Sign Languages: spinning and unraveling the past, present and future. TISLR9, forty five papers and three posters from the
9th. Theoretical Issues in Sign Language Research Conference, Florianopolis, Brazil, December 2006. (2008) R. M. de
Quadros (ed.). Editora Arara Azul. Petrópolis/RJ. Brazil. http://www.editora-arara-azul.com.br/EstudosSurdos.php.
Deixis, Anaphora and Highly Iconic Structures: Cross-
linguistic Evidence on American (ASL), French (LSF) and
Italian (LIS) Signed Languages1
Elena Antinoro Pizzuto1 , Paolo Rossini1,2 , Marie-Anne Sallandre3, Erin Wilkinson4
1Istituto di Scienze e Tecnologie della Cognizione, Consiglio Nazionale delle Ricerche, Roma
2Istituto Statale Sordi, Roma
3 UFR de Sciences du Langage, Université Paris 8 & UMR 7023, CNRS, Paris
4Department of Linguistics, University of New Mexico, Albuquerque, USA
This paper explores typological, presumably modality-specific features affecting deixis and
anaphora in signed languages (hereafter SL). Simplifying greatly for the sake of the present
exposition, we define deictic-anaphoric structures as text cohesion devices which allow speakers or
signers to introduce referents in discourse (deixis) and, subsequently, to refer back to them at later
instants in time (anaphora) (Lyons, 1977; Lombardi Vallauri , for a recent overview).
We focus on two major classes of dectic-anaphoric reference devices that have been
described in SL. The first, hereafter referred to as the ‘standard’ class has been rather extensively
investigated, and is realized through manual and visual indexes which establish marked positions in
space, often called “loci”, where referents can be symbolically assigned (for overviews highlighting
typological uniformity across SL with respect to this basic mechanism see, among others, Cuxac,
2000; Liddell, 2003; Lillo-Martin & Klima, 1990; McBurney, 2002; Rathmann & Mathur, 2002;
1 The present paper is a revised version of the presentation given by Erin Wilkinson and Marie-Anne Sallandre at the
TISLR 2006 meeting in Florianopolis, Brasil (Wilkinson, Rossini, Sallandre & Pizzuto, 2006). Each author contributed
equally, in different ways (and signed/written languages) to the work described in this paper. We gratefully
acknowledge partial financial support from different agencies, institutions, ongoing projects: the Fulbright Mason
Perkins Fund (2000-2001), for supporting Erin Wilkinson’s early research at the Istituto di Scienze e Tecnologie della
Cognizione of the Consiglio Nazionale delle Ricerche in Rome; the Italian and French National Research Councils
(CNR-CNRS Project “Language, its formal properties and cognition: what can be learned from signed languages”); the
Istituto Statale Sordi di Roma and the “Progetti Felicità” Funds (Italian project “Writing LIS and Sign Writing”); the
French Project ACI Cognitique “LS-COLIN” 2000-2002 (CNRS UMR 7023, IRIT, LIMSI). We are most grateful to all
the ASL, LIS and LSF signers who participated in the study, and to all the deaf and hearing colleagues at our
institutions for insightful discussions on the topics we explored. All the images of signers included in this paper are
reproduced with permission.
Pizzuto, 2004/to appear). Thus, to introduce a referent in discourse, a signer can produce a
standard (or “frozen”) manual sign2 for the referent, and deictically mark it in the signing space via
a manual and/or a gaze pointing, and/or also via morphological alteration of the sign’s place of
articulation (which may or may not be accompanied by visual indexes), thereby establishing a
position in space (or ‘locus’) for the referent symbolized. Anaphoric reference is then made re-
indexing roughly the same point in space via visual or manual pointings (see 1.1 for more details
and one illustrative example).
