on evidence from research in visual perception. As it turns out, the human<br />

mechanisms of biological motion recognition are extremely robust. Biological motion can be<br />

recognized from strongly impoverished stimuli, for example when the moving figure is<br />

reduced to a point-light display (classical experiment in Johansson 1973, Giese 2004a, b).<br />

Furthermore, motion categorization is learning-based, perspective-dependent and selective<br />

(ibid). Giese & Poggio (2003) argue that the robustness of motion categorization resides in<br />

two neural pathways, each of them representing motion in a specific way: A form-pathway<br />

recognizes biological motion as a sequence of ‘snapshots’ of the figure in motion, and a<br />

motion-pathway recognizes biological motion as a sequence of optic flow patterns. While<br />

human action perception seems to tolerate substantial variation in form features (Giese 2004<br />

a, b), motion patterns seem to be specific to particular types of actions, which explains why<br />

biological motion can be recognized only through the motion pathway and in the absence of<br />

form information (e.g. in point-light displays). This theory of motion recognition enables us to<br />

hypothesize which criteria will be relevant in the categorization and linguistic encoding of<br />

biological motion. Criteria related to the form-pathway of recognition are body shape and<br />

proportions (e.g. bulky vs. slim body, short vs. long legs), characteristic use of limbs (e.g.<br />

biped vs. quadruped; the isolated movements of the limbs) and, by extension, also species e.g.<br />

human vs. non-human). The series of ‘snapshots’ in a particular temporal order is what we<br />

will call the cycle of a particular type of biological motion. Relevant factors, related to the<br />

motion pathway of recognition will be characteristic speed, path (the presence vs. absence of<br />

translational motion), basic path shape (underspecified vs. globally specified, cf. Nikanne &<br />

van der Zee 2005, <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2004a, b). The view dependence of motion<br />

recognition will predict that factors like figure orientation (e.g. front forwards vs. front<br />

backwards, head up vs. head down), relative vector orientation (towards vs. away from vs.<br />

4


left-to-right vs. right-to-left) will play a part also in categorization, and possibly in<br />

lexicalization (cf. Jellema et al. 2002 for a bias for left-to-right human walking recognition in<br />

the macaque, also <strong>Dimitrova</strong>-<strong>Vulchanova</strong> & van der Zee, in preparation).<br />

2. Objectives<br />

From the brief discussion above, it has become clear that Manner and Path are pre-theoretical<br />

terms, useful in the description of major lexicalization strategies (e.g. the path / manner<br />

contrast). Detailed typological and conceptual structure research, however, ought to resort to a<br />

more refined system taking into account the different parameters that define these notions.<br />

With this in mind, we conducted an exploratory study which had three main goals. The first<br />

goal was to zoom in into the features comprising “manner” and to empirically test the validity<br />

of the parameters proposed in van der Zee (2000), Nikanne (2004), Nikanne & van der Zee<br />

(2005) and <strong>Dimitrova</strong>-<strong>Vulchanova</strong> (2004a, b), and to check whether there is variation in<br />

naming preferences when the scenes vary on different parameters. Our second objective was<br />

to find whether there is evidence of the presence of a translational component (the feature<br />

[+Path]) in the conceptual structure of common manner verbs, such as run, walk, crawl and<br />

climb (cf. Berman & Slobin 1994, Jackendoff 2002b, <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2003, 2004a, b).<br />

The third objective was to check whether in the description of motion scenes native speakers<br />

of the target languages would resort to motion verbs we define as basic-level in these<br />

languages (e.g., walk, run, crawl, climb) or to the super-ordinate general motion verbs (e.g.,<br />

go, come, move), or would perhaps use more concrete and detailed manner verbs (e.g., gallop,<br />

scurry, jog, lope and the like).<br />

3. The Verbs<br />

All three languages in the current study (English, Norwegian and Bulgarian) fall in Talmy’s<br />

satellite-framed category. Typically, they conflate translational motion with manner and<br />

ex<strong>press</strong> the path in such contexts through satellites or prepositional phrases, as shown in the<br />

examples in (1a-c). However, all three languages also have a dedicated category of verbs<br />

encoding pure translational (non-conflated) motion. Within the latter category, two major<br />

types can be distinguished. On the one hand, we have verbs that denote bounded motion (e.g.<br />

the boundary-crossing constraint of verb-framed languages), as in (1d) parallel to the Spanish<br />

example in (1e), and, on the other, verbs that do not encode any path specification (i.e. they<br />

lexicalize ‘pure translational motion’), as in English move, Norwegian bevege seg and<br />

Bulgarian dviza se (both meaning ‘move’) in the examples in (1f-h).<br />

5


(1) a. The bird flew out of the window<br />

(main verb specifying Manner) (path satellite)<br />

(English)<br />

b. Fuglen fløy ut<br />

bird-the flew<br />

out<br />

(main verb specifying Manner) (path satellite)<br />

(Norwegian)<br />

c. Pticata izletia prez prozoreca<br />

bird-the pref.-flew out window-the<br />

(main verb letja specifying Manner, plus a directional prefix iz-, specifying Path)<br />

(Bulgarian)<br />

d. Momcheto izleze ot kustata lazeiki<br />

boy-the pref.-went out house-the crawling<br />

(main verb ex<strong>press</strong>ing path - ‘exited’) (manner participle – ‘crawling’)<br />

“The boy exited the house crawling”<br />

(Bulgarian)<br />

e. El pajaro salió volando.<br />

(main verb specifying Path) (subordinated clause specifying manner)<br />

“The bird exited flying”<br />

(Spanish)<br />

f. The dot moves. (English)<br />

g. Prikken beveger seg. (Norwegian)<br />

h. Točkata se dviži. (Bulgarian)<br />

Since in all three languages, both verbs conflating manner and translational motion and<br />

verbs encoding pure translational motion can be felicitously employed given the appropriate<br />

context, a secondary goal in our pilot study was to check to what extent speakers of these<br />

languages would resort to the conflation verbs and to what extent the more general pure<br />

translational motion verbs would be used in the naming of motion scenes. <strong>In</strong> other words, we<br />

set out to check to what extent the three languages would behave as satellite-framed following<br />

the standard classification, and to what extent they will employ strategies claimed to belong in<br />

the verb-framed languages.<br />

6


With respect to terrestrial biological motion, at least four basic types of biological<br />

motion can be distinguished: walking, running, crawling, and climbing. The WALK type<br />

covers supported gaits usually involving ‘normal’ to slow speed. The RUN type applies to<br />

quick suspended motion. The CRAWL type defines slow supported motion, with the body<br />

near and/ or parallel to the ground. The CLIMB type covers motion in both directions along a<br />

vertical axis. Languages tend to have at least one lexical item encoding each basic type, but<br />

this is not necessarily so. For instance, the Asante dialect of Akan (a West African serial verb<br />

language) does not have verbs corresponding to e.g. run or fly (cf. Apraku 2005, <strong>Dimitrova</strong>-<br />

<strong>Vulchanova</strong> & Martinez, in preparation), and Ewe has only one lexical item, dzò for hop,<br />

jump and fly (cf. Agbelengor 2007). Also, some languages may display more than one verb<br />

for the same basic motion type (e.g. Bulgarian. tičam - ‘run1’ vs. bjagam –‘run2’) with no<br />

clear-cut distinction at first sight. 2 We address the Bulgarian data in 3.1 below.<br />

<strong>In</strong> the lexicons of English, Bulgarian, and Norwegian there are verbs of biological<br />

motion corresponding to all four basic types (cf. e.g., Rosch et al. 1976 for objects, and Zacks<br />

& Tversky 2001 for events, among others). <strong>In</strong> addition, all three languages have more<br />

concrete verbs (e.g. English strut, pace, Norwegian spasere –‘take a walk’, Bulgarian<br />

razxozdam se – ‘take a walk’). This is not surprising for two reasons. Firstly, these are the<br />

main modes of terrestrial locomotion typical of humans commonly observed in the world<br />

around us. Secondly, all three languages belong to the same language family. Given this fact<br />

we set out to investigate how speakers of these languages would use motion verbs in their<br />

descriptions of motion scenes.<br />

Here we will concentrate only on the following verbs in the domain of biological<br />

terrestrial motion. 3<br />

• For the walking type: English walk, Norwegian gå and Bulgarian hodja/ vurvja<br />

• For the running type: English run, Norwegian løpe and Bulgarian tičam / bjagam<br />

• For the crawling type: English crawl, Norwegian krabbe/ krype and Bulgarian<br />

pulzja/ lazja<br />

2 <strong>In</strong> Bulgarian the use of two verbs is generally motivated diachronically, as Old Bulgarian would have the socalled<br />

aspectual verb pairs, with one verb denoting the general scene (e.g. the specific motion pattern), and the<br />

other denoting the goal-oriented, bounded (telic) event. Such pairs are still found in Modern Russian (cf. Foote<br />

1967), e.g. xodit` (roughly corresponding to walk) vs. idti (go), and to a limited extent in Modern Bulgarian, e.g.<br />

xodja (walk, imperf.) vs. otivam (go, perf.). However, it would be premature to attribute this situation to a<br />

‘specific’ Slavic property, as German displays a similar situation (cf. e.g., Weisgerber, this volume for a<br />

discussion).<br />

3 For English, the main source was Cambridge Advanced Learner’s Dictionary, but Merriam-Webster Online,<br />

Oxford Dictionary online, Encarta Dictionary for English were also consulted for a counter-check. For<br />

Bulgarian, the dictionary used was Andreichin (2004), and for Norwegian, the online versions of ‘The Bokmål<br />

Dictionary’ and ‘The Nynorsk Dictionary’.<br />

7


• For the climbing type: English climb, Norwegian klatre and Bulgarian katerja se,<br />

slizam<br />

We predict that these verbs are basic-level verbs in the sense of Rosch (1973), thus being used<br />

more often than more abstract or more specific verbs in the description of events in that<br />

domain. <strong>In</strong> this section we briefly introduce the linguistic facts for each language, discussing<br />

data relevant for our basic hypotheses in the exploratory study.<br />

3.1 The Bulgarian verb pairs<br />

The Modern Bulgarian verbs for walking, running, and crawling deserve somewhat special<br />

attention. As is widely assumed, language does not tolerate redundancy in the lexicon. True<br />

synonyms are rarely attested cross-linguistically (Lyons 1977, among others). Now, a quick<br />

look up in standard mono-lingual Bulgarian dictionaries reveals very similar, if not<br />

overlapping, definitions. Quite often, one of the verbs is used to explain the meaning of the<br />

other. There is indeed some complementarity in the use of the respective verbs, xodja / vurvja<br />

(walk1/walk2), tičam / bjagam (run1/run2) and pulzja / lazja (crawl1/crawl2), and native<br />

speakers experience problems in explaining the potential difference between the two verbs in<br />

the pair 4 . Furthermore, it has not been attested in research whether they agree on how to use<br />

those verbs, either. One way to approach this issue is to look for differences on the level of<br />

language as a system revealed in patterns of occurrence and collocational restrictions, and in<br />

restrictions on the mapping of conceptual and semantic information to syntax. Another<br />

approach, empirical in nature, can simply test how native speakers use the items in question.<br />

We briefly sketch the differences observed at the level of grammar.<br />

One distinction are constraints as regards use in certain idiomatic ex<strong>press</strong>ions. Thus, of<br />

the two walk verbs, only xodja (walk1) occurs in fixed collocations, such as xodja na-PP<br />

(attend), as shown in (2a) below. <strong>In</strong> contrast, only vurvja (walk2) can be used in the context of<br />

Imperatives, most likely because Imperatives are usually associated with a bounded event, i.e.<br />

a directed motion situation. An example is given in (2a’) below featuring the two directed<br />

motion verbs vurvja (walk2)/otivam (go) vs. the ungrammatical xodja (walk1).<br />

Likewise, in aspectual processes involving reference to walking/running situations, only<br />

one of the two verbs (verb2) can give rise to an aspectually completed bounded event. This<br />

process is characterized by promotion of the prepositional phrase (PP) that ex<strong>press</strong>es the path<br />

to direct object accompanied by aspectual prefixation on the verb. <strong>In</strong> such cases, more often<br />

4 As revealed in random elicitation queries in the preparation of the experiment.<br />

8


than not, the aspectual prefix is a copy of the original preposition that heads the PP (cf.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong> 1998, 1999, and Guentcheva 2002, 2006 for an analysis and<br />

discussion). Consider the examples in (2b-b’, c-c’) below.<br />

(2) a. xodja na kino/uciliste/rabota<br />

Go to (the) cinema/school/work<br />

a’. Vurvi/otivaj/*xodi na rabota<br />

walk2/go/*walk1 to work<br />

b. vurvja/xodja [ PP po putja]<br />

walk on the road<br />

b’. Ivan iz-vurvja/*-xodi [ NP celija put]<br />

I. asp.pref.-walked the whole road<br />

c. bjagam/ticam do kustata<br />

run to the house<br />

c’. Ivan pro-bjaga/*- tiča razstojanieto<br />

I. perf.pref.-ran the distance<br />

<strong>In</strong> both pairs in (2b-b’) and (2c-c’) the contrast is between the possibility of promoting the<br />

prepositional phrase, po putja (on the road) and do kustata (to the house), respectively in the<br />

case of vurvja (walk2) and bjagam (run2), but not xodja (walk1) and tičam (run1). The PPpromotion<br />

construction in Bulgarian is similar, but not identical to the English walk NP (as in<br />

walk the corridors, sail the seas) construction, whereby the interpretation in the case of<br />

