+RZ 'R'HPRQVWUDWLYHV&RGH'LVWDQFH" $OIRQV0DHV &KULVWGH5RRLM Communication and Cognition Research Group Humanities Faculty Tilburg University maes@uvt.nl $EVWUDFW The interpretation of demonstratives in any context somehow relies on the spatial configuration of human communication (speaker – addressee – objects to be referred to). But how exactly do space and distance determine a demonstrative speech act? In a controlled production experiment, we collected baseline data on how humans construct a demonstrative contrast in different spatial configurations and referring to objects with a different degree of saliency. We asked 7 to 9 year old children to find salient and less salient differences between two nearly identical visuals presented to them, in one of two conditions, in both of which the two visuals had a different location and orientation with respect to the speaker. In the QHDU condition, two visuals were located within the arm reach of the child, one central in front of the child, the other left or right a bit further away; in the IDUcondition, the frontal visual had the same location, the left or right visual was placed a little bit further away, just outside the arm reach of the speaker. The results show that distance is a strong determinant for demonstrative behaviour, and that space slightly interacts with the saliency of objects. Congruent demonstratives (WKLV QHDU vs. WKDW IDU) were only fairly standard in the IDU condition, whereas in the QHDU condition the differences were much more referred to by using two proximate forms. The tendency to use congruent demonstratives slightly decreased in the case of salient objects. Also, salient objects more often took the first position in a demonstrative pair. Finally, distance appears to have a systematic effect on pointing behavior: in the IDUcondition participants pointed less systematically to the further away object than in the QHDUcondition (3). And, to top it all, a WKLV referent can even be more distant with respect to the deictic center than a WKDW referent Consider (4), in which the doctor uses WKLV to refer to an object which is further away from him than from the patient, and the patient uses WKDW to refer to an object which can hardly be more near to him, i.e., his own head. 6SDFHGLVWDQFHDQGWKHLQWHUSUHWDWLRQRI GHPRQVWUDWLYHV There are two claims about demonstratives in language that linguists around the world largely agree on: on the one hand, demonstratives somehow code aspects of the spatial constellation between speaker, addressee and reference objects (most typically near vs. far away from speaker or addressee), on the other hand, they never code literal or Euclidian space or distance (e.g. five vs. twenty inches away from speaker or addressee). The first claim is based on the assumption that all languages have adnominal or adverbial demonstrative variants which – at least in their basic exophoric use – somehow express differences in distance with respect to speaker, addressee or referent. In a survey of demonstrative systems in the recent world atlas of language structures, Diessel (2005) concludes that all languages seem to have at least a dual distance contrast, whereas the most complicated systems most probably have no more than a tripartite distance axis. The second contention is supported by the observation that demonstratives can be used to identify referents irrespective of the metrical distance with respect to the deictic center. Thus, the two objects referred to in (1) can both be centimeters or light years away from the speaker (e.g. two pens on the table vs. two stars in the sky). Demonstratives exhibit more degrees of freedom with respect to space and distance, however. They do not have to be in accordance with the relative distance between referents. The objects referred to in (1) may well have exactly the same distance with respect to the deictic center. Conversely, the same demonstrative term can be used to identify referents on a different distance with respect to the deictic center, as in (2) and (1) (2) (3) (4) Do you like WKLV one (KHUH or WKDW one (WKHUH ? Do you like WKLV one, or WKLV one? Do you like WKDW one, or WKDW one? Doctor: Is WKLV where it hurts (pushing with his thumb on the forehead