However, as noted elsewhere (2007), deictic-anaphoric reference is also produced, in SL, via
complex manual and nonmanual units which are neither pointing signs, nor are they listable as
standard signs. These units exhibit highly iconic features and are marked by specific eye-gaze
patterns which distinguish them from standard signs. Throughout this paper we will refer to these
units as Highly Iconic Structures (HIS) or ‘Transfers’, after Cuxac (1985; 1996; 2000). Aside from
specific eye-gaze patterns, HIS may be composed of: (1) manual forms encoding perceptually
salient features of the referents/ referential relations, most frequently described in the SL literature
with different terms such as “classifiers”, “productive morphemes”, “polysynthetic” or
“polycomponential” signs; (2) marked facial expressions, and/or directional modifications of head,
shoulders, whole trunk -- usually described in the literature as “role-shifting devices” but also with
other terms (see section 1.1 below for more discussion). As we will illustrate shortly hereafter,
different subtypes of HIS can be combined among themselves, or with standard signs, to
simultaneously encode information on two (or even more) referents, hence allowing the
specification of deictic-anaphoric reference in a multilinear fashion that appears to be unique of the
Although the manual and nonmanual devices mentioned above have been described from
different perspectives, and with different terminologies, there is little or no information on their use
in deictic-anaphoric reference operations, and/or on how they interact with the ‘standard’ devices
mentioned above. Our study provides relevant evidence on this topic through a comparative
examination of short narrative texts produced in three SL: American (ASL), French (LSF) and
Italian (LIS). ASL and LSF are known to be historically related, while LIS has no certain
relationship with either ASL or LSF3. The sample of language data we scrutinize thus allows us to
2 Standard signs are defined here, as in most SL literature, as those signs that are commonly listed in SL dictionaries,
and are often described as constituting ‘the frozen lexicon’. They are distinguished from the ‘productive’ highly iconic
forms described shortly hereafter.
3 It must be recalled that, due to the absence of appropriate written documentation of SL and to the complex language
transmission and variation phenomena proper of SL and of different communities of SL users (Cuxac, 2000; Fusellier-
Souza, 2004; 2006), it is still very difficult to ascertain the historical links among SL using the same criteria that are
applied in spoken/written language research. Thus, although there are no documented historical relationship between
LIS, on one hand, and ASL/LSF on the other hand, we cannot exclude that, at some point in past or recent history, these
SL have been in contact more than we are at present able to ascertain.
evaluate, albeit in part, the influence of language relations upon the phenomena investigated. If we
were to find similarities only between ASL and LSF, this could be taken as evidence that the
similarities identified can be attributed to the historical links between these two SL. If, however, we
were to find that similar patterns hold across the three SL investigated, this evidence would suggest
that similarities are grounded in more general features linked to the visual-gestural modality of
It is useful to clarify our theoretical framework and provide illustrative examples of the
structures mentioned above.
1.1. Theoretical framework and illustrative examples of the structures we focus upon
The present study stems from, and brings together theoretical and empirical work independently
developed, within our research groups, on SL morphology and on deixis and anaphora in SL as
compared to spoken or verbal languages (Pizzuto, 1978; 2007; Pizzuto, Giuranna & Gambino,
1990; Wilkinson, 2002), and on the crucial relevance that iconicity has in shaping SL lexicon,
grammar, discourse (Cuxac, 1985; 1996; 2000; Cuxac & Sallandre, 2007; Fusellier-Souza, 2004;
2006; Russo, 2004; Russo, Giuranna & Pizzuto, 2001; Sallandre, 2003; 2007). An appropriate
discussion of these theoretical foundations cannot be conducted here. We will only sketchily
outline some of the major points, especially with respect to the HIS mentioned above. Before doing
so, it is useful to consider briefly the ‘standard’ mechanism for deictic-anaphoric reference,
illustrated in Example (1) below with fragments taken from a LIS text reporting an ordinary event4.
4 From the ‘Necklace theft’ corpus (Fabbretti, 1997).
1c 1d 1e
(1) ‘a colleague’ [....] ‘he, the colleague, he...’
The still videoclip in 1a shows a standard nominal sign meaning ‘colleague’ through which
the signer deictically introduced in discourse this referent: the manual sign is dislocated at a
marked position at the signer’s right, and accompanied by a gaze pointing in the same direction.
Stills 1c-1e show how, at a later point in discourse, the signer anaphorically referred back to the
same referent first via a manual index directed towards the same position in space previously
marked (1c), then by a second instance of the sign ‘colleague’, this time articulated in an unmarked
position in neutral space (1d, compare with 1a-1b), and then by a second index (meaning ‘he, the
This ‘standard’ mechanism for deictic-anaphoric reference in signed discourse interacts
extensively with morphological modifications that different classes of verbs can undertake, altering
their locations in space, which we will not consider here (Cuxac, 2000; Pizzuto, 2007). The point
we wish to stress is that structures such as the above, described since early research on ASL (e.g.
Friedman, 1975; Klima & Bellugi, 1979; Wilbur, 1979), appears to be very similar across several
different SL of the world, and it seems plausible to assume that it is universal or nearly universal
(McBurney, 2002; Rathmann & Mathur, 2002; Pizzuto, 2007).
We turn now to briefly illustrate HIS and the theoretical model within which they are framed.
