Perspective in signed discourse: the privileged status of the signer’s locus and gaze

Abstract In gesture studies character viewpoint and observer viewpoint (McNeill 1992) characterize co-speech gestures depending on whether the gesturer’s hand and body imitate a referent’s hand and body or the hand represents a referent in its entirety. In sign languages, handling handshapes and entity handshapes are used in depicting predicates. Narratives in Danish Sign Language (DTS) elicited to make signers describe an event from either the agent’s or the patient’s perspective demonstrate that discourse perspective is expressed by which referent, the agent or the patient, the signers represent at their own locus. This is reflected in the orientation and movement direction of the manual articulator, not by the type of representation in the articulator. Signers may also imitate the gaze direction of the referent represented at their locus or have eye contact with the addressees. When they represent a referent by their own locus and simultaneously have eye contact with the addressee, the construction mixes referent perspective and narrator perspective. This description accords with an understanding of linguistic perspective as grounded in bodily perspective within a physical scene (Sweetser 2012) and relates the deictic and attitudinal means for expressing perspective in sign languages to the way perspective is expressed in spoken languages.


Introduction1
McNeill (1992) introduced a distinction between two types of viewpoint in co-speech gesture, character viewpoint and observer viewpoint.He described the distinction in terms of the type of representation in the gesture and the speaker's involvement in the story: in character viewpoint gestures (C-VPT) the speaker's hand(s) and body represent the referent's hand(s) and body and the narrator is felt to be "inside the story" (1992: 119), in observer viewpoint (O-VPT) the speaker's hand represents the referent as a whole, the speaker's body is excluded from the gesture space, and "the narrator keeps some distance from the story" (1992: 119).
McNeill characterized dual viewpoint gestures as gestures that "present a scene simultaneously from two viewpoints " (1992: 122).The two viewpoints may be a C-VPT and an O-VPT or two C-VPTs representing two different characters.In a number of studies, Parrill (2009Parrill ( , 2010Parrill ( , 2011Parrill ( , 2012) ) followed up on McNeill's initial suggestion of the two types of viewpoint, investigating especially gestures combining two viewpoints simultaneously.She examined the relationship of the two types of viewpoint to features of the linguistic message such as transitivity and information structure, but she explicitly excluded gaze directions from her studies, a feature that is central to the expression of perspective in signing.
In signed discourse, signers may also represent referents physically by means of their hands and their body such that their hand or their body is seen as a substitute for the referent (cf., among others, Kegl 1985, Engberg-Pedersen 1993, Engberg-Pedersen 1995, Liddell 2000, Liddell 2003, Dudis 2004 on body partitioning, and 2011).The term classifier has since Supalla (1982, but see Engberg-Pedersen 1993, Cogill-Koez 2000b, Liddell 2003, Schembri 2003) been used of manual articulator(s) with specific forms used to represent referents in certain predicate signs.Following Cormier et al. (2012) I shall call them depicting handshapes.Two types are especially relevant here, handshapes representing the hand(s) of an agent handling something, so-called handling handshapes (cf. the signer's right hand in Ill.1), and handshapes representing entities in their entirety, entity handshapes (cf. the signer's left hand in Ill.1).In sign languages such handshapes are more or less conventionalized with specific forms matching specific meanings (e.g., Engberg-Pedersen 1993, 2010, Liddell 2003).Besides special handshapes in predicates, signers use so-called loci to represent referents.Loci are manifested in the orientation and the movement direction of the hands in manual signs, which may indicate the direction of the location of a real or imagined referent (for similar uses of space in spoken language discourse, see Stukenbrock 2014).Depicting handshapes are the sign equivalents of the gesture handshapes underlying McNeill's classification C-VPT (corresponding to handshapes representing the hands of an agent) and O-VPT (corresponding to handshapes representing an entity in its entirety).McNeill also talks about the narrator either being "inside the story" or keeping "some distance from the story", but it is not clear what the articulatory gestural markers of this difference is.
Ill. 1.A signer's description of how somebody approaches a ruffian who is holding a club in his right hand (initial position).
Acknowledging the origin of parts of signing in gesture, sign linguists have tried to integrate the insights from work on co-speech gesture into work on viewpoint -or perspective -in signing (especially, Perniss 2007a, Perniss 2007b, Cormier et al. 2012; for an overview, see also Stec 2012).Perniss (2007aPerniss ( , 2007b) ) introduces a distinction between two types of what she calls event space projections, a life-sized space surrounding the signer with the signer as part of the event and a model-sized space in front of the signer with the signer outside the event (for a similar distinction in terms of surrogates and tokens, see Liddell 2003, and for further development of this analysis, Dudis 2011).In Perniss' analysis, the life-sized space corresponds to the use of handling handshapes, the model-sized space to the use of entity handshapes in what she terms prototypical alignments of event space projection and depicting handshapes.The life-sized space makes use of loci on the sagittal axis, the model-sized space of loci on the lateral axis, at least in cases of prototypical alignment of event space projection, depicting handshapes, and direction from the signer, according to Perniss.The prototypical alignment of a life-sized space with handling handshapes and use of the sagittal axis corresponds to McNeill's character viewpoint, the prototypical alignment of a modelsized space with entity handshapes and use of the lateral axis to observer viewpoint.But Perniss points out that there are many instances of non-prototypical alignment in signed discourse, for instance, a life-sized conceptualization of event-space with the use of entity handshapes (Ill. 1 would be a case in point).Such cases are presumably sign instances of McNeill's double-perspective gestures.