of the patient)? Patient: Yes, WKDW is where it hurts. These examples suggest that context may make the use of demonstratives more exciting than what one would expect on the basis of the simple proximal-distal dichotomy (see e.g., Enfield, 2003). There is a simple reason for this capricious behavior of demonstrative terms. On the one hand, demonstratives are certainly rooted in the way in which humans perceive and experience physical space. Their spatial experiential basis is part of a large range of linguistic conceptualizations based on bodily space, ranging from the many 'metaphoric' extensions of a simple spatial proposition like LQ to much more complex metaphoric conceptualizations, like PDUULDJH LV D MRXUQH\ or D FDUHHU LV DQ DVFHQGLQJ SDWKZD\ (e.g., Deutscher, 2005; Gibbs, 2005). On the other hand, the ultimate communicative function of demonstratives is not aimed at expressing distance, let alone metrical distance. Rather, they direct attention and create joint attention in human communication, as is convincingly argued for in Diessel (2006). The study of exophoric as well as anaphoric or discourse deictic demonstratives can be seen as different 1 In the remainder of this article, I use WKLVand WKDWas a short cut for all SUR[LPDWH and GLVWDO forms respectively. attempts to get a grip on the conceptual extensions of the spatial root of demonstratives. Exophorically used demonstratives allow conversants to create and orchestrate attentional regions or engagement spaces in face-to-face communication (e.g., Enfield, 2003; Levinson, 2004). &URVVOLQJXLVWLF UHVHDUFK has discriminated a large number of perceptual-spatial parameters in languages all over the world which play a part in shaping a communicative situation by way of demonstratives. They include locational factors (e.g., distance, orientational axes, cardinal directions of speaker, addressee or referent) as well as spatial-perceptual characteristics of referents (e.g., visibility, motion, posture, (un)boundedness, preciseness) (see for a survey Imai, 2003). All these parameters can be considered experientially based conceptual extensions of space. A similar interpretation can be claimed for DQDSKRULF and GLVFRXUVH GHLFWLF XVHV of demonstratives, albeit the extensions from the spatial near-far root are more conceptual and metaphoric (Maes, 2007). Thus, in discourse reference theories, demonstrative variance has been related to different degrees of ’mental nearness’. For example, Ariel (1990) attributes a higher degree of accessibility to WKLV than to WKDW entities and for Gundel, Hedberg and Zacharski (1993), WKLV NPs are more ‘given’ than WKDW NPs2. Other authors attribute more saliency to WKLV entities than to WKDW entities, capturing the difference in terms like focus (Gerner, 2003; McCarthy, 1994; Sidner, 1983; Webber, 1991), markedness (Halliday & Hasan, 1976), figure-ground (Hanks, 1992), or deictic force (Kirsner, 1979). Other studies relate demonstrative variation to undeniable but at the same time hard to control socio-cognitive, relational and perspectival subtleties. For example, WKLV and WKDW are considered to express some form of association with the realm of speaker and listener respectively (e.g. Cheshire, 1996; Glover, 2000; Kamio, 2001; Laury, 1997; Maes, 1996; Marchello-Nizia, 2005). Furthermore, many instances of WKDW involve the perspective of the addressee. They express shared knowledge with the addressee, as in the case of reminder or recognitional WKDW (e.g., Cornish, 2001; Himmelmann, 1996; Maes, 1996), or mental distance between speaker and entity, as in modal or emotional WKDW (e.g., Cornish, 2001; Lakoff, 1974), or they even act as turn-construction devices in dialogue (Hayashi, 2004). In sum, it is reasonable to assume that the interpretation of demonstratives in any context ultimately relies on the fundamental spatial configuration of human communication, and it is equally reasonable to assume that no demonstrative ever expresses space or distance directly, i.e. without the mediation of a speaker who conceptualizes space on the basis of his or her perspective, attention and intentions. 