Based on extensive analyses of LSF discourse and grammar in a cross-linguistic framework, Cuxac
(1985; 1996; 2000) has proposed that all SL are grounded upon, and exploit a basic capacity that
signers have in iconicizing their perceptual/practical experience of the physical world. One of the
effects of this iconisation process is to endow SL with an additional semiotic dimension compared
to verbal languages. In SL, unlike in verbal languages, there are two ways of signifying: either by
‘telling and showing’, thereby producing HIS or ‘Transfers’ that are unique of the signed modality,
or by ‘telling without showing’, using the standard lexicon and pointings, and producing structures
that are more comparable to those found in verbal languages.
These two ways of signifying mirror two different intents a signer can deliberately choose for
articulating his/her discourse: an illustrative and a non-illustrative one. The operations signers
perform when choosing an illustrative intent (and the resulting structures they produce) are defined
“Transfers”, and conceived as traces of cognitive operations whereby signers transfer their
conceptualization of the real world into the four-dimensional world of signed discourse (the three
dimensions of space plus the dimension of time).
The manual components of these complex structures are called ‘proforms’. ‘Proforms’
correspond to what, in most SL literature are defined as ‘classifiers’ (or with other terms previously
mentioned - see Schembri, 2003 and the collected papers in Emmorey, for overviews). The
difference between ‘proforms’ and ‘classifiers’, however, is not purely terminological but rather
substantial: it is grounded on a linguistic model that attributes to iconicity a crucial, formal role in
shaping SL discourse and grammar, on theoretical analyses showing the inadequacies of a
‘classifier-analysis’ for the elements under discussion, and on a detailed analysis of the distinctive
formal and articulatory features that characterize the HIS within which proforms are embedded,
most notably specific eye-gaze patterns (Cuxac, 1996; 2000). In contrast, analyses of these forms in
terms of ‘classifiers’ focus primarily on the manual components, disregarding or underestimating
the role that gaze patterns play in their specification.
Referring to Cuxac (1985; 2000) in this paper we distinguish three main types of Transfers
(but see Sallandre, 2003 for a much more detailed classification):
1. ‘Transfers of Form and size’ (TF) describe objects or persons according to their size or form
(no process or role involved). The object is described by means of proforms. Gaze is oriented
towards the hands, and facial expression specifies the form.
2. ‘Transfers of Situation’ (TS) involve the movement of an object or character (the agent,
specified by the dominant hand) relative to a stable locative point of reference (specified by the
non-dominant hand). The situation is shown as if the scene were observed from a distance. The
signer keeps his/her distance relative to what s/he is conveying. Gaze is oriented towards the
dominant hand and facial expression specifies the agent.
3. ‘Transfers of Person’ (TP) involve a role (agent or patient) and a process. The signer ‘becomes’
the entity to which s/he refers reproducing, in the course of the utterance, one or more actions
carried out or sustained by this entity. The entities referred to are usually humans or animals but
can also be inanimates.
The third type of Transfers, TP, have been defined in the literature with different terms
including “role taking”, “role shifting” (Padden, 1986) or also, in earlier work ,“Body and
Projected Body pronouns” (Kegl, 1976), “body markers” (Pizzuto & al, 1990), with a primary
focus on the nonmanual features they present (marked facial expressions, gazes, body postures).
As noted, a relevant feature of HIS is that subtypes of Transfers can be combined among
themselves, or with standard signs, to encode information on two (or even more) referents in a
multilinear, simultaneous fashion that has no parallel in speech. This phenomenon is called
‘Double Transfer’ (DT) in Cuxac’s terminology (e.g. combining simultaneously a TP for
specifying an agent, and a TS specifying locative information or also a second agent). Similar
phenomena have been described from different perspectives and with different terminologies by
several authors (e.g., among others, Dudis, 2004; Russo, 2004; the collection of papers in
Vermeerbergen, Leeson Crasborn, 2007). In this paper we refer to these phenomena as “Multiple
Reference” (MR) operations, and we assess their incidence in the signed narratives we analyzed.
Illustrations of these Transfer types are provided below. Examples (2) and (3) are taken from
LSF narratives and highlight the differences between standard signs and HIS. Note how when
producing the standard signs the signer’s gaze is directed towards the interlocutor. In producing
HIS the gaze is directed away from the interlocutor, and oriented either towards the hands (when
articulating TF and TS structures), or in different points in space reproducing the gaze of the entity
represented (when articulating TP structures. Example (2) shows a standard signs and a TF, both
encoding the same conceptual content: ‘tree’. The proform (handshape within the TF) describes the
form of the tree.