Theories in terms of event space projections (Perniss 2007a(Perniss , 2007b) ) and surrogates and tokens (Liddell 2003) tend to describe the representations by invisible projections in space rather than by what the manual and non-manual articulators do (see, however, Aarons & Morgan 2003, and to some extent Dudis 2007 andCormier et al. 2012).Furthermore, McNeill's and Perniss' classifications are limited to an analysis of representations and perspective at the level of the individual gesture in co-speech gesture or of the individual depicting construction in sign languages, the purpose is not to describe perspective at the clause or discourse level.The notion of dual viewpoint gesture or non-prototypically aligned representations do not do justice to the idea of perspective as a discourse notion manifested at the clause level: when more than one referent is represented, they are rarely on an equal footing (Janzen, O'Dea and Shaffer 2001, Janzen 2004, Janzen 2012, Dudis 2007) as suggested by the term dual viewpoint gesture.An analysis that limits itself to an enumeration of the different referent representations and describe their combinations as prototypical or non-prototypical alignment or single-or double-perspective does not explain why one referent representation is perceived as outranking others and therefore as embodying the perspective holder proper at the cost of other represented referents (cf.also Dudis' (2011) analysis of the signer's body as manifesting the vantage point from which a scene is depicted).
Based on narratives in Danish Sign Language (DTS) elicited to produce differences in perspective at the discourse and clause levels, I shall argue that in signed narratives, perspective depends on the orientation of the hands and the direction of their movement in relation to the signer's body and on the direction of the signer's gaze.The orientation of the hands and their movement direction indicate which referent is represented at the signer's locus, and the signer's gaze may either imitate the same referent's gaze direction or be used to indicate the signer as narrator through eye contact with the addressee.If signers' gaze direction imitates the gaze direction of a referent represented at their locus, the perspective is unambiguously with this referent, who may be an agent or a patient in, or an observer of, the event described.If signers have eye contact with the addressee and simultaneously represent a referent at their locus, signalled by the hands' orientation and movement direction, the perspective is mixed, consisting of both referent perspective and narrator perspective.Other articulators, i.e., the signers' facial expression, mouth movements, and their body posture, may add dramatic colour to the expression of a referent's perspective or pull the entire construction in the direction of the narrator perspective.

Perspective, quotation and constructed action
Perspective in the sense used in this paper is "the position from which a scene is viewed" (Langacker 1987: 117) or in Sweetser's words: "cognitive perspective starts with bodily viewpoint within a real physical Ground of experience" (2012: 1).Sweetser links visual viewpoint to proprioception and spatiomotor strategies for accessing and reaching objects.This view of perspective accords with the understanding of perspective in sign linguistics since sign languages conventionalize iconic means of representing an individual interacting with objects and with other individuals.
Sweetser points out that human beings cannot avoid being occupied with the actions of other humans in the immediate environment and with their interactions with the physical surroundings as seeing such actions activates some of the neural patterns needed to perform them.Human beings are thus constantly changing perspective on the scene in which they and others find themselves.But, as also pointed out by  , Ego is special among human beings: "when our mirror neurons fire in response to watching someone pick up an apple, the non-mirror ones do not fire; so we neither use our own muscles to grasp a phantom apple, nor hallucinate that we are picking up an apple" (Sweetser 2012: 2).The capacity to change perspective on the environment is reflected in spoken and signed discourse in changes in linguistically expressed perspective.
One obvious way that speakers and signers can change their perspective is by quoting others or by demonstrating what others did, do or may do in the future.Clark and Gerrig (1990) interpret direct quotations as a special type of demonstration: like the demonstration of a limp, direct quotations are non-serious actions performed as part of serious actions.For an understanding of what they mean by "non-serious", Clark and Gerrig quote Goffman (1974) who talks about "a given activity, one already meaningful in terms of some primary framework" being "transformed into something patterned on this activity but seen by the participants to be something quite else" (43-44).When individuals demonstrate a limp or quote the speech of somebody else, addressees do not understand the communicators as actually limping or saying and meaning what they ascribe to others.Demonstrations are furthermore, Clark and Gerrig claim, depictions in the sense that they are interpreted in part through direct perceptual experience in contrast to descriptions, which must be seen as standing for or denoting whatever is described, and indications, where communicators make addressees understand the intended communicative content by localizing it -or a token of it.But demonstrations, and in particular quotations, are selective depictions: when we demonstrate a limp or quote somebody else, we depict only some aspects of the source.In direct quotations, for instance, we typically ignore the source speaker's false starts and hesitations.Moreover, only the depictive aspects of a demonstration should be seen as the demonstration proper.Other aspects are supportive, e.g., a translation into the language of the current speech situation of something originally said in a different language.There may also be annotative aspects, which are the current speaker's simultaneous comments on the demonstrations, e.g., in the form of a smile or a sneer, and incidental aspects, which may, for instance, be the current speaker's voice quality, which differs from the source speaker's voice.
Traditionally, direct quotation is described as a rendition of what somebody said from the source speaker's perspective, especially as reflected in deictic terms and illocutionary markers.In (1) we understand the words after says as expressing the perspective of the referent of he.
(1) uh I think he says I fell asleep on the bench, sorry honey uh it was kind of embarrassing, a policeman had to wake me up and gave me a stern talking to uh it won't happen again The second deictic pronoun I is interpreted as referring to the source speaker, not to the current speaker, and the past event that 'he' is talking about is in the past of the source speech event, 'he' expresses his attitude by excusing himself and using a term of endearment, and 'he' makes the promise at the end; all of this should be ascribed to the source speaker, not the current speaker.In Clark and Gerrig's words, the current speaker demonstrates what ((s)he imagines that) the source speaker said.By contrast, in indirect quotation as in (2), the deictic expressions are switched to reflect the perspective of the current speaker, the illocutionary status of the source message is described in declarative sentences, and the attitudinal terms are left out or may be of the annotative type that should be seen as expressing the current speaker's attitude.
(2) so eh he's saying that he's eh been in an extended meeting or something trying to explain why he's so late Here the current speaker uses he, not I, about the source speaker -the source speaker is he from the current speaker's perspective; explain in the declarative sentence describes what the source speaker did.Still, however, the content of the quotation is attributed to the source speaker; only the source speaker, not the current speaker, is interpreted as lying to his wife.This gives a mixed perspective.