6WXG\LQJWKHLQWHUSUHWDWLYHEDVLVRI GHPRQVWUDWLYHV This conclusion, however, does not explain how exactly space and distance determine the use of demonstratives, and how they combine with conceptual and metaphoric extensions of demonstratives. The methods used in the two research traditions lack control to enable valid conclusions. Existing methods used in cross-linguistic as well as corpus based discourse research are not able to find conceptual (attentional, social, relational, emotional etc.) sediments in exophoric demonstratives, or spatial relics in non-exophoric demonstratives. Cross-linguistic research largely relies on grammars or explicit elicitation. Grammars are rarely specific enough to register refined pragmatic associations. For example, even for a wellrecorded language like Dutch, the pragmatic associations as described in Kirsner (1979) or Maes (1996) are largely absent in Dutch grammars. More recently, demonstratives have been studied cross-linguistically using explicit elicitation, e.g. on the basis of field manual instructions (Imai, 2003; Wilkinson, 1999). This may be reliable to find the parameters encoded in demonstratives, but not to elicit more conceptual factors, such as the attitude of participants towards objects or each other. Moreover, expert judgments about demonstrative variance are extremely whimsical. Simply ask a few colleagues whether they prefer WKLV or WKDW in an average stretch of discourse, and you will hear different intuitions and solutions. Finally, explicit elicitation is not a safe basis for the study of the actual use of language, as it treats participants as confederates. In discourse studies, the functional interpretation of demonstrative variants is mostly supported by exegetic analyses of attested examples, often complemented by distributional data from corpora. But neither method offers indisputable evidence about the spatial sediment in non-spatial demonstratives. In this paper, we will discuss the set up of a controlled production experiment which is intended to collect baseline data for Dutch on how distance determines the use of demonstratives in contrastive situations. The set up enables us to draw conclusions about the independent role of distance, to analyse the relationship between distance, demonstrative variation and pointing behaviour, and to explore the interaction of space and conceptual factors (in particular object saliency) which may be believed to determine demonstrative behaviour. Finally, as the experimental method resulted in a large and reliable collection of spontaneously produced contrastive demonstratives, it can be used to develop follow up experiments with gradually more finegrained conceptual, spatial and interactional variables. &ROOHFWLQJEDVHOLQHGDWDRQFRQWUDVWLYH GHPRQVWUDWLYHVDSURGXFWLRQ H[SHULPHQW 2 A notable exception is the indefinite use of WKLV, which allows speakers to introduce new entities vividly (Prince, 1981). 6HWXSDQGJRDOV An important methodological criterion in developing the experiment was the collection of a VL]DEOH corpus of VSRQWDQHRXV QDWXUDO and FRPSDUDEOH data from QDwYH speakers XQDZDUH of what we were interested in. To that end we asked the participants to play the well-known game to ILQG WKH GLIIHUHQFHV (either salient or non-salient) in two nearly identical visuals presented to them. This task resulted in a spontaneous and frequent production of demonstratives, which in addition was very much comparable for all participants. In this first phase, we restricted the contrastive use of demonstratives to the basic axis between D GHLFWLF FHQWHU WKH VSHDNHU DQG REMHFWV. Therefore, we made sure that the addressee had the same spatial perspective as the speaker. Thus, we intended to exclude all effects caused by different spatial perspectives of speaker and addressee. Furthermore, we opted for children (between 7 and 9) as participants. In a pilot study, we discovered that children of this age were much more inclined to use contrastive demonstratives than adults, who used much more descriptive (as opposed to pointing/demonstrative) referential expressions. According to all developmental studies, children of this age are perfectly able to use demonstratives appropriately. Finally, we varied distance on the basis of a parameter with a high cognitive value: we located the visual either within or outside