English is that of [+affected] for the direct object. The path PP-promotion pattern is an<br />

important test providing evidence of the underlying semantics of the verbs at hand in the<br />

sense of whether they encode directed traversal of the ground, or not.<br />

Furthermore, primarily xodja (walk1), but not vurvja (walk2) can be used in the derivation<br />

of new lexical items combining with semi-lexical/semi-aspectual prefixes to produce new<br />

verbs of motion further specifying various aspects of ‘manner’. Thus pro-xodja means to start<br />

walking (toddle), raz-xodja (stroll, Norwegian spasere), za-xodja ((the sun) set down). Based<br />

in our current research (cf. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> & Martinez, forthcoming) we propose that<br />

the two Bulgarian verbs in each pair, while both encoding traversal, differ in the presence of<br />

the feature [+directed motion]. Thus, xodja (walk1) encodes /low pace/ biological traversal<br />

motion, while vurvja (walk2) lexicalizes directed traversal motion. This find is coherent with<br />

an assumption that self-propelling based in force accumulation and discharge naturally leads<br />

9


to transition in space (cf. the analysis in <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 1996/99, 1999). <strong>In</strong> the study<br />

we set out to check whether the picture in this outline is supported by native speaker<br />

responses.<br />

3.2 Verbs of running<br />

The main verbs denoting running are English (run), Bulgarian (tičam, bjagam), and<br />

Norwegian (løpe, springe). They encode roughly the same meaning, which has been<br />

described by dictionary definitions in all three languages as ‘move fast by propelling oneself<br />

with quick and abrupt motions of the legs from the ground’. The two run-verbs in Bulgarian<br />

have more or less the same status with respect to register, however they display the [+/-<br />

directed motion] distinction discussed above. <strong>In</strong> contrast, in Norwegian, there is a distinction<br />

of register. The verb løpe belongs to the formal Bokmål variety, while springe is more<br />

informal, and is typical of particular dialects where it is employed consistently instead of løpe.<br />

Some examples are given in (5) below.<br />

(5) a. We had to run to catch up with him (English, Cambridge Learner’s Dictionary)<br />

b. Byaga za zdrave (Bulgarian)<br />

“He runs/ is running for health/ recreation”<br />

c. Detsata tičat… po polyanata. (Bulgarian, Andreichin 2004)<br />

“The children are running on the meadow”<br />

d. Jeg løp og løp (Norwegian)<br />

”I ran and ran”<br />

e. Han selv ikke ønsket å springe omkring i skogen … (Norwegian)<br />

”He himself did not want to run around in the forest”<br />

<strong>In</strong> addition to the meaning of biological motion, the verbs run, løpe, springe and bjagam<br />

(run2), but not tičam (run1) have the meaning of ‘flee, move away (by running) when in<br />

danger or being chased’.<br />

The English verb run, and the Norwegian verb løpe also have extended meanings along<br />

the lines of ‘move quickly’, ‘function’, ‘develop in time’, which are by far more widely used<br />

than the biological motion meanings, as revealed in a search in the ICAME corpora collection<br />

which yielded only hits for non-metaphorical run.<br />

10


3.3 Verbs of walking<br />

The most central verbs encoding walking are English walk, Norwegian gå, and Bulgarian<br />

xodja (walk1), vurvja (walk2), as illustrated in (6) below. All of them have the meaning of<br />

‘move forward by putting one foot in front of the other usually at a moderate pace’ 5 .<br />

(6) a. Peter walks along the road. (English)<br />

b. Petur vurvi/ xodi po putja. (Bulgarian)<br />

c. Petter går langs veien. (Norwegian)<br />

Observe that the dictionary definition of the walk verbs captures the [+traversal] feature<br />

we proposed for verbs of biological motion in the discussion of Bulgarian. The Norwegian<br />

verb gå is ambiguous between the manner meaning (=‘walk’) and a meaning of a general<br />

motion along a path, with which it is predominantly used. Gå is also used with a variety of<br />

abstract meanings of the type ‘happen’, ‘develop in time’.<br />

3.4 Verbs of crawling<br />

For the description of crawling scenes, English uses mainly the verb crawl, Bulgarian uses the<br />

verbs pulzja (crawl1) and lazja (crawl2), and Norwegian, the verbs krabbe and krype.<br />

Illustrations are given in (7) below.<br />

(7) a. There's an ant crawling up your leg. (English)<br />

b. For første gang så jeg henne krabbe fremover. (Norwegian, The Oslo Corpus)<br />

”It was the first time I saw her crawling forwards”<br />

c. En bille krøp over gulvet. (Norwegian)<br />

“A beetle was crawling across the floor”<br />

d. Zmiyata/ bebeto/ gusenitsata pulzi. (Bulgarian)<br />

“The snake/ baby/ caterpillar is crawling”<br />

e. Mravkata/ deteto lazi.<br />

“The ant/ baby is crawling”<br />

The verbs crawl (E), pulzja (Bg) and krype (N) mean: ‘move slowly in a prone position,<br />

with the body along close to or touching the surface’ 6 . The use of limbs is not required. Lazja<br />

5 As defined in Longman and Encarta Dictionary<br />

6 Cambridge Advanced Learner’s Dictionary<br />

11


(Bg) and krabbe (N), however, apply only to scenes where limbs are used, in addition to the<br />

specification for a ‘prone’ body position.<br />

3.5 Verbs of climbing<br />

For the description of biological motion along the vertical axis, and adhering to a vertical<br />

objector surface, English uses the verb climb, and Norwegian the verb klatre. Both verbs can<br />

be used to denote motion either up (8 a, b) or down, but in the latter case the direction is<br />

usually specified overtly by a directional ex<strong>press</strong>ion (8 d, e). Bulgarian splits the domain of<br />

vertical biological motion into two, using the verb katerja se for upward motion (8c), and<br />

slizam for downward motion (8f).<br />

(8) a. The koala is climbing the tree (English)<br />

b. Koalaen klatrer opp treet (Norwegian)<br />

koala-the climbs up tree-the<br />

c. Koalata se kateri (*nadolu) po durvoto (Bulgarian)<br />

koala-the refl. climb (downwards) on tree-the<br />

d. The sloth is climbing down the tree. (English)<br />

e. Dovendyret klatrer ned/ nedover treet. (Norwegian)<br />

Sloth-the climbs down1/ down2 tree-the<br />

f. Lenivecat sliza ot durvoto. (Bulgarian)<br />

sloth-the descends from tree-the<br />

An interesting difference between English and Norwegian is that while English climb<br />

encodes the upward direction by default, in Norwegian this has to be overtly specified, as<br />

shown in (8a-b). <strong>In</strong> a similar way, in Bulgarian the upward direction is lexically encoded in<br />

katerja se (climb) and need not be overtly ex<strong>press</strong>ed at all, as confirmed by the example in<br />

(8c) above. <strong>In</strong> the latter, the PP realizes a location, not a direction, as revealed by its English<br />

gloss, ‘on the tree’. Further confirmation of the inherent specification for upward direction<br />

comes from the ungrammaticality of an adverbial for downward direction (nadolu,<br />

downwards) in the context of katerja se (climb). While klatre, katerja se and slizam are<br />

usually restricted to biological motion and require a characteristic use of limbs and clinging to<br />

a vertical or steep surface, climb can be used with the more general meaning of ‘go up, move<br />

upwards’, very similar to German steigen (cf. Weisgerber, this volume). This use of climb is<br />

not considered here, as none of the scenes in the experiment show non-biological notion.<br />

12


3.6 Features in the representation of motion verbs<br />

Here we propose a finer-grained feature analysis for the representation of biological motion<br />

based in parameters independently argued to apply in the identification and categorization of<br />

motion scenes (cf. Giese & Poggio 2003, Shipley 2003, van der Zee 2000, Nikanne & van der<br />

Zee 2004, this volume, <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2004a,b, Matlock & Richardson 2004,<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong> & van der Zee, in preparation, among others).<br />

• Type of biological motion (e.g., terrestrial vs. aqua vs. air)<br />

• Species (human vs. non-human)<br />

• Characteristic use of limbs (bi-ped vs. quadro-ped vs. centi-pede)<br />

• Characteristic speed (fast vs. normal vs. slow)<br />

• Figure orientation (front-forwards vs. front-backwards vs. head-up vs. head-down)<br />

• Figure posture (erect vs. supine/lower than usual)<br />

• Path (presence vs. absence of translational motion)<br />

o Vector orientation (horizontal: towards vs. away from vs. left-to-right vs.<br />

right-to-left; vertical: up vs. down)<br />

o Path shape (underspecified vs. global specified vs. local specified)<br />

• Psychological state of the figure<br />

The above features can be represented in terms of a typed structure with a variation in<br />

the values set for each type. Thus, an instance of [terrestrial/ human/ bi-pedal/ supine/ slow/<br />

translational/ underspecified-path-shape] motion would be crawl. Most of the above types<br />

have default values and quite often are hierarchically related in terms of<br />

entailment/inheritance. For instance, human terrestrial motion is by default bi-pedal, head-up.<br />

Likewise, some features will not be relevant for the categorization, and respectively,<br />

representation of certain motion types. Thus, for the aqua type, the use of limbs does not play<br />

any role whatsoever, humans swim with arms and legs, fish swim with fins, etc. Rather, the<br />

type of contact with the surface and whether motion is self-propelled are important, e.g., swim<br />

vs. float (cf. Geuder & Weisgerber 2006, Lander et al., this volume). Thus, the main thrust of<br />

our analysis is that all of the above types can in theory combine on all of their values, thus<br />

potentially yielding a host of possible motion scenes, as confirmed by what is observed in the<br />

external world, and in the naming patterns reported in our experiment below. An interesting<br />

research question in this respect is whether all theoretically possible combinations are<br />

13


encoded linguistically across languages. Our current work suggests that languages rather<br />

choose to encode certain (proto-)typical combinations of features that correspond to higher<br />

perceptual salience and experiential frequency of the motion scene at hand. We refer to this as<br />

the default setting of parameters. Thus, cross-linguistically we expect these feature<br />

combinations to cluster systematically, giving rise to lexical items in the respective (motion)<br />

lexicon. For the naming of less prototypical or non-default situations, speakers are expected to<br />

resort to items already existing in the lexicon, a strategy we refer to as re-cycling (cf. Dekova<br />

& <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2006). Thus, a chimp moving on all fours (i.e., knuckle-walking),<br />

which is the default for the species, will be nevertheless characterised by walk, if slow or run,<br />

if fast, apparently picking on the head-up horizontal nature of the motion or, alternatively<br />

ignoring hand-contact. The above features can be represented along the attribute-value graph<br />

format proposed in <strong>Dimitrova</strong>-<strong>Vulchanova</strong> (2004a,b) in (9) below. 7<br />

Fig. 1<br />

Monodevelopment<br />

Element: Moverj<br />

Orientation of Moverj:<br />

+/- rotation<br />

front-forwards/front-backwards<br />

Psychological state of Moverj<br />

Phasing: multi-point<br />

Medium: Location<br />

Path<br />

Fixed parameters:<br />

Vector orientation:<br />

Horizontal (left-to-right/right-to-left/towards/away from)<br />

Vertical (upward/downward)<br />

Shape: (straight/curved/circular)<br />

Specifiable parameters:<br />

Path origin<br />

Path end<br />

Contiguous to: x2<br />

Path length<br />

Path-temporal function: path length<br />

tn<br />

The basic types in the representation include the figure in motion under the label of Mover,<br />

the medium with respect to which the change takes place, in the case of motion this is a<br />

7 All unidirectional processes of change, including motion are subsumed by the category Monodevelopment<br />

following the Sign Model framework (cf. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 1996/99, <strong>Dimitrova</strong>-<strong>Vulchanova</strong> & Hellan<br />

1995, <strong>Dimitrova</strong>-<strong>Vulchanova</strong> & Hellan 2000, Dekova 2006)<br />