A third type -so-called free indirect style, between direct and indirect speech -has long been acknowledged (for a historical overview, see Coulmas 1986, Clark andGerrig 1990).It shares features of both direct and indirect speech as the deictic terms are chosen from the current speaker's perspective, Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM but attitudinal elements and illocutionary marking of questions, statements, and commands express the source speaker's perspective -or the perspective of the source thinker, as free indirect style is very often used to convey thoughts in literary texts (Banfield 1973).This type also expresses a mixed perspective.Evans (2012) points out that direct and indirect speech are ideals or canonical points that are not necessarily manifested in all the world's languages, but against which other types of constructions can be characterized.Many languages follow what he calls a second person magnetism, i.e., second person values from the current speech situation dominate over the source deictic values, which means that, in such languages, it is never possible to use direct quotation in an equivalent of Peter said about you: "She is always so generous"; it would be Peter said about you you are always so generous.First person may also have a "magnetic" function.By contrast, some languages resist complete adaptation to the current speaker's deictic centre, for instance, by keeping the verb tense or mood of the source speech event in spite of changes in pronoun use.
Sign languages permit quotations, but sign languages also have constructions where signers depict states or events that animate referents take part in from the perspective of one of the referents.Following Tannen's (1986) description of quoted speech as constructed dialogue, Metzger (1995) talked about constructed action in sign language.2It is tempting to see such sequences as demonstrations in Clark and Gerrig's (1990) terminology.There is, however, an important difference.As mentioned, Clark and Gerrig see demonstrations as a sort of depiction, and they contrast depictions and descriptions.Descriptions are formed from "articulated symbol systems, such as the discrete words and sentences of a language" (1990: 767), whereas depictions "are formed from …dense, or nonarticulated, symbol systems, such as the analogical representations of painting or sculpture " (1990: 767).Without going into the extensive debate about what is discrete and symbolic and what is analogical in sign languages (see, especially, Cogill-Koez 2000a, 2000b;Liddell 2003), I will just say that it is not possible to describe the constructions that are the theme of this paper as "nonarticulated" or analogues of sculpture or painting, but they may nevertheless be seen as at least partial selective depictions.

Means of expressing perspective in sign languages
All articulators used in signed discourse, not just the hands, may contribute to expressing a certain perspective on an event.3Research on co-speech gesture has mainly focused on the hands, which may imitate the imagined manual action of a referent (see, however, Stukenbrock 2014), but in signing, the body, the face, including the mouth, and the gaze may also imitate what a referent is imagined to do.When signers describe a ruffian knocking somebody on the head with a club, they may hold their hands the way one would hold a club above one's shoulder, close their eyes, strain the muscles of their face, bend their head forward to one side, and then raise it as if to get more force into the blow when their hands move forward in the "blow" (see Ill. 2).
2 Dudis (2011) rightly points out that the integration of eye gaze, head movement, facial expression, and the body with a depicting construction may occur in depictions of events without an animate participant (Dudis' example is depiction of lightening strikes).He claims that the function of the non-manual signals is to indicate that "the depiction is being tracked through conceived time" (Dudis 2011: 35). 3 In Engberg-Pedersen (1993, 1995), I suggested that we analyse the contribution of each articulator separately.Following Jakobson's (1971) analysis of deictic shifters, I used the terms shifted reference, shifted attribution of expressive elements, and shifted locus of pronouns, signals of attitude, and the signer's locus respectively, when they were meant to be interpreted from the perspective of a referent different from the narrator's or signer's perspective.For instance, when a signer used the first person pronoun in a quotation to refer to the source sender, or an unhappy facial expression to express the unhappiness of a discourse referent, I would talk about a shifted pronoun or shifted attribution of the expressive element.The term shifted reference is, however, used by some sign linguists in a different sense, namely to indicate that signers move or lean their body in the direction of a locus that represents a discourse referent.Then shifted is used in the sense of physical movement (e.g., Janzen, O'Dea and Shaffer 2001; for an example where a speaker of German literally changes his position in space in a case of constructed speech, see Stukenbrock 2014) Loci as representations or projections of referents are not articulators, but are reflected in especially the hands' orientation and movement direction in some signs.In DTS loci are chosen so as to represent semantic, especially spatial or temporal, and pragmatic relations between the referents.In the example in Ill. 1, the locus to the left of the signer represents the person approaching the ruffian.It is seen in the signer's facing and looking leftwards and in her left hand which she moves with an entity handshape from the left towards herself.In general, only salient or recurring referents are associated with a locus, and referents that need not be distinguished or are pragmatically associated with each other may be represented by the same locus.The number of loci at play at any given moment in the discourse is generally the number of salient referents (or groups of referents) minus one because one referent is usually represented by the signer's own locus (cf.also for Swedish Sign Language, Ahlgren and Bergman 1994).This lending of their own locus to a referent appears from the signers' head or body orientation as when the signer faces slightly left in Ill. 1.The direction reflects not only the person approaching the ruffian, but also the ruffian's position in relation to the other person.This can be described as imitative body representation; by indicating the referent's orientation to a different referent, the signer's body and head orientation signals that the signer's locus has been taken over by a referent.
In Ill. 1 it may be difficult to see that we need to distinguish the signer's body, imitating the ruffian's posture, from her locus, representing the ruffian in relation to the sign made by her left hand.In other cases it is much clearer that the signer's body posture is irrelevant for the representations.Ill. 3 shows the way a signer described a policeman approaching a man lying on a bench.Here the signer's body posture is neutral, but the signer's left hand representing the policeman makes a semicircle from a little outside her left shoulder to a position close in front of her sternum.The signer's body as an articulator is neutral in relation to the representation; her body posture does not imitate the referent's body posture as the referent in question was lying down.But her position in space is relevant to an understanding of the direction and movement of the hand in the manual sign; the relationship between the referents is represented by the manual articulator's movement direction in relation to the signer's locus.This is why we need to distinguish head and body posture (as the head movement forwards and then back in Ill.2) from body orientation, which reflects the relation to other referents.