the arm reach of the speaker, so that the objects in the visuals were either graspable/touchable or not. The cognitive encoding of this distance parameter is convincingly shown in Kemmerer (1999). Moreover, a number of languages encode this distance parameter in that they use different demonstratives to point to referents within and outside the physical control of the speaker (Imai, 2003: 135 ff.). In sum, these conditions result in a natural communicative situation in which the use of demonstratives is controlled with respect to a variable of space (near – far) and attention (salient – non-salient). in Figure 1). Altogether, 23 differences were implemented in the three pairs, 14 of which were predefined as salient. 3DUWLFLSDQWV 42 participants took part in the experiment, all children between 7 and 9 years old (M=7,83, 23 boys, 19 girls). All children had Dutch as their native language. Pair 1 Pair 2 Pair 3 Figure 2. Three pairs of visuals in one of three presentation modes ,QVWUXPHQWDWLRQ DQG SURFHGXUH The visuals were located on a stand on the table. In the QHDU condition, one visual of each pair was located in front of the participant, on a 35 cm distance, the other in turn either left or right on a 45 cm distance. In the IDU condition, one visual had the same location (35 cm in front of the child), the other left or right on a 125 cm distance. That way, we varied the distance with respect to the speaker (absolute distance), but inevitably also the mutual distance between objects (relative distance). The temporal order of the three pairs was identical for all participants (see Figure 2). The position of the visuals (front vs. left/right) was varied as to guarantee that all visuals (and thus also all salient and non-salient differences) were presented equally often on all locations. 0HWKRG Participants were presented with three nonequidistant pairs of visuals in one of two between subject conditions: QHDU vs. IDU. Figure 1. Find the differences! One pair (pair 1) of visuals with 6 salient and 3 non-salient differences 0DWHULDOV Three sets of two nearly identical visuals were constructed. Two types of differences were implemented systematically. Differences are QRQVDOLHQW when the two versions of the object are equally normal with respect to our common knowledge, and hence none of the two versions is assumed to attract more attention or be more salient than the other (e.g. the rounds vs. stars in Figure 1). 6DOLHQW differences are based on the ILQG WKH PLVWDNHV format: one version is supposed to be uncommon and deviant with respect to common world knowledge, and hence may be expected to be more salient than the other (e.g. the bucket vs. the hat in the director's left hand Figure 3. Camera position in the QHDU (left picture) and IDU (right picture) condition. The experiment took place individually. The experimenter presented pair 1, asked the child to find the differences, and repeated the same procedure for pair 2 and 3. During the experiment, the experimenter (i.e. the second author) stood exactly behind the child (see Figure 3). That way, we ensured that speaker and addressee had the same perspective on the visuals. The experimenter acknowledged correct differences (RND\KPP\HV asked additional information when participants gave minimal answers, using a few standard cues (ZKDW":KDWH[DFWO\" and hinted at differences that could not be found ($QGWKH FRZ"$QGWKHFORWKHVRIWKHPDQ"(WF.). Each session was videotaped so that linguistic and visual aspects of demonstratives could be analysed properly. Each session was transcribed and divided in chunks, each chunk representing the communicative interaction about one difference. 5HVXOWV The 42 participants produced altogether 1076 chunks3. These chunks represent a large variety of referential strategies. Participants almost exclusively used a demonstrative strategy to refer to the differences. Only in a few cases participants did not use a demonstrative element spontaneously. Adverbial locative demonstratives (KLHU ’here’ vs. GDDU ’there’) were more frequent than pronominal demonstratives (GLWGH]H’this/these’ vs. GDWGLH ’that/those’), which were in turn more frequent than adnominal demonstratives (GLWGH]H 1 ’this/these N’ vs. 4 GDWGLH 1 ’that/those N’ ). Often, locative and pronominal demonstratives were combined. 