14


change of location, and whether the situation at hand is defined by multi-points (is a process)<br />

or by one/two-points (is punctual). The location includes specifications and attributes of the<br />

path defined as fixed parameters (those that are necessarily encoded in the meaning of the<br />

verb), and specifiable parameters (those that can be overtly referred to, and can take any value<br />

without causing contradiction with the original meaning of the verb). A separate path attribute<br />

is the path-temporal function which refers to the velocity of movement. There are also<br />

attributes of the Mover, the most important of which is its orientation in the movement. Also<br />

included in the representation is the possibility of certain verbs to specify for a psychological<br />

state of the mover (e.g., in Luganda (cf. Ndiwalana 2005), quite often verbs encode this<br />

feature).<br />

The current format is based in findings in current research of the perception and<br />

identification of biological motion (Johansson 1973, Giese 2004a, b, Jellema et al. 2002). It<br />

allows for highly detailed, while at the same time, naturally constrained representations (i.e.<br />

the array of features is limited) without resorting to descriptive loose notions, such as e.g.,<br />

manner. Such a format is also more coherent with the physical properties of translational<br />

motion in naturally including features of the figure in motion or the path or the path-temporal<br />

function (speed), and recognizing their salience in the recognition and categorization of<br />

motion scenes, rather than “degrading” them to concomitant, but not motion proper<br />

parameters.<br />

3.7 Hypotheses<br />

On the basis of the general motion encoding properties of the target verbs in the three target<br />

languages, our hypothesis is that the speakers of these three languages will use the verbs of<br />

biological motion, rather than pure path verbs, or more abstract verbs when describing scenes<br />

of biological motion. We expect this choice because the basic-level categories are<br />

informationally richer than the superordinate (0-level on Nikanne & van der Zee’s 2004<br />

classification) ones, as they encode translational motion ([+ Path]), in addition to other<br />

specifications of the path or figure in motion (e.g., path shape, figure orientation, specific use<br />

of limbs etc.). Furthermore, the target verbs, which are basic-level verbs describing terrestrial<br />

gaits, will be preferred to their more specific/concrete verbs (e.g., strut, gallop, lope), because<br />

of the level of encoding they represent, revealed, among other things in looking at the<br />

linguistic data in terms of patterns of occurrence.<br />

With respect to the categorization and lexicalization of biological motion, our prediction<br />

is that non-default scenes of biological motion will elicit a less uniform response across verb<br />

15


types, since such scenes are more difficult to categorize, and thus, more difficult to name than<br />

scenes with default setting of the parameters (cf. e.g. the study by Pavlova et al 2001,<br />

Thornton & Vuong 2004, among others). A default motion scene, such as e.g., walking,<br />

would feature a figure moving head-up horizontally left-to-right forward (cf. Giese & Poggio<br />

2003, Shipley 2003, among others). We return to default settings in the discussion in section<br />

5. below.<br />

<strong>In</strong> regard of the distribution of the target verbs across answers, we predict that the<br />

responses for Bulgarian will display the two-way split found in the lexicon, e.g. answers will<br />

feature the two verbs in the pair, verb1 and verb2. The distribution of these verbs will vary<br />

according to whether the context does or does not activate the feature [+directed motion] that<br />

the verbs in the pair do not share. Norwegian gå is expected to be used more frequently across<br />

all scenes compared to the walking verbs in English and Bulgarian, due to its general motion<br />

sense equivalent to English go (i.e. also for scenes which do not depict walking).<br />

4. Experimental design<br />

4.1 Material, method and participants<br />

The purpose of this experiment was to compare how some instances of biological motion are<br />

described by native speakers of Bulgarian, English, and Norwegian. For each of the three<br />

languages, the participants performed a free naming task: they watched a sequence of scenes<br />

on a computer screen and were asked to name the action they saw in each clip and to rate their<br />

level of confidence on a scale from 1 (lowest) to 7 (highest) for each response. 8<br />

46 stimuli were selected from documentaries or made by the experimenters with the aim<br />

to provide a range of biological motion scenes observable in nature or everyday life featuring<br />

intransitive actions performed by animate beings (humans, non-human primates, other<br />

animals). 17 out of the 46 scenes were distractors. The target clips, showed five full cycles of<br />

the action, or, for slower actions, a time interval of approximately 5 seconds. The scenes were<br />

shown in a stratified random order, to ensure that scenes that were similar were not presented<br />

close to each other. There were both typical and less typical to marginal examples of the<br />

target biological motion. The scenes covered variations with respect to cycle structure and/or<br />

body configuration (crawl on all fours, crawl on one’s stomach, bipedal walk, quadrupedal<br />

walk, bipedal run, quadrupedal run, quadrupedal trot), species (human, ape, bird, cat, dog,<br />

etc.), age differences among the actors (baby vs. adult), speed (slow vs. normal vs. fast),<br />

8 However, to the extent that some subjects had used an inverted scale (especially in the responses for<br />

Bulgarian), we did not process the level of confidence results statistically.<br />

16


translation vs. non-translation (regular running vs. running on the spot or running on a<br />

treadmill), vector orientation (horizontal: forwards vs. backwards vs. to the left/ right,<br />

vertical: up vs. down), figure orientation (front forwards, front backwards), shape of path<br />

(straight, circular), substrate (ground vs. branch vs. leaf, smooth vs. grassy surface). The<br />

stimuli were pre-screened and judged for prototypicality by 5 adult speakers for each<br />

language. Based in the judgements from the pre-test screening, we selected 29 target scenes<br />

(cf. Appendix 1), divided into 9 walking scenes, 9 running scenes, 8 crawling scenes, and 3<br />

climbing up/ down scenes.<br />

The participants were adult native speakers of the respective language: 16 for Bulgarian,<br />

16 for Norwegian, and 12 for English. They were randomly selected for age and sex. Each of<br />

the participants completed the experiment separately, in the presence of staff conducting the<br />

experiment.<br />

4.2 The procedure<br />

Subjects viewed the clips in a single session, with a brief break in between each two blocks of<br />

trials. They received detailed instructions from staff conducting the experiment before the<br />

start of each session, and also on screen before each block of trials. They were advised to<br />

provide the first word/description that came to mind and allowed to work at their own pace.<br />

Each image froze after the showing time expired and could not be played back for reference.<br />

Subjects were then prompted to enter their responses in a text box that appeared under the<br />

image. Below the text box there was a digit box for selecting the value for level of<br />

confidence. After filling in this box subjects could proceed to the next clip.<br />

4.3 Design<br />

The scenes presented in the experiment were not matched on parameters like settings,<br />

physiological characteristics of the agents, viewing angle, etc. We wanted to elicit preliminary<br />

responses to as many different instances of the actions as possible, and include not only<br />

humans but other species as well. For this purpose, we had to use natural settings and<br />

deliberately ignore the context uniformity requirement. As the experiment was planned as an<br />

exploratory study for a sequence of further experiments on the encoding of biological motion,<br />

its purpose was not to check systematically for all of the above factors, but to give general<br />

indications of their role and potential significance in motion categorization and naming, and<br />

thus help to direct the attention to specific features for further research. We are aware that the<br />

results should be interpreted accordingly. Thus, the current study has provided a good starting<br />

17


point for further research in the area, generating concrete direct experimental developments<br />

(cf. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> & van der Zee, in preparation), as well as work on setting up a<br />

cross-linguistic typology of ex<strong>press</strong>ions of biological motion (cf. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> &<br />

Dąbrowska, in progress).<br />

4.4 Results<br />

As predicted, in all three languages, the participants preferred to use the basic-level motion<br />

verbs to describe the target scenes, even though the lexicons of all three languages display a<br />

rich inventory of motion words. Thus, clearly our study points towards a distinction between,<br />

on the one hand, properties of the lexicon (in terms of whether it encodes a concept or not),<br />

and naming strategies, on the other, similar to Slobin’s (1996) habitual attention/language use<br />

tendencies. <strong>In</strong> the responses for all languages, the verbs that are the focus of study were more<br />

frequent than other more abstract or more specific (gait) verbs, thus showing that these verbs<br />

are basic-level verbs (in the sense of Rosch 1973). Most of the respective motion scenes<br />

elicited clear responses that fall into groups, according to verb use and in correspondence to<br />

the scene type viewed (e.g. run verbs for running, walk verbs for walking, etc.). A small<br />

number of scenes failed to classify as either (cf. discussion below).<br />

As predicted, English and Norwegian opt for one basic term per biological motion type,<br />

except for the split in Norwegian in the domain of crawling (see fig.2). General (non basiclevel)<br />

motion verbs were used very rarely in all three languages, and only in connection with<br />

scenes with non-default parameter settings.<br />

18


Fig. 2 Overall frequencies of the target verbs in English<br />

22,13<br />

20,98<br />

21,55<br />

17,53<br />

8,33<br />

5,46<br />

0,57 0,86<br />

2,59<br />

run walk crawl climb general motion more specific<br />

biological<br />

motion<br />

intrinsic motion tandem non-motion<br />

Fig. 3 Overall frequencies of the target verbs in Norwegian<br />

24,78<br />

23,92<br />

14,22<br />

9,27<br />

9,91<br />

8,19<br />

5,60<br />

3,66<br />

0,43<br />

gå løpe/springe krabbe krype klatre general motion more specific<br />

biological<br />

motion<br />

intrinsic motion<br />

non-motion<br />

19


Fig. 4 Overall frequencies of the target verbs in Bulgarian<br />

15,95<br />

14,87<br />

11,64<br />

9,48<br />

9,91<br />

8,84<br />

10,34<br />

4,53<br />

5,60<br />

5,17<br />

1,51<br />

0,86<br />

1,29<br />

vurvja hodja ticham bjagam pulzja lazja katerja se slizam general<br />

motion<br />

more<br />

specific<br />

biological<br />

motion<br />

intrinsic<br />

motion<br />

tandem<br />

motion<br />

nonmotion<br />

The histograms in Fig. 2-4 show the scoring of lexical items across all output in target<br />

scenes for each language. As expected, the cumulative percentage of target verbs across all<br />

scenes considerably exceeds the percentage of both general (non basic-level) motion verbs as<br />

well as the percentage of other (more specific) gait or path-shape verbs. Thus, there is a close<br />

match across the three languages with a total of 68,97% of target verbs for English, 73,49%<br />

for Norwegian and 68,75% for Bulgarian.<br />

A possible confound concerning this result is that basic level motion verbs are generally<br />

phonologically shorter than the more concrete (manner) verbs (cf. e.g., walk vs. meander),<br />

and subjects might have resorted to these responses for brevity. However, this is counterevidenced<br />

by our data. As a matter of fact, many responses did contain rather detailed<br />

descriptions, including prepositional phrases, manner phrases etc. An interesting observation<br />

concerns the use of general motion verbs in Norwegian at a mere 0,43%. Most likely, this<br />

result confirms our prediction that gå which covers both go and walk pre-empts the need to<br />

use other lexical items above the basic-level category. <strong>In</strong> addition, in responses for all three<br />

languages we find verbs describing not the type of motion in the scene, but rather its purpose<br />

(e.g. play, plan/intend to take a bath). This finding is highly coherent with ideas put forward<br />

in Tomasello (2001) and Zacks & Tversky (2004) that human categorization is highly<br />

susceptible to the teleology of events, and that humans attend from very early on to agents’<br />

intentions and goals. Furthermore, recent research in infant cognition based in eye-tracking<br />

20


has demonstrated that infants as young as 12 months can predict other people’s action goals<br />

(cf. Falck-Ytter, Gredebäck & von Hofsten 2006).<br />

The category of “other motion verbs” in the histograms in figures 2, 3 and 4 include<br />

non-target motion verbs encoding a more specific gait pattern (e.g., strut, step, lollop) or<br />

verbs encoding path-shape, such as circle, Bulgarian obikaljam (go round) or verbs denoting<br />

tandem motion (e.g., chase) and the like.<br />

4.4.1 Running scenes<br />

The use of running verbs in all three languages and across all 29 scenes was limited only to 9<br />

out of the 29 target scenes, which were defined as running scenes based on the pre-screen<br />

results. As predicted, the use of target verbs in Bulgarian was divided between the two target<br />

verbs, while English and Norwegian (with the exception of one case mentioned below) used<br />

only one verb. The frequency of each Bulgarian verb alone was much smaller, compared to<br />

that of the single main running verb in English or Norwegian, but the combined frequency of<br />

the two verbs is similar to that of the single verbs in the other two languages.<br />

Both Bulgarian run-verbs appear in the description of each running scene, but there is a<br />

preference for ticham (run1) over bjagam (run2). The only scene that stands apart from this<br />

pattern is the one of a chimpanzee disappearing rapidly into a wood, moving away from the<br />

camera. Most likely, in this scene, subjects have focused on the ‘directed motion’ feature of<br />

the scene, with the chimpanzee appearing as running away from danger. Also, Bulgarian<br />

showed a clear difference from English and Norwegian in the naming of the scene of a dog<br />

running round a tree. <strong>In</strong> this case Bulgarian subjects used also verbs encoding path-shape (e.g.<br />

obikaljam –‘go round’), or figure rotation (vurtja se – ‘spin around’) at 18,75%, in addition to<br />

(basic-level) biological motion verbs at 47,5%. <strong>In</strong> English, run is the main verb used for all<br />

running scenes in 83,33% of the cases. Although a variety of other verbs of running were used<br />

in the naming of these scenes, their frequency in comparison with run was negligible (8,33%).<br />

<strong>In</strong> Norwegian, løpe was the main verb used to describe running scenes. Two of the<br />

participants used consistently another verb (springe) instead of løpe, stating that the verb løpe<br />

did not exist in their dialect of Norwegian. Some of the other participants reported hesitation<br />

between løpe and springe, but used løpe in the end. 9 This shows that springe and løpe may be<br />

viewed as a synonymous pair, similar to the Bulgarian pairs of motion verbs in this study,<br />

however with the distinction found not in the lexicon of the language, but rather in<br />