Like the first person pronoun I in English, the signer's locus can be used of the signer or of a quoted sender in speech quotations.In fiction English I can also be used of other referents outside quotations.In signed discourse this type of lending one's own locus to a referent is quite common in descriptions of interactions between several referents, also in nonfiction.The signer's locus has one more function: it is used in the base form of many manual signs to represent the agent or actor.The base form of a verb like SEND-MAIL (Ill.4a) is identical to the form used in the equivalent of 'I sent the letter to someone', but in 'I got a letter' the hand orientation is reversed (Ill.4b).In constructions like the ones in Illustrations 1-3, an understanding of who is represented by the signer's locus must be deduced from the discourse context.
Ill. 3. A signer's description of a policeman approaching a man lying on a bench.The signer's locus represents the man on the bench who is being approached, but her body posture is neutral.Signers' gaze behaviour is also crucial to understanding how perspective is expressed in signed discourse in DTS (Engberg-Pedersen 2003).When signers have eye contact with their addressee(s), they take on their role as sender or narrator, except in one case to which I will return.Through eye contact with the addressee, they can keep track of the current speech situation and check the addressee's understanding.Signers also take on the role of sender/narrator when they use their gaze for referencetracking, i.e., looking briefly in the direction of a referent's locus.This happens often in the beginning of a sentence when the topic changes to a new referent.Configurational or locational gaze is seen when signers describe a complex static configuration or the relationship between two or more referents by representing them in space; it is as if they direct addressees' gaze to the representation itself.The final type of gaze that indicates the sender/narrator is signers' looking away in no particular direction at a major syntactic break or when they hesitate.
In contrast to the different types of gaze behaviour indicating the sender/narrator, imitative gaze consists of signers' imitation of a discourse referent's observed or imagined gaze behaviour.When the signer's locus represents a referent in a narrative, their gaze direction often, but not always, imitates that referent's gaze in the represented event (cf.Ill.1).
The signer's facial expression may imitate a referent's facial expression of a particular feeling in relation to a state or an event.In Ill. 2 the signer imitates the ruffian's supposedly strenuous expression, and in Ill. 1 the signer's face expresses a joyful expectation that should probably be attributed to the ruffian, not the narrator, at the prospect of the future assault.But it is not always easy to decide whether a facial expression should be attributed to one or the other.Dudis (2004) points out that in American Sign Language (ASL), the signer's mouth or lower face region may be "partitioned-off" to produce different information than the rest of the face.In Ill. 5 from DTS the signer describes a man sitting on a bench in a park, watching what is going on around him.The signer's locus represents the man, as can be seen from the fact that the back of her hand in the sign meaning 'watch' is oriented towards her body and the tips of the fingers towards the imagined park scenery, i.e., what the man sees.Her gaze is imitative of the man's gaze sweeping over the imagined scenery.The corners of her mouth are slightly drawn out, an expression that may be either attributed to the man on the bench or seen as a conventional mouth pattern in DTS that may accompany process verbs and indicates that the action is relaxed and carefree.In a different story the signer represents a boy washing a dog; she accompanies a prolonged rendition of the verb WASH (derived with a handling handshape) with repetitions of the mouth pattern of the spoken-language equivalent meaning 'wash'.It is unlikely that the boy would say the word again and again during the act, so the mouth pattern should not be seen as imitative, but be attributed to the narrator.In what follows, I will show that the choice between handling and entity handshapes (corresponding to McNeill's C-VPT and O-VPT, respectively) is irrelevant for the expression of perspective at the clause and discourse level.The decisive factors are the signer's gaze direction and who is represented by the signer's locus, i.e., the orientation of their body and the orientation and movement direction of their hand(s) in predicate signs.When the signers' gaze direction represents the narrator (eye contact with the addressee on predicates), but their locus is used of one of the referents as reflected in the orientation of their body or the orientation or movement direction of their hands in manual signs, the perspective is mixed as in noncanonical speech quotation constructions (Evans 2012), even when the construction does not occur in a quotation, but is constructed action.The signer's facial expression or body posture may add further colouration to a referent perspective.

Elicitation materials and participants
The cartoons used for eliciting the data of this study were constructed in a project with Peter Harder, The University of Copenhagen, to collect data from Danish, English, and DTS in 2004.They consist of booklets of 4-5 drawings of one of two types.In one series of booklets, called the A-stories (A for agent), the protagonist is seen in a series of situations leading up to an event where the protagonist interacts with a second individual in such a way that the protagonist has the active role.In the P-stories (P for patient), the protagonist is the patient referent in the last picture.There are four A-stories and four P-stories, and they occur in pairs such that the last picture of each set of A-and P-story is identical, for instance, a ruffian knocking down a man holding money in his hands.In the A-story, the ruffian is the protagonist throughout the story, in the P-story, the small man is the protagonist.Each participant saw either the A-story or the P-story from each set.Each participant saw altogether four stories, one A-story with human referents, one A-story with a human agent protagonist and an animal patient referent, one P-story with human referents, and one P-story with an animal protagonist and a human agent.
The two P-stories with human protagonists had a sequel, a single picture where the protagonist is seen returning home where he is met by his wife.The instruction to this picture, which was shown to the participants after they had told a story on the basis of the first four pictures, was (translated into English): "Here you see him when he comes home to his wife.She doesn't know what has happened.He tells her about what happened.Try to tell the story the way he would tell it to her." The participants were told to leaf through the booklets first and then tell the stories without looking at the pictures again.Five adults contributed to the data set and were videotaped when telling the stories.All five are native signers of DTS, they are aged between 35 and 60, and they happened all to be women.As it will appear, the signers told very similar stories with respect to the features focused here.As the purpose of the study was not to investigate the full range of possible variants in DTS, the data set from five signers was judged to be sufficient.