6WDQGDUG FRQWUDVWLYH GHPRQVWUDWLYHV DQG VSDFH The largest part of the data are standard cases of demonstrative contrast (58,5%, n=629), which we – for analytical purposes – defined as having the following three characteristics: (i) they contain at least one demonstrative device referring to one object, and at least one other demonstrative device referring to the other objects; (ii) they are produced spontaneously, without an intervention of the experimenter; and (iii) they contain enough semantic content to enable successful identification without the help of multimodal aspects of reference (pointing, gaze, etc.). Examples are given in (5) to (7). (5) And I see that here [F] is a window, but there [L/R] is not one. 5 [pp 8; far – congruent] (6) Euh, that [F] has no window and that [L] does. [pp 26, near – double distal] (7) Here [L] is a window and here [F] not. [pp 42, near – double proximal] We consider these examples the best test bed for answering our main research question. Overall, in about half of these standard cases (52,9%), participants used the WKLVWKDW contrast congruently (WKLV=QHDU, WKDW=IDU). Incongruent pairs were rare (n=31, 4,9%), but not restricted to mistakes. In 16,5% of all cases, participants used two distal variants; two proximal variants were used in 25,6%. Table 1 shows that the distance variable 3 The experimenter took care that all differences (42 * 23 = 966) were found. But some differences were mentioned twice, and now and then, the children mentioned differences which were not there. 4 The demonstrative system in Dutch is quite similar to English: a two term demonstrative system (proximal vs. distal), with locative adverbs (KLHUvs. GDDU and identical adnominal and pronominal forms (GLWGH]Hvs. GDWGLH , the distribution of which is related to the grammatical gender and number of the following noun or object referred to (neuter singular: GLWvs. GDWmasculine and feminine singular and all plurals: GH]Hvs. GLH 5 The examples are semi-literal translations from Dutch. Interventions of the experimenter are italicized. All examples refer to the same object, representing a salient difference in visual pair 1 (see Figure 1), i.e., the dormer window in the circus tent. The codes [F] or [L/R] refer to the position of the visual, i.e. either in front or left/right of the speaker. influenced the proportions of these four classes drastically 2 (3) = 215, p < .0001). In the IDU condition, children used almost exclusively either congruent demonstratives (83,5%) or two distal variants (13,7%). The proportion of congruent cases in the QHDU condition was much lower (28,8%), even lower than the use of two proximal variants (43,6%). Distal variants were more frequent in the near than in the far condition (18,8% vs. 13,7%). Incongruent cases were exclusively found in the near condition (8,8% vs. 0%). Table 1. Number (percentage) of types of contrastive demonstratives (this-that, *this-that, this-this, that-that) related to condition (near vs. far) (n=629). 1HDU )DU 7RWDO This – That 101 (28,8) 232 (83,5) 333 (52,9) *This – That 31 (8,8) 0 (0) 31 (4,9) That – That 66 (18,8) 38 (13,7) 104 (16,5) This – This 153 (43,6) 8 (2,9) 161 (25,6) 351 278 629 (100) Total 6WDQGDUG FRQWUDVWLYH GHPRQVWUDWLYHV DQG REMHFW VDOLHQF\ The effect of object saliency can be measured in different ways. First, saliency does not alter the general proportions of congruent, incongruent, double proximal and double distal demonstratives 2(3) = 5,68, p = .13). Although not significant, salient differences tend to decrease the proportion of congruent demonstratives (WKLV=QHDU, WKDW=IDU): 58% of all non-salient references are congruent, vs. 49,6% of all salient references. This indicates that the normal distance mechanism is weakened under the influence of salient objects. One may equally expect that saliency decreases the gap between near and far in the case of congruent demonstratives, again because of the fact that saliency would weaken the effect of distance. As Table 2 shows, a small tendency can be observed, which is however not 2 VLJQLILFDQW (1) = 0,95, p = .33). Finally, it can be noted that most of the incongruent cases (18 vs. 11) were linked to salient objects. Table 2. Number (percentage) of congruent salient and non-salient demonstratives related to condition (near vs. far) (n=330). 