9 This was confirmed in a brief interview after the experiment.<br />

21


individual/sub-system (dialectal) lexicons. As a result, in the study we have treated<br />

løpe/springe as individual variants of the same verb.<br />

Another interesting observation is that, in Norwegian, for two scenes (the man running<br />

on the spot, the dog on a treadmill), the more specific manner of motion verb jogge (‘jog’)<br />

was used (for the man running on the spot in 68,75% of the cases even surpassing the<br />

frequency of løpe at 31,25%). There are two things which might have caused this deviation:<br />

the running is done in a slower, steady pace, and it could be straightforwardly interpreted as a<br />

recreational activity, in which case Norwegian employs a verb borrowed from English,<br />

namely jogge.<br />

The scenes of a man running on the spot and a dog running on a treadmill stand apart<br />

also in another way due to the absence of translational motion. Except for jogge, this did not<br />

cause a preference for a different verb, but resulted in the addition of adverbial phrases with<br />

the meaning ‘on the spot’/ ‘on a treadmill’ for all three languages. We discuss this finding in<br />

section 5. below.<br />

4.4.2 Walking scenes<br />

Although walking verbs were used throughout all scenes, the main bulk applied to the<br />

nine walking scenes, with occasional use for other scenes, mainly the ones in the ‘grey zone’<br />

between walking and crawling 10 . The difference between the number of walking verbs used<br />

in walking scenes and the number used across all scenes was minimal (with Bulgarian walk1<br />

used at 85,71% and walk2 at 87,84%, Norwegian gå at 87,83% and English walk at 93,15%),<br />

showing that the target verbs indeed encode the basic-level concept of low-to-normal pace<br />

space traversal. Again, both Bulgarian verbs co-occurred in the naming of each scene but one<br />

of them, hodja (walk1), was the preferred one. Across all languages the scenes of humans<br />

walking elicited the highest target verb scores, supporting ideas about a fundamental<br />

difference between the identification of human motion and motion performed by other species<br />

(cf. Thornton & Vuong 2004, Pavlova et al. 2001 for adult categorization, Shipley 2003).<br />

The answers elicited by a scene of a woman walking backwards, did not show a<br />

difference in target verb use, but target verbs were always accompanied by satellites/<br />

adverbial phrases. <strong>In</strong> Bulgarian also a path verb meaning ‘regress’ occurred once.<br />

<strong>In</strong> four scenes there were lower frequencies of the target verbs for some or all of the target<br />

languages. The scene of monkeys walking around a tree elicited a significant use of target<br />

10 These included a tortoise walking very slowly and a beetle crawling/walking on a twig.<br />

22


verbs only in Norwegian. <strong>In</strong> Bulgarian the target verbs were totally absent, but the answers<br />

contained verbs encoding path-shape (at a total of 43,75 % for that scene) or figure rotation<br />

(12,5%). <strong>In</strong> English, both target verbs and path-shape verbs were used (the latter at 41,67 %<br />

for that scene).<br />

Two other walking scenes – that of a crocodile, and that of a long-legged bird – elicited<br />

few target verbs. We assume that the relevant factor for these two scenes was the atypical<br />

movements of the body or head that accompanied the gait. This triggered the use of more<br />

specific manner of motion verbs (included in Fig. 2-4 as ‘other motion verbs’), like English<br />

strut, waddle, Norwegian sprade (‘strut’), vagge (‘sway’), Bulgarian razhozdam se<br />

(‘promenade’), kracha (‘pace’). <strong>In</strong> Norwegian also the factor of substrate showed to be<br />

important in one scene – that of a chameleon walking along a tree branch: For this scene, the<br />

most frequently occurring Norwegian verb was balansere (‘balance’) at 37,5 % . Another<br />

interesting result was that in Norwegian the verb gå was used at 50% also for a scene showing<br />

a beetle, which the other two languages predominantly described as an instance of crawling.<br />

4.4.3 Crawling scenes<br />

Generally, the three languages exhibited much more variation in the naming of crawling<br />

scenes compared to the two types already discussed. All the expected crawl verbs were used<br />

in the context of scenes of crawling. The image of a human crawling on all fours elicited<br />

systematic crawl verb responses, in the case of Norwegian a 100% of all responses for that<br />

scene yielded krabbe. <strong>In</strong> Bulgarian, both pulzja (crawl1) and lazja (crawl2) were used in the<br />

naming of the images of humans crawling. However, across all scenes, as expected, the verb<br />

pulzia (crawl1) had a much higher frequency than lazia, because of the underspecification for<br />

[+ directed motion]. Also being more general (and underspecified) of the two, pulzia was the<br />

only verb used to name scenes of limbless locomotion (snakes), and the one predominantly<br />

used where the body of the agent has contact with the surface. Both verbs were almost equally<br />

acceptable in scenes based in a clear limb cycle (with species that had limbs).<br />

The two snake scenes showed the greatest variation and differences across the three<br />

languages in terms of choice of verb. Only in Bulgarian, they were characterized<br />

overwhelmingly by target verbs. <strong>In</strong> English the verb predominantly used for the snake moving<br />

in a ‘default’ way (technically called ‘lateral undulation’) was slither, with clear emphasis on<br />

the friction between the body and the substrate. <strong>In</strong> Norwegian the same scene and the scene of<br />

the man crawling on his belly were predominantly characterized by verbs denoting the<br />

23


winding body- and, consequently, path-shape (åle seg – ‘move like an eel’). Also non-humans<br />

(beetles, tortoises) elicited non-coherent responses in terms of target verbs.<br />

Norwegian stands apart from the other two languages in the classification of crawling<br />

scenes. The verb krabbe seems to be restricted to humans and is applied exclusively to human<br />

locomotion, with a non-significant use for other legged creatures (a tortoise), but not for<br />

insects. For locomotion scenes of legged creatures, several verbs were used (krype – ‘creep’,<br />

kravle ≈ ‘crawl’) but none of them was given significant preference.<br />

4.4.4 Climbing scenes<br />

The three climbing scenes elicited the target verbs for English and Norwegian, while for<br />

Bulgarian this applied primarily to the climbing up scene. As mentioned previously, English<br />

and Norwegian use only one verb to cover motion both up and down, confirmed by<br />

preferences in our experiment, while Bulgarian uses two different verbs, one for each vector<br />

orientation. Bulgarian results differ from those of the other two languages in that other verbs<br />

were used as well: kachvam se –‘go up’ used at 31.25% and spuskam se –‘go down’ at 25%<br />

of verbs used for climb scenes (both verbs can be classified as ‘pure-path’ [+directed motion]<br />

verbs).<br />

The scene of a koala climbing a tree scored highest for target verbs in English and<br />

Norwegian, while the scene of a koala climbing a tree in small hops scored highest for the<br />

Bulgarian verb katerja se. The scene of a koala climbing a tree in small hops scored lower in<br />

English, where in 25% of the cases the subjects chose hopping as the main activity in the<br />

scene’ (compared to 12,5% in Norwegian and none in Bulgarian). As expected, in the<br />

climbing down scene, there was a difference between the three languages. For Bulgarian, both<br />

[+directed motion] verbs and the basic-level motion verb specific to climbing down were<br />

used. <strong>In</strong> English and Norwegian, though the target verbs were used consistently, the naming<br />

commonly included prepositional phrases and satellites, such as Norwegian nedover/ English<br />

down.<br />

4.5 Cluster analysis<br />

<strong>In</strong> order to check the results obtained through the percentage distribution of verbs across<br />

scene types in responses we conducted a cluster analysis and calculated Simpson’s diversity<br />

index (D), essentially following the idea in Majid et al.(2007), however employing a different<br />

linkage method. Cluster analyses have proved to be very reliable in revealing patterns of<br />

grouping in collections of objects, and thus potential similarity (cf. e.g., the seminal work by<br />

24


Tversky 1977), and are specifically appropriate in the field of matching perceptual stimuli<br />

(action scenes) to lexical items to reveal patterns of lexical preference, as shown in a recent<br />

cross-linguistic experiment (cf. Majid et al. 2007). <strong>In</strong> the analysis we matched languagestimulus-response<br />

to check the degree of similarity across motion scenes as revealed in verbs<br />

used in the responses. The main aim was to check whether patterns in naming reflected any<br />

commonality in the stimuli on the general assumption that the stimuli were analysable in<br />

terms of features of the motion scene displayed. Thus, the general idea was that the pattern of<br />

clustering of motion scenes according to verbs used would provide evidence for the<br />

conceptual features that underlie the respective lexical choice for each language, and,<br />

respectively, across the three languages. The method we used in our analysis was hierarchical<br />

agglomerative clustering with average linkage. We employed a multiset distance measure<br />

formally defined in Appendix 3.<br />

The clustering of data according to the multiset distance appears to reflect the data in the<br />

current study somewhat more accurately than the average linking of the Jaccard distance 11 , as<br />

indicated by the higher Cophenetic Correlation Coefficients (CPCCs) for the clusterings with<br />

the multiset distance (respectively C= 0.991 for English, 0.986 for Norwegian and 0.966 for<br />

Bulgarian, compared to C=0.876, 0.894 and 0.916 for the Jaccard analysis for the same<br />

languages). 12 We reason that this is due to the fact that the multiset distance measure also<br />

takes into account the frequency of occurrence, which we believe is more appropriate in<br />

handling experimental linguistic data reflecting language usage. 13 For the sake of comparison,<br />

however, we also applied a Jaccard distance measure to the data.<br />

The cluster analysis generally confirmed the results obtained in the percentage<br />

(histogram) analysis. For all three languages, the more similar the stimuli were in terms of the<br />

basic features adopted in our analysis, the closer clustering was obtained in terms of verbs<br />

used. Thus the clustering confirmed the features we had preliminarily selected for the<br />

conceptual analysis of verbs of biological motion briefly introduced in 3.5 above. An<br />

interesting, but not unexpected result was that Bulgarian displayed a much lower average<br />

Simpson’s index (D) than English or Norwegian (0.39 vs. 0.56 and 0.62) meaning that the<br />

11 The Jaccard measure is based in a binary vector for each stimulus (1 or 0) whereby a single occurrence of a<br />

verb is sufficient for a similarity result. Thus the Jaccard measure will produce a 100% similarity between the<br />

two sets {walk, walk, walk, walk, run} and {walk, run, run, run, run} despite the difference in verb frequency. <strong>In</strong><br />

the study in Majid et al. (2007) a Jaccard measure was employed, apparently producing reliable results in view<br />

of the small populations/sets the study handled.<br />

12 Majid et al. (2007) also used average linkage for the clustering.<br />

13 Thus from the same two sets {walk, walk, walk, walk, run} and {walk, run, run, run, run}, with set<br />

intersection {walk, run} and union {walk, walk, walk, walk, run, run, run, run} the multiset measure yields a<br />

similarity of 2/8 = 0.25.<br />

25


former has higher diversity in the motion lexicon. This result is highly coherent with the data<br />

and observations at the outset, namely the two-verb split for each motion scene. An inspection<br />

of the two plots revealed another interesting general result worth mentioning. While the<br />

multiset–based cluster analysis provides a grouping based on core similarities across stimuli<br />

and scene types, the Jaccard-based analysis, provides grouping based on similarity regarding<br />

marginal features, a fact which may be useful in planning future research and obtaining<br />

subtler results. We address the more detailed results in the discussion in section 5. below (cf.<br />

also <strong>Dimitrova</strong>-<strong>Vulchanova</strong>, Matínez, Edsberg & Eshuis, in preparation). The dendrograms<br />

for each language are enclosed in the Appendix.<br />

5. Discussion<br />

5.1 General comments<br />

Our basic results confirm that English and Norwegian have one basic term per motion scene<br />

(modulo the crawl verbs in Norwegian), and confirm the two-way split for each motion scene<br />

in Bulgarian. The target verbs in all three languages were used for the naming of basic motion<br />

scenes involving humans, with canonical figure orientation (front forwards), vector<br />

orientation (left-to-right or towards), and underspecified (straight path shape) performed at<br />

characteristic rate. Most scenes featuring a variation from the default on one or more of these<br />

conditions produced an increase in the range of lexical items preferred across subjects. Below<br />

we discuss the most salient findings. We first consider the selection of factors that have<br />

influenced responses for a particular scene type and for particular lexical items.<br />

5.2 Factors<br />

5.2.1 Biological factors – cycle, body and limb structure, species<br />

The most distinctive feature of at least three of the four types of targeted biological motion<br />

(running, walking and crawling), was their cycle – the number of limb movements, and the<br />

pattern of limb movement - the way the Agent moves its limbs and body, in order to achieve<br />

translational motion. <strong>In</strong> this respect, two factors were of utmost importance: suspended vs.<br />

supported motion (i.e. running vs. walking/ crawling), and erect posture vs. ‘supine’/ ‘lower<br />

than usual’ posture (i.e. walking vs. crawling).<br />

Our prediction that the naming of the target scenes would consistently reflect the value<br />

of the parameter supported vs. suspended was borne out. All scenes with suspended motion<br />

were overwhelmingly identified with verbs of running in all three languages. <strong>In</strong> addition, the<br />

cluster analysis revealed a major split between the clustering of suspended (running) scenes<br />