Analysis
The cartoons used to elicit the stories for this study were designed to test the difference between the linguistic treatment of agent protagonists versus patient protagonists.Each story includes a climax where the protagonist interacts with a human being or an animal as either the agent or the patient.Table 1 shows whether the signers described the crucial event in such a way that their own locus was used to represent the agent or the patient of the story in the predicate.Most signers used several constructions to describe the crucial event, however.In her narrative about the man who was knocked down and robbed by the ruffian (a P-story), Signer3 described what happened by means of the sequence seen in Ill. 6.The first still photo shows a depicting construction of the man holding his money (a handling handshape in her left hand) with an entity handshape of someone approaching the man ('While he was holding his money, someone approached him from the right').Next Signer3 signed MAN followed by a depicting construction with two handling handshapes ('It was a man who knocked him with a club') (the way signs and signed sentences are rendered in writing in this paper is explained in the Appendix).In the depicting construction with handling handshapes (Ill.6c) the signer's locus represents the agent.In the final part of the example, the signer used the sign STEAL in its base form ('He stole (the money)').That is, this signer described the crucial event, the knocking down, not from the protagonist's perspective, but from the agent's perspective (cf.Table 1).But she described the ruffian approaching the man holding the money from the latter's perspective, and having introduced someone by means of the entity handshape in her right hand (Ill.6a), she identified this individual by means of a noun (MAN in Ill.6b) and then switched the perspective to this individual.
Table 1.The signers' choice of locus representation in the predicate about the crucial events in the narratives as seen in the orientation and/or movement direction of the hand(s).A = only agent, P = patient in at least one predicate.Accidentally, Signer5 was not asked to tell the P-story as if told to the man's wife (Human protagonist + human, speech quotation).

Human protagonist + animal
Human protagonist + human After having finished the story, Signer3 was asked to recount what the man said to his wife after returning home.In the speech quotation, she indicated the ruffian, i.e., the agent (MAN), and made a depicting construction with a handling handshape from the forward right upward direction towards her own head ('A man knocked me on my head') (Ill.7).Then she signed 1.p MONEY Handling:countingmoney ('I was counting my money'), and keeping her nondominant hand with the handling handshape from Handling:counting-money, she signed the first person pronoun (1.p) and once more the depicting construction with the handling handshape from the forward right upward direction ('while counting my money, I was knocked on my head'), after which she added STEAL POINT fru ('he stole it'), using the base form of the sign STEAL with an agent representation at her own locus followed by a point in the forward right upward direction to refer to the agent of both the knocking action and the stealing action.In the speech quotation (the man explaining the event to his wife), this signer thus used two identical depicting constructions where her own locus represented the protagonist, i.e., the patient in the event.In Table 1, A means that the signers used their locus only of the agent of the event, P means that at least once the signers used their locus to represent the patient in a verb form or the form of a depicting construction.The following forms are used for the classifications in Table 1: -A depicting construction with one or two handling handshapes about knocking someone down with a club (Ill. 2 and Ill.7b-c); this form is used both with an agent (two hands) and a patient (one hand) represented at the signer's locus in the narratives.-AWAKEN: a lexical verb derived with a handling handshape (handling by touching or pushing as in Ill.

Animal protagonist + human
8); in the narratives the verb is only used with a patient represented at the signer's locus, but it can be used also with an agent represented at the signer's locus.-ARREST: a lexical verb derived with one or two handling handshapes; in the narratives the verb is only used with an agent represented at the signer's locus, but it can also be used with a patient represented at the signer's locus.-WASH: a lexical verb derived with a handling handshape; in the narratives it is used with an agent and a patient represented at the signer's locus.-SHOOT 'shoot (with a gun)': a lexical verb derived with two handling handshapes, the verb cannot be used with a patient representation at the signer's locus since the handling iconically reflects handling a gun, not the patient.In the narratives the sign is only used with an agent representation at the signers' locus.-A depicting construction with an index hand following the trajectory of (in these narratives) a bullet; this form is used both in A-stories with the bullet's starting point represented at the signer's locus, and in a P-story with the target (the patient shot) represented at the signers locus (the tip of the index finger is brought into contact with the signer's neck in a description of how a cow is shot by a hunter, see Ill. 9).To these signs can be added constructions in other parts of the narratives like the one in Ill. 1, specifically the depicting construction on the signer's left hand in relation to her locus.What is described here is not one referent doing something to a different referent, but one referent approaching another referent represented by the signer's locus.In Ill. 1, the situation is described from the perspective of the referent being approached, i.e., the perspective of the referent represented by the signer's locus.The depicting construction is not used in any of the narratives with a movement of the hand away from the signer's body, but it is used in both A-stories and P-stories about the prospective agents being approached by their potential victims and about the victims being approached by the prospective agents.
Table 1 shows that all the signers used only A-forms and no P-form when the protagonist was the agent in the crucial event of the stories, and that they had a strong tendency to use a P-form when the protagonist was the patient of the crucial event, especially in the stories about a human protagonist interacting with another human referent, less so in the stories about an animal protagonist (a cow or a puppy) interacting with a human agent.The fact that there are at all A-forms in the P-stories is evidence that there is a strong tendency to make agents topics of sentences, a tendency which is also reflected in the fact that the base form of many verbs is identical to a form with the agent represented at the signer's locus.But the results in Table 1 indicate that the use of the signer's locus is the decisive factor in determining discourse perspective.