1HDU )DU Congruent salient objects 58 (32,6) 120 (67,4) Congruent non-salient objects 42 (27,6) 110 (72,4) General 100 (30,3) 230 (69,9) In sum, saliency appears to have only a slight effect on the use of demonstrative variants. Apart from that, saliency is expected to decrease the natural preference to refer to the frontal object first. Overall, participants refer to the frontal object first in 68,5% of all cases. In the case of non-salient objects the percentage decreases to 66,6% (vs. 71,2 for salient objects), again a slight indication that saliency may more easily overrule the standard demonstrative pattern (nearest object = first reference = WKLV), but again not a significant 2 RQH (1) = 2,379, p = .12). A final indication for the importance of object saliency is this: the two versions of the non-salient objects acted equally often as first referent in a demonstrative contrast (51% vs. 49%, n=270), which was what we expected because each visual and thus each object version was presented equally often in frontal and lateral (l/r) position. For salient objects however, the salient version was much more frequently used as first referent (72,1% vs. 27,9%, n=351), which shows the effect of saliency on the construction of a demonstrative contrast. This saliency effect is distance dependent, as it is stronger in the QHDU condition than in the IDU condition. This suggests that saliency is becoming more important as distance is less compelling6. 1RQVWDQGDUG FRQWUDVWLYH GHPRQVWUDWLYHV The remaining demonstrative acts (n=447, 41,5%) somehow deviate from the standard contrastive scheme. A number of cases were left out of the analysis for a number of reasons, either, because the difference was not found spontaneously, or because objects were not recognized, as is illustrated in (8), or because children had problems to come up with an appropriate formulation of the difference. In all these cases, the demonstrative endeavor of the children resulted in a mix of deictic and anaphoric referential devices, interesting for a later analysis, but too complex and not enough comparable for current purposes. (8) I do not know what this [R] is, but it is there, and not here [F] [pp. 11, near] There are two non-standard demonstrative acts, which are used systematically and which underline the role of multimodal aspects of reference in the given experimental task. First, participants spontaneously often refer to one of the two objects only. Hence they do not meet the first and second characteristic – as in (9) – or all three characteristics – as in (10) – of our definition of a standard demonstrative contrast above. Second, participants refer to the two objects in a way that does not allow identification without the help of the multimodal information included in the pointing gesture, as in (11). (9) There [L] is a brown window. :KHUH" There [L]. 2.. [pp 10, near] (10)That [F] $QGZKDWLVWKHGLIIHUHQFH" Well, there [F] is such a thing and there [L] is not [pp 1, near] 6 A more fine grained analysis, including also the nonstandard cases, should shed more light on the role of saliency in relation to distance. It is clear that some predefined salient differences really worked well: participants noticed them relatively early in the course of the task, and they referred first to the deviant version much more often than to the normal version. Other predefined salient objects apparently were less salient as expected. This certainly has weakened the effect of saliency in this experiment. (11) Here this one [R] and here [F] not. :KDWGLG\RXVD\" Here [R] that window and here [F] not. [pp 19, near] As it was the purpose of the study to collect as much cases of comparable standard verbal contrastive demonstratives as possible, these minimal strategies were discouraged by interventions of the experimenter, as is shown in (9) to (11). These interventions resulted in participants gradually aligning their contributions more to the 'standard' needs of the experimenter. But in fact, these minimal strategies fitted perfectly in the experimental setting. By using these strategies, children exploited the situation of sharing their perspective with the experimenter, who was standing right behind them. Many children (n=17, 40,5%) opted for one or both of these minimal strategies as their first demonstrative act. They only identified one of the two objects (n=15, 35.7%), and/or referred to the two objects by using empty demonstratives only (n=8, 19%). 