26


and respectively the verbs naming them and the supported motion scenes and verbs (e.g.<br />

walking and crawling). Further support for the relevance of the feature supported vs.<br />

suspended comes from the cluster analysis which clearly displays a greater semantic closeness<br />

between crawl and climb seen in how these verbs cluster on a jaccard similarity plot in<br />

English and Bulgarian. Apparently, the common denominator here is supported motion in<br />

both cases, usually referred to as ‘clambering’ in the context of climb.<br />

For the two types of supported biological motion, walking and crawling, a further<br />

dividing factor may prove to be posture. While in walking, the default value is erect posture,<br />

or at least erect limbs and some distance between the torso and the ground, in crawling the<br />

default is ‘supine’ posture, non-erect limbs, and nearness or contact with the ground. These<br />

features are clearly present, and respectively influence, the naming of walking vs. crawling of<br />

humans. <strong>In</strong> other cases, however, it is unclear what the difference between ‘erect’ and<br />

‘supine’ means, as the species in question have only one posture available. This was<br />

demonstrated in the naming of the scenes of a beetle, and a tortoise, responses to which varied<br />

on the walking – crawling continuum. Likewise, for images of quadrupeds moving at a<br />

moderate pace (e.g., a koala, a tiger and a chimp) the most common verb employed was walk<br />

across all three languages, despite the non-erect posture. Most likely, in this case the decisive<br />

feature is ‘head-up’, probably levelling the otherwise observable distinction between figure<br />

posture and figure orientation.<br />

Special patterns of locomotion can elicit more concrete verbs. For instance, for the scene<br />

of a crocodile walking towards water, some English and Norwegian participants chose to<br />

stress the sideways swaying motion of the body by using the verbs waddle (Eng)/ vagge (No).<br />

<strong>In</strong> the scene of a long-legged bird walking, English and Norwegian participants preferred the<br />

verbs strut (Eng)/ sprade (No) to the target verbs, because of the specific mode the bird<br />

carried its head and shoulders. <strong>In</strong> Bulgarian, some participants put an emphasis on the<br />

separate steps, by using the verb kracha (‘stride’).<br />

An interesting observation was the difference across the three target languages in the<br />

focus on features encoded in the respective verbs used. This was particularly salient in the<br />

scenes of snake locomotion, where one and the same motion event was lexicalized differently<br />

in each language, selecting different features. <strong>In</strong> English, the most salient feature was the<br />

friction occurring between the body and the ground (ex<strong>press</strong>ed in the verb slither). <strong>In</strong> contrast,<br />

in Norwegian the most salient feature was the sideways undulation of the body (ex<strong>press</strong>ed by<br />

the verb åle seg – ‘move like an eel’), the latter apparently a ‘metaphor’ for path shape<br />

specification. <strong>In</strong> Bulgarian both of these features were ignored, and the scene was lumped<br />

27


together with the other cases of ‘supine’ locomotion. This however, does not mean that each<br />

respective language lacks the means to ex<strong>press</strong> the features selected by the other two. All<br />

three languages have verbs covering each of the three features (English crawl, wind;<br />

Norwegian krype, skli – ‘slide’; Bulgarian izvivam se – ‘wind’, pluzgam se –‘slither/slide’),<br />

but they are not habitually applied to the situation in question.<br />

A biological factor which proved to be important was species, though it did not have<br />

equal weight in all cases. Thus, for Norwegian krabbe, ‘species’ is a crucial feature,<br />

overriding cycle or posture, with the verb used exclusively for humans, and with high<br />

uniformity of naming for the human crawling scenes. An interesting question is to what extent<br />

motion categorization is driven by features (e.g., shape, size) of the object moving or by the<br />

pattern of movement, similar to the influence of object features on object categorization (cf.<br />

Madoley & Smith 2005). We are currently investigating this option in a controlled experiment<br />

(cf. Coventry et al., in progress).<br />

The domains of walking and crawling generally showed to be ‘anthropocentric’, with<br />

the clips of human locomotion scoring highest in the use of target verbs. The variation and<br />

uncertainty in the naming of motion by non-human species, such as insects and reptiles, can<br />

be attributed exactly to this anthropocentric emphasis at the core of lexical items denoting<br />

walking and crawling, clearly supporting the view in Shipley (2003, current research). If<br />

distinctions such as ‘erect’ and ‘supine’ are not relevant for the species, the choice between<br />

existing verbs will be difficult. At the same time, it will be difficult to use any other verb, as<br />

our whole lexicon is generally anthropocentric, reflecting the perception and categorization of<br />

human locomotion, which prove to be easier, if the one categorizing and naming also has<br />

relevant experience in planning and executing the same locomotion patterns (cf. Shipley<br />

2003, Pavalova et al. 2001, among others). Also, since categorization is a learning-based<br />

process (Giese & Poggio 2003, Giese 2004), it will be natural for a particular type of<br />

biological motion to be most commonly associated with the species that exhibit it most often<br />

in our experience (for the clustering of naturally co-occurring features in categorization, cf.<br />

Rosch 1976).<br />

5.2.2 Velocity and substrate<br />

A factor directly related to the cycle of biological motion is velocity. This is theoretically<br />

supported by the Froude equation, where speed, alongside gravity and limb length, is one of<br />

the three main parameters that influence gait pattern and gait-transition (cf. Alexander 1989,<br />

1996). It has been discovered that the transition from supported to suspended gaits happens<br />

28


across species at particular Froude numbers, which means that, generally, supported<br />

locomotion will correspond to a lower velocity, while suspended motion corresponds to<br />

higher velocity. Thus, velocity is a major factor distinguishing walking from running, with<br />

even greater variation across gait-types in quadrupeds (e.g. walk vs. amble, trot vs. pace,<br />

canter vs. gallop). Our experiment was not explicitly designed to check for the role of<br />

velocity, as we used scenes shot in natural surroundings which were not subject to any<br />

manipulation 14 . There are some indications, however, that each of the three main gait types:<br />

walking, running and crawling has its ‘default’ velocity: high for running, low for crawling<br />

and normal-to-low, for walking. <strong>In</strong> some verbs of biological motion, velocity is more salient<br />

than in others, even to the degree that all other meaning components of these verbs can be<br />

factored out in extended uses where the verbs denote purely speed. Thus, English run and<br />

Bulgarian ticham/ bjagam are often used to indicate high velocity and ease of motion, rather<br />

than a specific suspended cycle (9a, b). English crawl, Bulgarian pulzja/ lazja, and Norwegian<br />

krype (but never krabbe), on the other hand, are used to denote slow and laborious motion (9<br />

c, d, e).<br />

(9)a. She ran over to the shop to get some sugar (English)<br />

b. Tja iztiča do magazina za da kupi zahar (Bulgarian)<br />

c. Traffic crawls along at 10 miles an hour. 15 (English)<br />

d. Kolata pulzeshe edvam-edvam (Bulgarian)<br />

“The car was barely crawling”<br />

e. Lastebilen krøp oppover stigningen / Temperaturen krøp oppover (Norwegian)<br />

”The truck was creeping up the slope” / “The temperature was creeping up”<br />

<strong>In</strong> addition, the dendrograms in the cluster analysis revealed a basic split between<br />

lexical items naming high speed scenes (e.g. run) vs. all others (e.g. walk and crawl) the latter<br />

both performed at a normal-to-low speed. These results confirm independently the salience of<br />

velocity in the identification and naming of basic-level biological motion, with a clear split in<br />

the lexicon between items for fast motion vs. items for slow or normal, clearly making motion<br />

at high velocity the non-default feature in the pair.<br />

Some of the data in our experiment also suggest that, when not overtly modified, the<br />

target verbs refer to a particular speed range, while divergence from the default is usually<br />

14 Except for one clip which was generated backwards.<br />

15 Example from Merriam Webster Online.<br />

29


specified explicitly through modification. The responses elicited by one of our target scenes (a<br />

tiger walking slowly) in all three languages displayed a high degree of modification with<br />

adverbial phrases spelling out the unusually slow velocity of the action.<br />

5.2.3 Presence vs. absence of translation<br />

The presence of a translational component could not be checked across all types of motion in<br />

our set up. There were only two scenes explicitly targeting this parameter, a man running on<br />

the spot and a dog running on a treadmill, where the translational motion parameter was<br />

removed. Prior to the experiment we hypothesized that the concept of run is a kind of dot<br />

object concept (cf. Pustejovsky 1995, and Jackendoff 2002), deriving its conceptual structure<br />

as inherited from two main types, biological motion (a manner verb) and translational motion<br />

(a path verb), as indicated in (10). 16<br />

(10) [biological motion ● translational motion]<br />

Thus, English run can be used with both notions feeding the mapping to syntax and<br />

interpretation demonstrated in the examples in (11) below.<br />

(11) a. John ran.<br />

b. The motor was running.<br />

c. The river runs parallel to the road.<br />

The example in (11a) is ambiguous between the traversal motion interpretation and the<br />

pure “manner” basic-level biological motion interpretation, especially in the absence of overt<br />

path specification. <strong>In</strong> the example in (11b), however, in an extended use of run, the<br />

translational motion component has been factored out, thus ex<strong>press</strong>ing only the complex<br />

pattern of movement, in this case roughly meaning “work/be in action”. <strong>In</strong> contrast, in (11c)<br />

we witness an instance of the “manner” component being removed, with a focus on the<br />

traversal component, in this case meaning “extends”. <strong>In</strong> terms of the system put forward most<br />

recently in Jackendoff (2002), this would be an instance of the EXT (extension) function<br />

represented formally as EXT (x, Path) which includes a path and applies to non-temporal<br />

ex<strong>press</strong>ions. If metaphorical extensions based in run can feature both components<br />

16 This is also recognized in Berman & Slobin (1994) for some “manner” verbs.<br />

30


independently, there is sufficient ground to assume that both are originally present in the<br />

conceptual structure of the verb. Otherwise it remains a mystery what the source of the<br />

respective extensions in (11b-c) can be.<br />

Our expectations regarding the two scenes where the path-traversal component in run<br />

had been factored out, were that subjects could still felicitously apply a run verb in the<br />

naming task. Our hypothesis was confirmed by the clusering in the dendrograms, whereby the<br />

multiset average plot for both English and Bulgarian yielded a high similarity score for the<br />

two stimuli (0,8 for Bulgarian and 0,65 for English) with both branching from the same node.<br />

Likewise, for all three languages the jaccard plot displayed a high degree of similarity<br />

between the two in-place scenes and the central run scenes (e.g., a woman running). It is<br />

however worth noting that the translational interpretation is the default one when the target<br />

run-verbs are used in their basic-level biological motion sense. Thus, in our experiment, while<br />

the two non-translational run scenes elicited overwhelmingly target run verbs, most<br />

commonly this was accompanied by modification through adverbial phrases explicitly<br />

spelling out the absence of translation. We return to the connection between non-default<br />

values and modification shortly below.<br />

5.2.4 Axis and vector orientation<br />

Verbs of climbing clearly instantiate a group of biological motion verbs where ‘Manner’<br />

cannot be dissociated from Path. <strong>In</strong> Norwegian klatre, and Bulgarian katerja se/ slizam, both<br />

the contact of the limbs with the vertical surface and the translation in space are necessary<br />

features, without which no scene can be felicitously described as an instance of climbing. It is<br />

language specific, however, to what degree these verbs specify for a vertical path. Both in<br />

English and Norwegian the respective verbs encode motion along a vertical axis, but the<br />

direction is not specified, it can be either up or down (cf. the analysis for German<br />

steigen/klettern in Weisgerber, this volume). Still there is some evidence that motion upwards<br />

is the default (it is the direction implicitly understood when the verbs appear unmodified),<br />

while motion downwards is usually specified through modification with satellites or adverbial<br />

phrases. This was clearly demonstrated in responses to the climbing down scene in the<br />

experiment. <strong>In</strong> Bulgarian, each of the two climbing verbs is specified for a particular (vector)<br />

orientation along the vertical axis.<br />

Although the target verbs of walking, running and crawling appear to be underspecified<br />

for axis of motion, motion along the horizontal axis seems the default. Thus, some of the<br />

scenes where the surface was slightly inclined (the scene of a caterpillar, a snake, and a beetle<br />

31


moving along an upward 15-20’ angle path) elicited satellites overtly ex<strong>press</strong>ing this nondefault<br />

path orientation (English up, Norwegian opp, oppover), and in few cases triggered the<br />

use of climbing verbs. These results are coherent with the representation for German and<br />

English in Weisgerber (this volume) on which the path explcitly is specified for a difference<br />

in height between the starting point and the goal, irrespective of which way the difference<br />

goes (e.g., whether a positive or negative value). To the extent that a vertical path cannot vary<br />

on the features for vector orientation left-to-right/right-to-left and away/towards (at least not<br />

linguistically), in our representation in (9) above, we conflate the two factors.<br />