Janzen, O'Dea and Shaffer ( 2001) define passive in ASL as transitive clauses constructed such that (1) the signer presents the clause from the point of view of the patient rather than the agent, even though an agent is understood to be carrying out the action; and (2) the agent is demoted, which often means that the agent is not mentioned, even though there are instances of a more weakly demoted agent still present in the clausal structure.Critical morphological features are the direction of movement of the verb toward the signer and the eye gaze of the signer, which is associated with the patient rather than the agent of the action… (Janzen, O'Dea & Shaffer 2001: 283-284) Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM The authors here point to two of the same features that I have looked at in DTS: the orientation and movement of the manual articulator(s) in the verb sign or depicting construction and the signer's eye gaze.They also claim that "[i]n the active construction, the signer's shoulder leans slightly in the direction of the agent positioned in the signing space…" (Janzen, O'Dea and Shaffer 2001: 289).This feature is not seen in the narratives in DTS.Janzen, O'Dea and Shaffer find that in passives in ASL the patient is more accessible or given information than the agent.This, too, is not necessarily the case in DTS.A nominal referring to the agent may be the last preceding nominal as in example (3) (see Ill. 10), but the patient referent is overall more salient as it is the protagonist of the narrative (cf.also example 7 in Janzen, O'Dea and Shaffer 2001: 296).----------------------------------------------- Example (3) also shows that, contrary to what is the case in ASL according to Janzen, O'Dea and Shaffer (2001), the signer needs not change her body orientation.The signer here changes her body posture along the sagittal axis, not her orientation, as she imitates the agent in the forward movement of the blow, and the Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM patient in leaning back when signing Handling:move-|club|-to-own-head.Janzen, O'Dea and Shaffer find that in what they call prototypical passives in ASL, the articulator in the verb sign moves from a neutral or semantically empty direction towards the signer.This is also true of one example in the narratives from DTS, i.e., in a description of a man walking along and counting his money and then being hit; the aggressor is only identified afterwards.But in the descriptions of a man on a bench being woken by someone, the verb AWAKEN is made in two P-stories on the signers' left shoulder (cf.Ill.8), and then the leftward direction is used to represent the policeman.Notably, the sign AWAKEN is made by one signer with her right and by another with her left hand, so the choice of the left shoulder is not phonologically determined, but meant to signal the direction of the -up to that point in the narrative -not yet identified agent.In Ill.7b the starting point of the hand in the depicting construction meaning 'hit with a club' is to the right, which is the direction of the thief's locus.
Janzen, O'Dea and Shaffer ( 2001) are undoubtedly right in pointing to similar functions of the relevant constructions in ASL and passives in spoken languages: the perspective is with the patient, and the agent is demoted or unidentified.The reason why I hesitate to talk about passives in DTS is the similarity between the constructions that form the basis of the classifications as P in Table 1 and constructions such as the one in Ill. 1 with a depicting construction of someone approaching the holder of the perspective.The difference might be ascribed to transitivity; the verbs used for the classifications in Table 1 might be said to be transitive while the depicting construction in Ill. 1 describes an individual moving towards another individual, a movement that might be stopped at any point.Janzen, O'Dea and Shaffer do indeed describe their constructions as transitive constructions that have been detransitivised.But it has to be argued on independent grounds that transitivity (cf.Matthews 1981) is relevant to sign languages, and that there is a difference in transitivity between the depicting construction in Ill. 1 and the depicting construction in Ill.7b.A proper treatment of passive and transitivity in the present context is, however, not possible.
In seven of the eight different stories used to make the classifications in Table 1, the relevant sign or depicting construction is either made with a handling handshape or is a lexicalized version of such a sign.In these seven cases a human agent manipulates a human or animal patient in close encounter.In one story the manipulation is a shooting incident, and here the signers use SHOOT followed by a sign that denotes the consequence.SHOOT is a lexical sign made with two handling handshapes, and, as mentioned, it does not depict the handling of the patient, but of the gun.The two predicates showing the consequence of the shooting are FALL, i.e., a lexical version of a depicting construction with an entity handshape, and a depicting construction showing the bullet hitting the cow's neck by means of the signer's own neck (Ill.9).McNeill's classification of O-VPT and C-VPT gestures is based on the type of depiction (whether the speaker's hand represents the referent as a whole or whether the speaker's hand(s) and body represent the referent's hand(s) and body), but the examples with SHOOT and one individual approaching another individual (cf.Ill. 1) demonstrate that the choice of handshape type, handling or entity, does not determine the type of perspective, but only reflect the type of action represented.The action may involve close contact with manual manipulation (e.g., AWAKEN with a handling handshape) or it may take place without such contact (SHOOT with two handling handshapes, 'approach' with one entity handshape).AWAKEN and 'approach' are used in these stories only with patient referent perspective, and SHOOT can only be used with agent perspective in the sign.
If we now turn to the other features that may be used for referent imitation, we find many combinations from minimal imitation to full alignment of all features with the use of the signer's locus.Table 2 is a classification of all the other features in the twenty-four descriptions of the crucial event.The four rows Human + human, self-report represent the descriptions of what two of the protagonists told their wives in P-stories (quoted speech with the patient as the source speaker).As can be seen from Table 2, there are few descriptions with full alignment of all features to represent either a referent or the narrator (eye contact with the addressee except in the case of quoted speech, see below).This is partly due to the use of mixed perspective in signed discourse, partly to the fact that many features are ambiguous and can be interpreted as representing either a referent or the narrator.If the signers use the forward direction to represent the patient in a predicate, the sign token is identical to the base form of the sign (indicated as A/neutral in Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM Table 2, see Ill. 4a), as in Signer3's description of a boy washing a puppy (Ill.11).In Ill.11, the signer's face expresses pleasure, but is she smiling to represent the boy's feelings or as a comment to the story?Some signers more or less smile all through the stories, probably because of their harmless nature.Signer3's body orientation in Ill.11 is irrelevant; her orientation is aligned with the manual sign, but her body posture (the shoulders pulled up) expresses the boy's pleasure at performing the action.Finally, her gaze imitates the boy's looking at the puppy.