3RLQWLQJDQGGLVWDQFHThe strong tendency of children to minimize their linguistic effort implies the use of pointing gestures. Examples like (10) and (11) require accompanying pointing gestures to result in adequate referential acts. But also in standard cases of demonstrative contrast, where pointing is not really needed to identify objects, pointing is almost omnipresent. Table 3 shows the pointing behaviour of participants on the basis of a partial analysis of the data (i.e. the first pair of visuals of the first 16 participants including 151 chunks). In most cases, participants point spontaneously to the two contrastive objects with one hand. Pointing with two hands, simultaneously (n=12) or nonsimultaneously (n=5) is rare. And pointing to the frontal object only, either spontaneously or after intervention of the experimenter, is restricted almost exclusively to the IDU condition. Small as the sample may be, the results suggest that putting objects outside the arm reach results in a less abundant use of pointing gestures. Furthermore, the overall high number of pointing gestures suggests that pointing behaviour is not affected by the four types of demonstrative contrasts or vice versa. But a larger scale analysis of the data will have to (dis)confirm these suggestions and trends. Table 3. Number (percentage) of different types of pointing related to condition (near vs. far) in a subset of the data, 16 participants, pair one of visuals (n=151). 1HDU Q 6SRQWDQHRXVSRLQWLQJ )DU Q at both objects with same hand 58 (76,3) 50 (66,7) at both objects with two hands 14 (18,4) 3(4) 0 (0) 11 (14,7) 1 (1,3) 1 (1,3) 1 (1,3) 10 (13,3) at the frontal object only at the left/right object only 3RLQWLQJDIWHULQWHUYHQWLRQ to the frontal object only 1RSRLQWLQJ Analyses of individual participants make clear that children can use different strategies to set the scene in the experiment. For most of the QHDU children gaze and pointing makes clear that their engagement space includes the region of the two visuals together, which explains their frequent use of double proximal demonstratives. Some of the IDU children also extend their engagement space to encompass the far away visual by leaning a bit left or right in the direction of the further away visual and using wide gestures. Other IDU children focus on the frontal visual, and display subdued pointing behaviour with respect to the far away visual. &RQFOXVLRQ The data presented in Table 1 show a pervasive influence of peri-personal distance in using contrastive demonstratives. Thus far, it is unclear whether peripersonal distance or relative distance of objects is the most determining factor. But the results enable us to draw a number of provisional conclusions. First, the distribution in the QHDU condition suggests that, in a standard contrastive situation, with two objects located near to the speaker and near to each other, and with two clear suggestions for a contrastive treatment of the two objects, i.e. the difference in position (frontal vs. left/right), and the difference in distance with respect to the speaker (35 vs. 45 cm.) the standard demonstrative contrast (this vs. that) is less standard than what we would have expected. This may be strange as intuition tells us to use the standard this-that format, even for equidistant objects. Second, the distribution in the IDU condition makes clear that the standard and congruent contrastive scheme is triggered by a larger distance (from the speaker and from the other object). It is striking that in this condition not a single incongruent case of demonstrative contrast is produced. In sum, if we only consider the use of demonstrative variants, an insignificant 100 cm difference in distance seems to cause an identical task space (i.e. the two visual fields relevant in the task) to be conceived of as one region (in the near condition) or as a dichotomic space (in the far condition). However, the inclusion of the multimodal behaviour of speakers as well as a more refined analysis of the saliency of objects referred to may well make this conclusion less finished. 