5.2.5 Vector orientation and Figure orientation<br />

Although the concepts of objects and actions that humans entertain have been considered by<br />

some authors as three-dimensional, and view-independent (Jackendoff 1996), recent studies<br />

have shown that objects, as well as events have privileged views, which are recognized most<br />

easily. Recent studies (Giese 2004b) show that biological motion (especially walking) is best<br />

recognized in a left-to-right view (as opposed to motion form right to left, towards the viewer<br />

and away from the viewer). Our experiment was not explicitly designed to test the parameter<br />

of vector orientation, but it displayed evidence in support of the ‘privileged’, default view<br />

idea. Thus, two scenes representing motion away from the camera produced a large variation<br />

in naming results. The scene of a chimpanzee running away from the camera, and the image<br />

of a woman walking right-to-left, both being instances of non-default vector orientation,<br />

“jammed” coherent responses and triggered overt ex<strong>press</strong>ion of the non-default value.<br />

As for Figure orientation, the default in horizontal motion is ‘head up’, facing in the<br />

direction of motion. If the default is to be overridden it is specified overtly by the use of<br />

adjuncts. Thus, in the scene of a sloth climbing down where the climbing was head down,<br />

quite often subjects resorted to overt modification. <strong>In</strong> contrast, the default setting was never<br />

overtly specified, e.g. for the scenes with a koala climbing head up. These findings support<br />

the lexical encoding model proposed in Koenig et al. (2003) whereby lexically-encoded<br />

semantic features do not necessarily coincide with overt syntactic realization. Also, our results<br />

suggest that when more than one parameter is set to non-default setting, subjects` responses<br />

are less coherent and less focussed on the primary lexical exponents of the category. Thus,<br />

when non-default vector orientation is combined with non-default figure orientation,<br />

responses spread across a wide range of lexical items, rather than triggering the expected<br />

target verbs.<br />

32


5.2.6 The salience of path shape and direction<br />

According to Nikanne & van der Zee (2005) and Nikanne & van der Zee (this volume), Path<br />

encoding in languages can be represented at three grain levels: neutral, global, and local. At<br />

the neutral grain level, path shape is underspecified, ex<strong>press</strong>ing directed motion, where the<br />

path is seen as the shortest/ unmarked/ default route between the point in space occupied by<br />

the Figure, and the Ground, as in the English verbs go, arrive, enter, the Norwegian verbs<br />

komme, and the Bulgarian verbs otivam ‘go’, idvam ‘come’. The global grain level can<br />

provide general and non-composite specifications of the trajectory of motion, such as e.g., a<br />

curve (in English – circle, swerve, veer, Bulgarian - obikalyam ‘go round’, zavivam ‘make a<br />

turn’, krazha ‘circle’). At the local grain level, path shape is specified as composed of smaller<br />

segments, which may or may not lead in a definite general direction (in English - wind, spiral,<br />

zigzag, in Bulgarian krivolicha, lakatusha, viia se ‘go in an irregular path’, in Norwegian -<br />

bukte seg, slynge seg ’wind ). The overarching prediction is that path (-shape) supersedes<br />

other parameters in the categorization of motion scenes and, more specifically in the linguistic<br />

encoding of motion. Our experimental data included two scenes designed to check for the<br />

relevance of this parameter. The scenes showed animals walking/ running in a circular path.<br />

The results for the three languages were different.<br />

For Bulgarian, path shape indeed did override other ‘Manner’ parameters, such as e.g.<br />

cycle, inducing the predominant use of verbs denoting path shape in the case of the scene of<br />

monkeys walking around a tree (at 56,25% for that scene), but not in the scene of a dog<br />

running around a tree. We hypothesise that the overriding factor in the latter case was the high<br />

velocity of motion. This result also raises the issue of a possible hierarchy of factors in motion<br />

identification to be attested in future research.<br />

<strong>In</strong> English, path-shape verbs were also used, but not so commonly (41,67% for the<br />

monkeys-walking-round-tree scene). <strong>In</strong> Norwegian, path shape verbs were not used at all with<br />

participants giving a clear preference to biological motion verbs (75%). The important<br />

observation, here is that for both scenes across the three languages, path shape was overtly<br />

spelled out through the use of adverbial ex<strong>press</strong>ions.<br />

The above data show that though path shape is an important factor in motion<br />

conceptualization and encoding in language, its salience in verb choice differs across<br />

languages. From the three target languages in this study, Bulgarian has the strongest tendency<br />

to give preference to this factor, while Norwegian has the lowest, with English occupying an<br />

intermediate position on the scale. Thus, in Bulgarian the stimuli featuring a non-default pathshape<br />

form their own branch in the dendrograms from the cluster analysis, while in both<br />

33


English and Norwegian they are grouped with other scenes of similar gait pattern, with<br />

subjects resorting to basic level biological motion verbs in the naming. The use of adverbial<br />

modifiers also suggests that there is a default value (a straight path), the divergence from<br />

which has to be explicitly realised in language.<br />

5.2.7 The use of satellites<br />

The general pattern observed in our study was that satellites are indeed very frequent in all the<br />

languages under consideration, thus confirming Talmy’s (1985) general proposal for a<br />

typology in terms of this property. However, our data produced a surprising result in regard of<br />

expectations concerning the role of satellites. Generally, it has been assumed in the literature<br />

(cf. Talmy 1985, 2000, Bowerman et al. 2002, however see Slobin 2004 for a critical<br />

discussion) that satellites serve to ex<strong>press</strong> overtly the specification of the path parameter (here<br />

referred to as [+/- Path]) in the group of languages called satellite-framed languages (with<br />

English and German usually cited as good examples). The free naming task in our study<br />

provided a good opportunity to check this tendency, as our set up was very similar to (but not<br />

exactly replicating) the recent tradition of “frog stories” across languages (cf. Berman &<br />

Slobin 1994, Slobin 2004). <strong>In</strong>deed, two of the languages in our study, English and<br />

Norwegian, moderately confirmed the role traditionally assumed for satellites, in that subjects<br />

had chosen to describe the scenes at hand by using a path satellite along with the head verb.<br />

However, our data revealed quite an orthogonal tendency, especially for Bulgarian. Responses<br />

for the latter systematically used satellites in the cases of non-default settings, that is, when a<br />

default parameter was modified. <strong>In</strong> all three languages, satellites explicating the removed<br />

[Path] parameter were used for the scene of a man running on the spot, therefore we believe<br />

that this is a tendency for all three languages in this study. <strong>In</strong> English and Norwegian,<br />

however, the tendency was obscured in the data for two reasons. Firstly, it is impossible to<br />

isolate the pure non-default setting explication uses from the general tendency for explicating<br />

the path in Germanic. Secondly, both English and Norwegian only have one major verb per<br />

scene, thus making the path parameter difficult to test independently. <strong>In</strong> Bulgarian, in<br />

contrast, the competition of two verbs provides a good testing environment for the tendency at<br />

hand. We propose non-default explication function as a preliminary label for this tendency.<br />

6. Concluding remarks<br />

The current study has provided input to the main objectives we set out to investigate. We<br />

obtained good evidence for the relevance and respective salience of the factors in biological<br />

34


motion scenes that we had selected for the conceptual representation and lexical encoding of<br />

such scenes. Our experiment has also demonstrated that subjects generally prefer the basiclevel<br />

biological motion verbs to both more general (super-ordinate) and more concrete (underordinate)<br />

verbs. Our general results have shed some light on the issue of universality of<br />

cognitive concepts and their encoding in language. Our data provide support to the view that<br />

many conceptual primitives are indeed based in universal cognitive categories. The main<br />

difference across languages resides in how these features/primitives are encoded (or ‘bundle’<br />

to use Jackendoff’s term) in lexical items and underdetermine their felicitous use. As it turns<br />

out, a possible distinction might be along the lines of how these features split into sufficient<br />

and necessary conditions, and whether these conditions apply in conjunction or disjunction<br />

(cf. a new function proposed by Jackendoff 2002 for the latter type of disjunction).<br />

Apparently, there are robust features that underlie the core concepts (which are also<br />

semantically relevant) confirmed in our study by the overall high confidence level for<br />

canonical (default) scenes across all three languages. Our data also indicate that in naming<br />

processes there appears to be a prototypical packaging of the features that characterize<br />

biological motion (e.g., walking upright front-forwards left-to-right, as also shown in Jellema<br />

et al. 2002), while marginal members of the category are notoriously difficult to name. An<br />

interesting question here is what is more salient in motion identification, the motion pattern<br />

(cycle) or the body-shape/form, or rather the combination of both (cf. Madoley & Smith 2003<br />

for a developmental study). <strong>In</strong> our study, non-default settings of more than one parameter<br />

dramatically influence subjects’ responses or produce what we call “jamming” effects. The<br />

latter were also observed with images showing “nonsensical movements”, such as e.g., a<br />

baby-tortoise struggling to walk on a sand beach. That is, movements that are biologically<br />

possible (e.g. following the laws of gravity, use of limbs etc.), which are, however, rarely<br />

observed or culturally non-salient. 17 <strong>In</strong> such cases the preferred naming strategy is to use a<br />

verb at grain level 0 (underspecified on all parameters) or try out obsolete or rare words.<br />

Satellites would be used in conjunction with level 0 verbs or basic-level biological motion<br />

verbs when more than one parameter is set at a non-default value. This strategy is clearly at<br />

deviance from previously suggested functions of satellites, and we choose to label it the nondefault<br />

explication function. Future cross-linguistic research will provide further support to<br />

this tendency.<br />

17 These images included, among others, a penguin sliding down an ice slope on full body, a baby sea turtle<br />

trying to walk, a chimp baby trying to walk, a fish using its fins to “walk” out of water.<br />

35


References:<br />

Alexander, R. McNeil 1989. Dynamics of dinosaurs and other extinct giants. NY: Columbia<br />

University Press.<br />

Alexander, R. McNeil 1996. Optima for animals, ch. 3, pp. 45-64. Princeton: Princeton<br />

University Press.<br />

Ameka, F., & J. Esegbey 2000. Serialising languages: Verb-framed, satellite-framed or<br />

neither? <strong>In</strong> Larry Hyman and Ian Maddieson, eds., African Comparative and Historical<br />

Linguistics: Proceedings of the Thirty-second Annualn Conference on African<br />

Linguistics. Lawrenceville, NJ: Africa World Press.<br />

Apraku, Prince 2005. Conceptual Structure of Motion Events in Akan as Compared to<br />

English. MPhil thesis, <strong>NTNU</strong>.<br />

Bejan, A. & J. Marden 2006. Unifying constructal theory for scale effects in running,<br />

swimming and flying. The Journal of Experimental Biology 209:238-248.<br />

Berman, R. & D. Slobin 1994. Relating events in narrative: a crosslinguistic developmental<br />

study. Hillsdale, NJ: Lawrence Erlbaum Associates.<br />

Bowerman, M. & S. Choi 2001. Shaping meanings for language: universal and languagespecific<br />

in the acquisition of spatial semantic categories. <strong>In</strong>: Bowerman, M. & S.<br />

Levinson (eds.) Language acquisition and Conceptual development, Cambridge: CUP.<br />

Bowerman, M. et al. 2002. Putting things in places: Developmental consequences of linguistic<br />

typology, talk.<br />

Clark, E. 2001. Emergent categories in first language acquisition. <strong>In</strong>: Bowerman, M. & S.<br />

Levinson (eds.) Language acquisition and Conceptual development, Cambridge: CUP.<br />

Coventry, K. et al., in progress. Biological motion and object features: an experimental study.<br />

Dekova, R. 2006. Lexical Encoding of Verbs in English and Bulgarian, PhD thesis,<br />

Trondheim: <strong>NTNU</strong> trykk.<br />

Dekova, R. & M. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2006. Event lexicalization across languages. <strong>In</strong>:<br />

Koeva, S. & M. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> (eds.) Proceedings from FASSBL 5, Sofia,<br />

Bulgaria.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. 1996/99. Verb semantics, diathesis and aspect. München/<br />

Newcastle: Lincom.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. 1998. A Typology of Measures. <strong>In</strong>: Timo Haukioja (ed.) Papers<br />

from the XVIth Scandinavian Conference of Linguistics, pp. 50-63. Turku: Publications<br />

of the Department of Finnish and General Linguistics of the University of Turku 60.<br />

36


<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. 1999. <strong>In</strong>cremental Walls. Nordic Journal of Linguistics. 21 (1),<br />

pp. 1-16.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. 2003. The complexity of results. <strong>In</strong> Beerman, D. & L. Hellan<br />

(eds.) TROSS Proceedings, Trondheim.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. 2004a. Verbs of Motion and their Conceptual Structure, Motion<br />

Encoding Workshop, Åbo Akademi, Turku<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. 2004b Paths in Verbs of Motion, invited talk at Argument<br />

Structure CASTL Conference, November 4-6, 2004, Tromsø University.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. & L. Martínez (in preparation) Verbs of terrestrial biological<br />

motion in Akan – a pilot study.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. & E. van der Zee (in preparation) Run, walk, and crawl: a forcedchoice<br />

motion experiment.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M. & M. Weisgerber, in <strong>press</strong>. ‘Concept’, ‘Context’ and beyond:<br />