Table 2.All the means that may be used for representing a referent in the twenty-four descriptions of the crucial events in the stories.Accidentally, Signer5 was not asked to recount the protagonist's self-report to his wife in a P-story.The white cells indicate expressions that can be unambiguously ascribed to the narrator or one of the referents, either the agent (A) or the patient (P) of the event described.The grey cells represent ambiguous or neutral representations.The coding is explained in the Appendix.P/neutral in the column Body + head orientation is found in the stories of a man on a bench woken by a policeman.Since the man is not aware of the policeman's existence until he is awake, he is not turned in the direction of the policeman when the policeman touches him.Therefore, the signers' body (see Ill. 8) and head orientation is neutral when they make the sign AWAKEN in the P-stories of this episode and in quoting the man's self report to his wife, but their orientation may equally well be described as imitative.

Manual
Table 2 shows that the columns Face, Mouth, Body and/or head orientation, and Body posture tend to align with the manual sign in the sense that they are either neutral or conventional, or if they do represent a referent, the signer's locus in most cases also represents either the agent or the patient (the column Manual sign).There is one case of an A-value in the column Face and three cases of A-values in the column Body posture with A/neutral in the column Manual sign.In all the other thirty-four cells with an unambiguous A-or P-value in the columns Face, Mouth, Body and/or head orientation, and Body posture, this value is aligned with the value of the column Manual sign.Since the value in the column Manual sign reflects who the signer's locus is used of, one of the referents or the narrator, this distribution indicates the privileged status of the signer's locus.
With the value in Manual sign as the point of departure, we see that twenty-one out of thirty cells in the rows with a P-value in Manual sign have P-values.That is, 70% (21/30) of the values align with a manual sign indicating the patient perspective by means of the signer's locus.This fact and the fact that there are only P-values in the column Manual sign when the patient of the event is the protagonist of the stories are again strong indicators of the importance of the signer's locus in expressing perspective in DTS.Twenty-eight out of forty-eight cells in the rows with an A-value in Manual sign have A-values.Here 58% (28/48) of the values are aligned with the A-perspective.Of the cells in the rows with A/neutral in the Manual-sign column, as few as five out of forty-two cells have unambiguous A-values, i.e., 12% (5/42).This distribution is a symptom of the fact that the signer's locus is strongly associated with the agent in the formation of lexical signs.The agent perspective is close to a neutral perspective or the narrator's perspective.
If we look specifically at the gaze values in relation to the values in the column Manual sign, we find that in twelve out of thirteen cases where the gaze direction is imitative (an A or a P in the Gaze column), the column Manual sign also has an A or a P (I shall return to the only deviant case below).In two more cases the signer's gaze direction partially imitates the agent (A followed by narrator) when the locus use is A/neutral, that is, these cases may also be analysed as alignment of gaze direction with an agent representation at the signer's locus.There are four cases of narrator gaze in rows with an A or a P in the column Manual sign, and these constitute mixed perspective in the clauses.The gaze direction is thus less used for expressing the perspective of a referent than the signer's locus in a manual sign, which is again evidence of the privileged status of the signer's locus in expressing a referent's perspective and point to the crucial role of the signer's gaze in expressing narrator perspective.In three of the four cases with mixed perspective (A or P value in Manual sign, narrator in Gaze), the signers use their own locus to represent the agent, and they use eye contact during more than one sign; their discourse is so to speak in the narrator perspective, but they mix into the story an element of referent perspective, i.e., the representation of the patient referent by a locus different from their own.In only one case a signer uses her locus to represent the patient when having eye contact with the addressee, namely Signer3 saying that the protagonist cow was hit in the neck by the hunter's shot in a P-story.Signer3 first introduces the hunter, describes his shooting and then the consequence, that the cow was hit (Ill.9) and fell.Interestingly, the example occurs in a story where the interaction between the protagonist and the second referent is distant.The signer does not have the possibility of describing the shooting itself from the patient's perspective as the hands in SHOOT cannot change direction.She can only describe the consequence, being hit, from the patient's perspective.But she cannot describe where the cow was hit without using her own neck (cf.Dudis 2011).It is thus not clear which factor influences the construction in Ill.9, that the protagonist is the patient of the event, or that the signer shows details of the shooting on her own body.Possibly the two factors go together in the depicting construction in Ill.9.
One cell in Table 2 is marked eye contact rather than narrator.The reason is that the description occurs in quoted speech where the protagonist explains his mishap to his wife.One signer here uses eye contact with the real addressee when quoting the protagonist's self-report to his wife.I have seen other cases of signers using the real addressee as a substitute for addressees of the original speech event in quotations in DTS, but this phenomenon needs further investigation.
In conclusion, most of the articulators that may be used to represent a referent either align with the representation at the signer's locus manifested in the manual signs, or they are neutral, ambiguous, or conventional (the latter about mouth movements).The signer's gaze direction is unique in being able to indicate the narrator unambiguously.Combinations of the signer's locus representing a referent (A or P) with other articulators representing the same referent more or less strongly can be seen as more or less dramatic cases of constructed action.By contrast, cases where the signer's locus is used of a referent and the signer has eye contact with the addressee can be seen as cases of mixed perspective, except in quotations where the addressee may be used as a stand-in for the source addressee.In the data used for this study, mixed perspective is seen mainly when the signer's locus is used of the agent.When the signer's locus is used of the patient, there is a stronger trend towards alignment of other features with the manual sign than when it is used of the agent.Stukenbrock (2014) describes a German-speaking man's recount of an everyday incident in a reality TV-show.The man quotes what he overheard another man say to his wife in a hospital room, imitating the man's dialect, gaze direction, and body posture in a combination which, apart from the dialect imitation, is very similar to what can be seen in signed discourse.Immediately afterwards, the man in the reality show demonstrated his own reaction of disbelief by standing "still in an upright static, position" (Stukenbrock 2014: 85) and directing his gaze towards the imagined position of his girlfriend with a meaningful facial expression while at the same time he described his body posture and engaged his girlfriend by looking at her, in Stukenbrock's translation into English "I am standing, I am looking at my girlfriend" (2014: 85).