5HIHUHQFHV Ariel, M. (1990). $FFHVVLQJ 1RXQ3KUDVH $QWHFHGHQWV London/New York: Routledge. Cheshire, J. (1996). That jacksprat: An interactional perspective on English that. -RXUQDO RI SUDJPDWLFV 369-393. Cornish, F. (2001). ’Modal’ that as determiner and pronoun: the primacy of the cognitive-interactive dimension. (QJOLVK /DQJXDJH DQG /LQJXLVWLFV (2), 297-315. Deutscher, G. (2005). 7KH XQIROGLQJ RI ODQJXDJH $Q (YROXWLRQDU\ 7RXU RI 0DQNLQG V *UHDWHVW ,QYHQWLRQ: Metropolitan Books. Diessel, H. (2005). Distance contrasts in demonstratives. In B. Comrie (Ed.), :RUOG$WODVRI/DQJXDJH6WUXFWXUHV (pp. 170-173). Oxford: Oxford University Press. Diessel, H. (2006). Demonstratives, joint attention, and the evolution of grammar. &RJQLWLYH /LQJXLVWLFV (4), 443-462. Enfield, N. J. (2003). Demonstratives in Space and Interaction: Data from Lao Speakers and Implications for Semantic Analysis. /DQJXDJH MRXUQDO RI WKH /LQJXLVWLF6RFLHW\RI$PHULFD 82-117. Gerner, M. (2003). Demonstratives, articles and topic markers in the Yi group. -RXUQDO RI SUDJPDWLFV 947-998. Gibbs, R. W. (2005). Embodiment in metaphorical imagination. In D. Pecher (Ed.), *URXQGLQJ &RJQLWLRQ WKH UROH RI SHUFHSWLRQ DQG DFWLRQ LQ PHPRU\ ODQJXDJH DQG WKLQNLQJ (pp. 65-92). Cambridge: Cambridge University Press. Glover, K. D. (2000). Proximal and distal deixis in negotiation talk. -RXUQDORISUDJPDWLFV 915-926. Gundel, J. K., Hedberg, N., & Zacharski, R. (1993). Cognitive status and the form of referring expressions in discourse. /DQJXDJH(2), 274-307. Halliday, M. A. K., & Hasan, R. (1976). Lexical cohesion, &RKHVLRQLQ(QJOLVK (pp. 274-292). London: Longman. Hanks, W. F. (1992). The indexical ground of deictic reference. In C. Goodwin (Ed.), 5HWKLQNLQJ &RQWH[W ODQJXDJH DV DQ LQWHUDFWLYH SKHQRPHQRQ (Vol. 43-76). Cambridge: Cambridge University Press. Hayashi, M. (2004). Projection and grammar: notes on the ’action-projecting’ use of the distal demonstrative are in Japanese. -RXUQDORISUDJPDWLFV 1337-1374. Himmelmann, N. P. (1996). Demonstratives in narrative discourse: a taxonomy of universal uses. In B. Fox (Ed.), 6WXGLHV LQ $QDSKRUD (pp. 205-254). Amsterdam/Philadelphia: John Benjamins Publishing Company. Imai, S. (2003). 6SDWLDO 'HL[LV Unpublished PhD, State University of New York, Buffalo. Kamio, A. (2001). English generic we, you, and they: an analysis in terms of territory of information. -RXUQDO RI SUDJPDWLFV 1111-1124. Kemmerer, D. (1999). ’’Near’’ and ’’far’’ in language and perception. &RJQLWLRQ 35-63. Kirsner, R. S. (1979). Deixis in discourse: an exploratory quantitative study of the modern Dutch demonstrative adjectives. In T. Givon (Ed.), 6\QWD[ DQG VHPDQWLFV (Vol. 12, pp. 355-375). New York: Academic Press, Inc. Lakoff, R. (1974). Remarks on This and That. &/6 345-356. Laury, R. (1997). 'HPRQVWUDWLYHV LQ LQWHUDFWLRQ 7KH HPHUJHQFH RI D GHILQLWH DUWLFOH LQ )LQQLVK. Amsterdam/Philadelphia: John Benjamins Publishing Company. Levinson, S. C. (2004). Deixis and pragmatics. In G. Ward (Ed.), 7KH+DQGERRNRI3UDJPDWLFV (pp. 97-121). Oxford: Blackwell. Maes, A. (1996). 1RPLQDO DQDSKRUV 0DUNHGQHVV DQG WKH &RKHUHQFHRI'LVFRXUVH Leuven: Peeters. Maes, A. (2007). Spatial and conceptual demonstratives. In L. Lundquist (Ed.), &RPSDULQJ $QDSKRUV %HWZHHQ 6HQWHQFHV7H[WVDQG/DQJXDJHV (34 ed.). Copenhagen: Samfundslitteratur Press. Marchello-Nizia, C. (2005). Deixis and subjectivity: the semantics of demonstratives in Old French (9th-12th century). -RXUQDORISUDJPDWLFV 43-68. McCarthy, M. (1994). It, This and that. In M. Coulthard (Ed.), $GYDQFHV LQ ZULWWHQ GLVFRXUVH DQDO\VLV (pp. 266275). New York: Routledge. Sidner, C. L. (1983). Focusing in the Comprehension of Definite Anaphora. In R. C. Berwick (Ed.), &RPSXWDWLRQDO PRGHOV RI GLVFRXUVH (pp. 267-330). Cambridge-Massachusetts: The MIT press. Webber, B. L. (1991). Structure and ostension in the interpretation of discourse deixis. /DQJXDJH DQG &RJQLWLYH3URFHVVHV(2), 107-135. Wilkinson, D. (1999). 0DQXDOIRUWKH)LHOG6HVVLRQ. Nijmegen: Max Planck Institute for Psycholinguistics.