Manner of motion verbs encoding and relevant bits of information. <strong>In</strong>: <strong>Dimitrova</strong>-<br />

<strong>Vulchanova</strong>, M. & S. Koeva (eds.) Selected papers from FASSBL5, thematic volume,<br />

Southern Journal of Linguistics.<br />

<strong>Dimitrova</strong>-<strong>Vulchanova</strong>, M., L. Martinez, O. Edsberg, R. Eshuis. in preparation. A Cluster<br />

Analysis of Biological Motion Stimuli in English, Norwegian and Bulgarian.<br />

Falck-Ytter, T., G. Gredebäck & C. Von Hofsten 2006. <strong>In</strong>fants predict other people’s action<br />

goals. Nature Neuroscience, advance online publication.<br />

Foote, I. 1967. Verbs of Motion. <strong>In</strong> Studies in the Modern Russian Language. Cambridge:<br />

Cambridge University Press.<br />

Gennari, S. et al. 2002. Motion events in language and cognition. Cognition 83, 49-79.<br />

Geuder, W. & M. Weisgerber 2006. Manner and Causation in Movement Verbs <strong>In</strong>: Ebert,<br />

Chr. & C. Endriss (eds.) Proceedings from Sinn und Bedeutung 10, Berlin, ZAS Papers<br />

in Linguistics pp. 125-138.<br />

Giese, M. 2004a. Biological principles of the representation of complex shapes and<br />

movements. (http://www.uni-tuebingen.de/knv/arl)<br />

Giese, M. 2004 b. Learning as Principle of Action Recognition in Visual Cortex. Second<br />

Summer school for Cognitive Vision. Bonn, Germany 16th – 20th August 2004.<br />

(http://www.ipb.uni-bonn.de/events/summerschool04, last accessed 30 th Jan. 2005)<br />

Giese M. A. & T. Poggio 2003. ‘Neural mechanisms for the recognition of biological<br />

movements and action’. Nature Reviews Neuroscience 4, 179-192.<br />

37


Guentchéva, Z. 2002. The semantics and functions of prefixes. <strong>In</strong>: <strong>Dimitrova</strong>-<strong>Vulchanova</strong>,<br />

M., I. Krapova, D. Dyer & C. Rudin (eds.) Balkanistica 15, special issue on South<br />

Slavic and Balkan languages.<br />

Hellan, L. & M. <strong>Dimitrova</strong>-<strong>Vulchanova</strong> 2000. Criteriality and grammatical realization. <strong>In</strong>:<br />

Coopmans, P., M. Everaert & J. Grimshaw (eds) Lexical Specification and insertion,<br />

CILT series 197, 165-194, John Benjamins.<br />

Jackendoff, R. 1996. ‘The architecture of the Linguistic-Spatial <strong>In</strong>terface’. <strong>In</strong> P.Bloom, M.A.<br />

Peterson, L. Nadel & M.F. Garrett (eds.), Language and Space. Cambridge,<br />

Massachusetts: The MIT Press.<br />

Jackendoff, R. 2002a. What’s in the Lexicon. <strong>In</strong>: Nooteboom, S et al. (eds.) Storage and<br />

Computation in the Language Faculty. Dodrecht: Kluwer Academic Publishers.<br />

Jackendoff, R. 2002b. Foundations of Language. Oxford: Oxford University Press.<br />

Jellema, T, C.I. Baker, M.W.Oram, & D.I. Perrett (2002) Cell populations in the banks of the<br />

superior temporal sulcus of the macaque and imitation. <strong>In</strong> Meltzoff, A. (ed.), Imitative<br />

Mind: Development, Evolution and Brain Bases. NY: Cambridge University Press.<br />

Johansson, G. 1973. Visual perception of biological motion and a model for its analysis.<br />

Perception & Psychophysics, 14, 201-211.<br />

Lyons, J. 1977. Semantics, vol. 1. Cambridge: Cambridge University Press.<br />

Madole & L. Smith 2005. <strong>In</strong>: Carlson, L. & E. van der Zee (eds.). Oxford: OUP.<br />

Majid, A., M. Gullberg, M. van Staden & M. Bowerman 2007. Semantic categories in<br />

Germanic languages: “Cutting and breaking”. <strong>In</strong>: Journal of Cognitive linguistics. Vol.<br />

18-2.<br />

Mandler, J 1996. Preverbal representation of language, in Bloom P., M. Peterson, L. Nadel, &<br />

M. Garrett (eds.), Language and Space. Cambridge, MA: MIT Press.<br />

Matlock, T., & Richardson, D.C. (2004). Do eye movements go with fictive motion?<br />

Proceedings of the 26th Annual Conference of the Cognitive Science Society.<br />

Ndiwalana, M. 2003. Verbs of Movement in Luganda: A Frame Semantics and Sign Model<br />

Perspective. MA thesis, <strong>NTNU</strong>.<br />

Nikanne, U. 2002. What is in a construction? Talk at the 2 nd <strong>In</strong>ternational Conference on<br />

Construction grammar, Helsinki September 2002.<br />

Nikanne, U. 2004. Remarks on Ex<strong>press</strong>ions of Motion in Finnish. Motion Encoding<br />

Workshop, Åbo Akademi, Turku<br />

Nikanne, U. & E. van der Zee 2005. The grain levels in the linguistic ex<strong>press</strong>ions of motion,<br />

talk at 21 Scandinavian Conference of Linguistics, Trondheim.<br />

38


Papafragou, A. et al. 2002. Shake, rattle, 'n' roll: the representation of motion in language and<br />

cognition. Cognition 84, 189-219.<br />

Pavlova, M., Krageloh-Mann, I., Sokolov, A. & Birbaumer, N. (2001). Recognition of pointlight<br />

biological motion displays by young children. Perception, 30, 925-933.<br />

Pustejovsky, J. 1995. The Generative Lexicon. Cambridge, Massachusetts: MIT Press.<br />

Rosch, E.H. (1973) Natural categories, Cognitive Psychology 4: 328-50.<br />

Rosch, E., C.B. Mervis, W.D. Gray, D.M. Johnson, & P. Boyes-Braem 1976. Basic objects in<br />

natural categories. Cognitive psychology, 8, 382-439.<br />

Shipley, T.F. 2003. The effect of object and event orientation on perception of biological<br />

motion. Psychological Science, 14(4), 377-380<br />

Slobin, D. 1996a. From ‘thought and language’ to ‘thinking for speaking’. <strong>In</strong>: Gumperz, J., J.<br />

John & S. Levinson (eds.) Rethinking Linguistic Relativity. Cambridge: CUP, pp. 70-96.<br />

Slobin, D. 1996b. Two ways to travel: verbs of motion in English and Spanish. <strong>In</strong>: Shibatani,<br />

M. & S. Thompson (eds.) Grammatical Constructions. Their form and meaning.<br />

Oxford: Oxford University Press.<br />

Slobin, D. 2001. Form-function relations: how do children find out what they are? <strong>In</strong>:<br />

Bowerman, M. & S. Levinson (eds.) Language acquisition and Conceptual<br />

development, Cambridge: CUP.<br />

Slobin, D. 2004. The many ways to search for a frog: Linguistic typology and the ex<strong>press</strong>ion<br />

of motion events. <strong>In</strong>: Stromqvist, S. & L. Verhoeven (eds.), pp. 219-257.<br />

Snell-Hornby, M. 1983 Verb-descriptivity in German and English. A Contrastive study in<br />

semantic fields. Heidelberg: Carl Winter Universitätsverlsg.<br />

Stromqvist, S. & L. Verhoeven (eds.) 2004. Relating events in narrative, vol. 2 Typological<br />

and contextual perspectives. Mahwah, NJ: Lawrence Erlbaum Associates.<br />

Talmy, L. 1985. Lexicalization patterns. Semantic structure in lexical forms. <strong>In</strong>: Shopen, T.<br />

(ed.) Language Typology and Syntactic Description. Vol. 3: Grammatical Categories<br />

and the Lexicon, pp. 57-149. Cambridge: Cambridge University Press.<br />

Talmy, L. 2000. Towards a cognitive semantics. Cambridge Mass.: MIT Press.<br />

Thornton, I. & Q. C. Vuong 2004. <strong>In</strong>cidental Processing of Biological Motion, <strong>In</strong>: Current<br />

Biology.<br />

Tomasello, M. 1995. Pragmatic Contexts for Early Verb Learning. <strong>In</strong>: Tomasello, M. & W.<br />

Merriman (eds.) Beyond Names for Things: Young children's acquisition of verbs.<br />

Hillsdale,New Jersey/Hove, UK: Lawrence Erlbaum Associates.<br />

39


Tomasello, M. (2001). Perceiving intentions and learning words in the second year of life. <strong>In</strong><br />

M. Bowerman & S. Levinson (Eds.), Language Acquisition and Conceptual<br />

Development. Cambridge University Press.<br />

Troje, N. 2002. Decomposing biological motion: A framework for analysis and synthesis of<br />

human gait patterns. <strong>In</strong>: Journal of Vision, 2, pp. 371-387.<br />

Tversky, A. 1977. Features of similarity. Psychological Review, vol. 84, no 4, pp. 327-352.<br />

van der Zee, E. & U. Nikanne (eds.) 2000. Cognitive <strong>In</strong>terfaces. Oxford: Oxford University<br />

Press.<br />

van der Zee, E. 2000. Why we can talk about bulging barrels and spinning spirals: curvature<br />

representation in the lexical interface. <strong>In</strong>: van der Zee, E. & U. Nikanne (eds.) 2000.<br />

Cognitive <strong>In</strong>terfaces. Oxford University Press, Oxford.<br />

van der Zee et al. 2004. Motion encoding in language: Reference frame use during the<br />

employment of spatial locatives in a motion context, conference talk.<br />

Zacks, J. & B. Tversky 2001. Event Structure in Perception and Conception. <strong>In</strong>:<br />

Psychological Bulletin, vol. 127, nr. 1, pp. 3-21.<br />

Zlatev, J. & P. Yangklang 2004. A third way to travel: The place of Thai in motion event<br />

typology. <strong>In</strong> Stromqvist, S. & L. Verhoeven (eds.).<br />

Dictionaries and corpora<br />

Andreichin, L et al. 2004. Bulgarski tulkoven rechnik (Dictionary of the Bulgarian<br />

Language.) Fourth Edition. Sofia: Nauka i Izkustvo.<br />

Atanasova, T. et al. 1983. Bulgarian-English Dictionary. Second Photoype Edition. Sofia:<br />

Nauka i Izkustvo<br />

Atanasova, T. et al. 1985. English-Bulgarian Dictionary. Third Edition. Sofia: Bulgarian<br />

Academy of Sciences<br />

Cambridge Learner’s Dictionary (http://dictionary.cambridge.org/)<br />

Encarta Dictionary (http://encarta.msn.com/encnet/features/dictionary/dictionaryhome.aspx)<br />

Merriam-Webster Online Dictionary (http://www.m-w.com/)<br />

Oxford Dictionary Online (http://dictionary.oed.com/)<br />

The Oslo Corpus of Tagged Norwegian Texts - bokmål and nynorsk parts<br />

(http://www.tekstlab.uio.no/norsk/bokmaal/english.html)<br />

The Bokmål Dictionary (http://www.dokpro.uio.no/ordboksoek.html)<br />

40


Appendix 1<br />

This appendix contains still images of the 29 motion scenes used in the analysis and the<br />

motion verbs (and their frequencies) used for each scene in each of the target languages. Nonmotion<br />

verbs which are not taken into account in the cluster analysis are not present.<br />