Here is possibly a difference between spoken and signed discourse: the combination of bodily enactment and simultaneous verbal description is hardly possible in signed discourse.But eye contact with the current addressee during constructed speech in signed discourse may serve the same function, namely that of indicating the individual in constructed interaction with whom the narrator empathizes, in the narratives in DTS it is the man recounting his accident to his wife, and in the German story it is the man who overheard the other man's comment to his wife (i.e., himself).Both the signer's eye contact with the current addressee and the German man's verbal description of his bodily reaction and of his looking at his girlfriend point to the current narrator in the here-and-now of the communication situation rather than the past referents and have the function of involving the current addressee in the evaluation of the narrated events.

Conclusion
The narratives used for this study demonstrate how perspective in signed discourse can be related to perspective in spoken languages.In both modalities, the means for expressing perspective have deictic and attitudinal value.In spoken languages, we distinguish canonical direct speech from canonical indirect speech by whether the deictic terms are anchored in the source communication situation (direct speech) or in the current communication situation (indirect speech).Indirect speech represents mixed perspective as the deictic terms are chosen from the current speaker's perspective and the contents should be ascribed to the source speaker.Non-canonical forms may deviate from the canonical forms by various types of mixes of deictic anchor points and ascription of attitudinal expressions to either the source or the current speaker.
In sign languages signers may also use deictic and attitudinal means to represent and imitate referents or to indicate their role as current sender; the means are their own locus, their facial expression and body posture, and their gaze direction.The iconicity of the constructions furthermore relates the means used for expressing perspective in sign languages closely to the neural basis of perspective pointed out by Sweetser (2012): signers see others act, and they demonstrate (Clark and Gerrig 1990) such actions in iconic ways, controlled by restrictions on signing in general and by the conventions of specific sign languages.For instance, the fact that the signers do not lie down when describing the man on the bench being woken by a policeman may be seen as an example of Clark and Gerrig's (1990) incidental aspect of a demonstration, but also as a consequence of the requirement that signers, when signing, stay in one place with one main body posture and orientation adapted to the current communication situation.
McNeill's classification of gestures as expressing character viewpoint and observer viewpoint based on the type of representation in the hands obscures the similarity between the means used for expressing perspective in spoken and signed languages.Moreover, it obscures the similarity between examples such as someone moving towards somebody else (Ill. 1) and somebody knocking somebody else on their head (Ill.7b-c); in both cases one referent, the individual approached or knocked down, is represented by the signer's locus as holder of the perspective even though one construction makes use of an entity handshape (cf.observer viewpoint) and the other of a handling handshape (cf.character viewpoint).
A further advantage of the analysis presented here is that it prioritises the means of imitation.The use of the signer's locus to represent one of the referents makes this referent the holder of the perspective, with one caveat: the base form of many predicate signs resembles the form with the agent represented at the signer's locus and the patient represented in the forward direction.This is in itself a strong indication of how predicate signs reflect the neural mirroring of actions iconically.Evidence that use of the signer's locus as reflected in the manual signs is the strongest indicator of referent perspective is found in the facts that predicate signs with the patient referent at the signer's locus only occurs in P stories, and that there is a stronger trend towards aligning other features with the manual sign when the signer's locus is used unambiguously to represent either the agent or the patient.
Eye contact with addressees is anchored in the current communication situation; it permits signers to keep track of the addressees' understanding.But signers often break off eye contact to imitate a referent's gaze direction.When they do so, their gaze direction imitates the referent represented by their locus, and in such cases, there is full alignment with this referent as the holder of the perspective.Mixed perspective is found when the signers represent one referent by their locus, but have eye contact with the addressee.The only exception is that signers may use the current addressee as a substitute for the source addressee when quoting others.In this case, eye contact with the current addressee may be a way of pointing to the referent of the two people interacting in the source communication situation, namely the source speaker.
Finally, the other means of imitation, especially the signers' facial expression and their body posture, may align with the referent representation expressed by the signer's locus and add dramatic colouration.
. To avoid confusion, I shall here talk about the articulators as representing or indicating either a discourse referent or the sender/narrator.Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM Ill. 2. A signer's description of how a ruffian knocks another man on the head with a club (initial position).
Ill. 4a.Initial and final position of SEND-MAIL, neutral sign or from signer locus to non-first person (with permission from The Danish Sign Language Dictionary, www.tegnsprog.dk)Ill.4b.Initial and final position of SEND-MAIL from non-first person to signer locus (with permission from The Danish Sign Language Dictionary, www.tegnsprog.dk).Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM

Ill. 5 .
A signer's description of a man watching the scenery in a park at his leisure (initial position).Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM Signer3's description of the man counting his money being approached by another man and knocked on his head.(a) is a depicting construction describing the man holding money (left hand) at the signer's locus and somebody approaching him (right hand).(b) is the sign MAN (initial position).(c) is the initial position of the depicting construction 'knock with a club' (both hands).(d) is the sign STEAL (initial position).Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM Signer3's description of what the man who was mugged told his wife.(a) is the sign MAN (initial position), (b) and (c) are the initial and final position of a depicting construction describing the knocking on somebody's head with the patient represented at the signer's locus.The signer who told this story explains the use of her left hand in (b) and (c) as "a startled reaction gesture on the part of the man.The man didn't have a clue that this was going to happen, but it happened anyway" (my translation).As the only movement of the signer's left hand is the movement that brings it into position, the reaction represents backgrounded information in relation to the blow (Engberg-Pedersen 2011).
Brought to you by | Copenhagen University Library (Det Kongelige Bibliotek) Authenticated Download Date | 9/24/15 2:51 PM Ill. 8.The sign AWAKEN (in the signer's right hand) made with the patient represented at the signer's locus.Ill.9. A depicting construction showing somebody being hit in the neck (final position).