Running scenes (scenes 1 - 9)<br />

1 Chimpanzee running<br />

2 Koala running<br />

3 Dog running fast<br />

English: 9 run, 1 lollop, 1 gambol, 1 chase<br />

Norwegian: 14 løpe/springe (run)<br />

Bulgarian: 2 ticham (run), 10 bjagam (run),<br />

1 turcha (run spec.), 1 prepuskam<br />

(run spec.)<br />

4 Dog running in circles<br />

English: 8 run, 2 bound, 1 gallop, 1 hop<br />

Norwegian: 11 løpe/springe (run), 1 sprette<br />

(jump), 1 hoppe (hop), 1 galoppere<br />

(run spec.), 1 lunte (run spec.), 1<br />

bevege seg (move)<br />

Bulgarian: 4 ticham (run), 6 bjagam (run),1<br />

turcha (run spec.), 1 pripkam (run<br />

spec.), 1 prepuskam (run spec.), 2<br />

podskacham (hop), 1 obikaljam<br />

(circle)<br />

5 Dog running on treadmill<br />

English: 9 run, 1 lope, 2 sprint<br />

Norwegian: 14 løpe/springe (run), 1 sprinte<br />

(run spec.)<br />

Bulgarian: 7 ticham (run), 3 bjagam (run), 3<br />

prepuskam (run spec.), 1 gonja<br />

(chase), 1 presledvam (pursue)<br />

6 Lizard running<br />

English: 10 run, 1 race<br />

Norwegian: 15 løpe/springe (run)<br />

Bulgarian: 5 ticham (run), 1 bjagam (run),<br />

2 turcha (run spec.), 2 obikaljam<br />

(circle), 1 vurtja se (turn)<br />

English: 8 run, 1 jog, 1 trot, 1 walk<br />

Norwegian: 12 løpe/springe (run), 2 jogge<br />

(run spec.), 1 gå (walk/ go)<br />

Bulgarian: 6 ticham (run), 5 bjagam (run), 1<br />

pripkam (run spec.), 1 podtichvam<br />

(run spec.)<br />

English: 7 run, 1 lope, 1 trot, 1 scurry, 1<br />

move<br />

Norwegian: 13 løpe/springe (run), 1 pile<br />

(run spec.), 1 ile (run spec.), 1<br />

bevege seg (move)<br />

Bulgarian: 7 ticham (run), 7 bjagam (run), 1<br />

turcha (run spec.), 1 presledvam<br />

(pursue)<br />

41


7 Lizard running on hind legs<br />

8 Man running on the spot<br />

9 Woman running<br />

English: 8 run, 1 sprint<br />

Norwegian: 13 løpe/springe (run)<br />

Bulgarian: 7 ticham (run), 3 bjagam (run), 1<br />

pritichvam (run spec.)<br />

English: 8 run, 4 jog<br />

Norwegian: 5 løpe/springe (run), 11 jogge<br />

(run spec.)<br />

Bulgarian: 7 ticham (run), 5 bjagam (run)<br />

English: 10 run, 2 jog<br />

Norwegian: 12 løpe/springe (run), 4 spurte<br />

(run spec.)<br />

Bulgarian: 9 ticham (run) ,4 bjagam (run), 1<br />

vturvam se (run spec.)<br />

Walking scenes (scenes 10 – 18)<br />

10 Woman walking<br />

11 Woman walking backwards<br />

12 Chimpanzee walking<br />

English: 11 walk, 1 chancé<br />

Norwegian: 14 gå (walk/ go), 1 spasere<br />

(walk spec.), 1 marsjere (walk<br />

spec.)<br />

Bulgarian: 11 hodja (walk), 5 vurvja (walk)<br />

English: 11 walk<br />

Norwegian: 15 gå (walk/ go), 1 spasere<br />

(walk spec.)<br />

Bulgarian: 10 hodja (walk), 4 vurvja (walk),<br />

1 dviza se (move), 1 vrushtam se<br />

(return)<br />

English: 1 lope, 9 walk 1 step, 1 lumber<br />

Norwegian: 13 gå (walk/ go), 1 spasere<br />

(walk spec.), 2 rusle (walk spec.)<br />

Bulgarian: 8 hodja (walk), 1 vurvja (walk),<br />

3 razhozdam se (walk spec.), 2<br />

dviza se (move), 1 otivam (head<br />

for)<br />

42


13 Long-legged bird walking<br />

14 Crocodile walking<br />

15 Monkeys walking round a tree<br />

English: 6 walk, 5 strut, 1 stalk<br />

Norwegian: 9 gå (walk/ go), 1 spasere (walk<br />

spec.), 4 spankulere (walk spec.), 1<br />

sprade (walk spec.), 1 vagge (walk<br />

spec.)<br />

Bulgarian: 5 hodja (walk), 1 vurvja (walk),<br />

3 kracha (walk spec.), 1 pristupvam<br />

(walk spec.), 3 razhozdam se (walk<br />

spec.), 1 dviza se (move)<br />

16 Tiger walking<br />

English: 3 walk, 1 step, 6 waddle, 1 paddle,<br />

1 stroll<br />

Norwegian: 8 gå (walk/ go), 1 lunte (walk<br />

spec.), 1 luske (walk spec.), 1 stavre<br />

(walk spec.), 4 vagge (walk spec.),<br />

1 vralte (walk spec.)<br />

Bulgarian: 5 hodja (walk), 1 vurvja (walk),<br />

1 pristupvam (walk spec.), 1<br />

shljapam (walk spec.), 1 lazja,1<br />

tutrja se (crawl spec.), 3 dviza se<br />

(move), 1 otivam (head for)<br />

17 Koala walking<br />

English: 2 walk, 1 chase, 1 follow, 5 circle<br />

Norwegian: 2 løpe/springe (run), 12 gå<br />

(walk/ go)<br />

Bulgarian: 1 gonja (chase), 1 sledvam<br />

(follow), 7 obikaljam (circle), 2<br />

vurtja se (turn)<br />

18 Chameleon walking on twig<br />

English: 8 walk,1 pad, 1 slope, 1 prowl, 1<br />

stalk<br />

Norwegian: 12 gå (walk/ go), 2 lunte (walk<br />

spes.), 1 spankulere (walk spes.), 1<br />

tusle (walk spes.)<br />

Bulgarian: 9 hodja (walk), 2 vurvja (walk),<br />

3 razhozdam se (walk spec.), 1<br />

dviza se (move), 1 minavam (pass)<br />

English: 1 lope, 10 walk, 1 stroll<br />

Norwegian: 1 trave (run spec.), 13 gå (walk/<br />

go), 1 spasere (walk spec.), 1 lunte<br />

(walk spec.)<br />

Bulgarian: 1 pritichvam (run spec.), 8 hodja<br />

(walk), 2 vurvj (walk), 1 krach<br />

(walk spec.), 1 dviza se (move), 1<br />

zaputvam se (make for)<br />

English: 2 move, 1 make one’s way, 8 walk,<br />

1 crawl<br />

Norwegian: 5 gå (walk/ go), 1 spasere (walk<br />

spec.), 1 luske (walk spec.), 1 snike<br />

seg (walk spec.), 1 krabbe (crawl), 1<br />

klatre (clamber)<br />

Bulgarian: 9 hodja (walk), 2 vurvja (walk),<br />

1 kracha (walk spec.), 1<br />

promukvamse (walk spec.), 1 lazja<br />

(crawl), 1 pulzja (crawl), 1 dviza se<br />

(move)<br />

43


Crawling scenes (scenes 19 - 26)<br />

19 Baby crawling<br />

20 Woman crawling<br />

21 Man crawling on his stomach<br />

English: 12 crawl<br />

Norwegian: 16 krabbe (crawl)<br />

Bulgarian: 8 lazja (crawl), 7 pulzja (crawl),<br />

1 presledvam (pursue)<br />

22 Caterpillar crawling<br />

English: 1 move, 1 walk, 10 crawl<br />

Norwegian: 16 krabbe (crawl)<br />

Bulgarian: 3 hodja (walk) ,5 lazja (crawl), 7<br />

pulzja (crawl)<br />

23 Beetle crawling on twig<br />

English: 1 move, 9 crawl, 1 creep<br />

Norwegian: 1 snike seg (walk spec.), 1<br />

krabbe (crawl), 7 krype (creep), 7 åle<br />

seg (wriggle)<br />

Bulgarian: 1 lazja (crawl), 13 pulzja<br />

(crawl), 1 dviza se (move)<br />

24 A slow tortoise<br />

English: 1 move, 1 make one’s way, 9<br />

crawl, 1 creep<br />

Norwegian: 3 krabbe (crawl), 3 krype<br />

(creep), 2 kravle/kråle (crawl), 6 åle<br />

seg (wriggle), 2 bukte seg (wind/<br />

wriggle)<br />

Bulgarian:<br />

English: 1 scurry, 1 climb ,1 move, 1 walk,<br />

6 crawl, 2 creep<br />

Norwegian: 8 gå (walk/ go), 2 krabbe<br />

(crawl), 4 krype (creep), 1<br />

kravle/kråle (crawl), 1 klatre<br />

(clamber)<br />

Bulgarian: 2 hodja (walk), 2 vurvja (walk),<br />

4 lazja (crawl), 3 pulzja (crawl), 1<br />

dviza se (move)<br />

English: 1 walk, 1 mosey, 10 crawl<br />

Norwegian: 4 gå (walk/ go), 1 stabbe (walk<br />

spec.), 4 krabbe (crawl), 7 krype<br />

(creep)<br />

Bulgarian: 3 hodja (walk), 1 vurvja (walk),<br />

2 lazja (crawl), 4 pulzja (crawl), 1<br />

mukna se (crawl spc.), 4 dviza se<br />

(move)<br />

44


25 Snake crawling<br />

26 Snake sidewinding<br />

English: 2 crawl, 9 slither, 1 slide<br />

Norwegian: 4 krype (’creep’), 1 kravle/kråle<br />

(’crawl’), 7 åle seg (’wriggle’), 2<br />

bukte seg (’wind/ wriggle’), 1<br />

slange seg (’move in a snake-like<br />

fashion’)<br />

Bulgarian: 2 promukvam se (‘sneak’), 12<br />

pulzja (‘crawl’), 1 dviza se<br />

(‘move’)<br />

English: 2 move, 1 crawl, 3 slither, 4<br />

sidewind, 1 slide, 1 ripple<br />

Norwegian: 1 flyte (‘flow’),1 krype,1<br />

kravle/kråle (’crawl’), 8 åle seg<br />

(’wriggle’),4 bukte seg (’wind/<br />

wriggle’),1 slange seg (’move in a<br />

snake-like fashion’)<br />

Bulgarian: 10 pulzja (‘crawl’), 1 pripluzvam<br />

se (‘slither’), 1 kachvam se<br />

(‘ascend’), 1 izvivam se (‘wind’), 2<br />

dviza se (’move’)<br />

Climbing scenes (scenes 27 – 29)<br />

27 Koala climbing a tree<br />

28 Koala climbing a tree in small hops<br />

29 Sloth climbing down a tree<br />

English: 11 climb, 1 crawl<br />

Norwegian: 15 klatre (‘clamber’)<br />

Bulgarian: 10 katerja se(‘clamber’), 5<br />

kachvam se (‘climb’)<br />

English: 9 climb, 3 hop<br />

Norwegian: 2 hoppe (hop’), 14 klatre<br />

(‘clamber’)<br />

Bulgarian: 14 katerja se (’clamber’)<br />

English: 8 climb, 2 descend, 1 move, 1 walk<br />

Norwegian: 1 gå (‘go’), 15 klatre<br />

(‘clamber’)<br />

Bulgarian: 1 hodja (‘walk’), 1 lazja<br />

(‘crawl’), 1 pulzja (‘crawl’), 7<br />

slizam (‘go down’), 4 spuskam se<br />

(‘descend’)<br />

45


Appendix 2<br />

Dendrograms showing the clustering of the 29 analysed motion scenes<br />

Fig. 1a.<br />

Jaccard tree for Bulgarian<br />

46


Fig. 1b.<br />

Multiset tree for Bulgarian<br />

47


Fig. 2a.<br />

Jaccard tree for English<br />

48


Fig. 2b.<br />

Multiset tree for English<br />

49


Fig. 3a.<br />

Jaccard tree for Norwegian<br />

50


Fig. 3b.<br />

Multiset tree for Norwegian<br />

51


Appendix 3<br />

Formal description of the multiset distance measure<br />

We employed a multiset distance measure formally defined in the following manner: Given a<br />

set of verbs V and two verb lists l 1 and l 2 , we define the multiset distance between l 1 and l 2 to<br />

be<br />

d _ multiset(<br />

l , l<br />

1<br />

2<br />

) = 1−<br />

∑<br />

∑<br />

υ∈V<br />

υ∈V<br />

min( n<br />

max( n<br />

υ,<br />

l<br />

1<br />

υ,<br />

l<br />

1<br />

, n<br />

, n<br />

υ,<br />

l<br />

2<br />

υ,<br />

l<br />

2<br />

)<br />

)<br />

(1)<br />

where n υ,l is the number of occurrences of the verb υ in the verb list l. (The numerator is really<br />

the cardinality of the multiset intersection of the verb lists, and the denominator is really the<br />

union.)<br />

To highlight the difference between the multiset distance and the Jaccard distance, the<br />

Jaccard distance between l 1 and l 2 can be formulated as<br />

d _<br />

jaccard(<br />

l , l<br />

1<br />

2<br />

) = 1−<br />

∑<br />

∑<br />

υ∈V<br />

υ∈V<br />

min( p<br />

max( p<br />

υ,<br />

l<br />

1<br />

υ,<br />

l<br />

1<br />

, p<br />

, p<br />

υ,<br />

l<br />

2<br />

υ,<br />

l<br />

2<br />

)<br />

)<br />

(2)<br />

where p υ,l is 1 if υ is present in l and 0 otherwise. (The numerator is really the cardinality of<br />

the set intersection of the verb lists, and the denominator is really the union.)<br />

Simpson’s diversity index (D) for a given verb list was calculated with the formula<br />

(http://www.wku.edu/~smithch/biogeog/SIMP1949.htm)<br />

∑ n<br />

,<br />

× ( n<br />

,<br />

−1)<br />

υ∈<br />

υ l υ<br />

( l)<br />

=<br />

N × ( N −1)<br />

D<br />

V l<br />

(3)<br />

where n υ,l is the number of occurrences of the verb υ in l, and N is the length of the given verb<br />

list.<br />

52

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!