(2) ‘tree’ via a standard sign (2a) and via a TF (2b-2c)
Example (3) shows a standard sign (3a) and a TP (3b), both encoding the same conceptual
content ‘horse’. In the TP structure, all the manual and non manual features (gaze, facial
expression, body and the hands) reproduce those of the embodied entity.
(3) ‘horse’ via standard sign (3a) and via a TP (3b)
Example (4) below illustrates three TS structures taken from, respectively, the LIS (4a), ASL
(4b) and LSF (4c) narratives analysed for the present study. Both the LIS (4a) and the ASL (4b)
example refer to ‘the falling of a dog out of a windowsill’. The LSF example (4c) describes ‘a
horse jumping over a fence’. In all TS, gaze is towards the dominant, then the non-dominant hand.
The dominant hand expresses the agent with the process (‘dog-falling’, ‘horse-jumping’), while the
non-dominant hand expresses the locative, and the object implicated in the locative relation
(‘windowsill’, ‘fence’). The facial expression is congruent with the process represented.
The examples in (5) illustrate two occurrences of MR taken from, respectively, the LIS (5a),
and ASL (5b) narratives analysed for the present study. In 5a, the TP and TS structure produced
allow the signer to anaphorically refer simultaneously to a ‘child holding a dog in his arm’ and to
‘the dog licking the child on his cheek’. In 5b, the same type of structures allows the signer to
represent ‘a dog’ with ‘a jar’ around his neck.
Note that TS structures such as those illustrated in (4) are also examples of MR.
1.2. Evidence from previous studies and questions investigated in the present study
Previous studies conducted on LSF have provided clear evidence on the extensive use of HIS in
LSF texts of different genres. This has been shown in Cuxac’s (1996; 2000) analyses of large texts
produced by a small number of LSF signers, and in more recent work conducted by Sallandre
(2003) on a large corpus of short narratives and ‘prescriptive’ (cooking recipes) texts produced by
19 signers. Sallandre’s results also highlight important differences with respect to discourse genres:
HIS are much more frequent in narrative (on average 70%) compared to prescriptive texts (30% on
average). Disregarding terminological differences among authors, similar indications on the
widespread use of HIS in different genres of signed discourse can be drawn from analyses and
observations reported for LIS (Pizzuto, 2007; Pizzuto & al, 1990; Russo 2004; 2005; Russo & al,
2001; Wilkinson, 2002), for BSL (Brennan 2001), ASL (Emmorey & Reilly 1998; Emmorey,
2003), DSL (Engberg-Pedersen, 1993; 2003).
No study that we are aware of, however, has explicitly addressed the question we aimed to
clarify with the present, cross-linguistic study:
1. How frequently are HIS used for performing deictic-anaphoric reference operations?
2. Are HIS more or less frequently used, for deictic-anaphoric reference purposes,
compared with ‘standard’ signs and manual pointings?
3. In performing deictic-anaphoric reference operations via HIS, what is the incidence of
MR, i.e. how frequently does the use of HIS allow signers to simultaneously introduce
or re-introduce in discourse two (or even more) referents?
2. Data used for the present study
The data we used for the present study were drawn from more extensive corpora of signed
discourse of different genres that have been collected in France, Italy and USA on a fairly large
number of native and non native ASL, LIS and LSF signers (Wilkinson, 2002; Sallandre, 2003;
2007; Pizzuto, Rossini, Russo & Wilkinson, 2005). In the present work we analyze short narrative
texts elicited through two different pictured stories that have been widely employed in much
research on both SL and spoken languages. The LIS and ASL narratives were elicited through the
same story, “Frog where are you?” (Mayer, 1969). The LSF narratives were elicited through “The
Horse” story (Hickmann, 2003).
In the book version we used as elicitation material, the ‘Frog’ story is made of 24 pictures. It
tells the adventures of a boy, his dog, a frog. The boy (main protagonist) finds a frog, brings it
home, puts it in a jar placed in his bedroom, then goes to sleep with the dog. During the night,
while the boy and the dog are sleeping, the frog jumps out of the jar and escapes. The next
morning, the boy wakes up and finds out that the frog has disappeared. The boy and the dog then
start searching everywhere for the frog.
The ‘Horse’ story is made of 5 pictures, hence much ‘shorter’ than the ‘Frog’ story. It
narrates simple actions performed by a horse, a cow and a bird within the space of a lawn divided
by a fence. The main protagonist is the horse, who gallops happily over one side of the lawn and is
observed by the cow and the bird positioned, respectively, on the other side of the lawn, and on a
picket of the dividing fence. At one point the horse jumps over the fence to join the cow on the
other side, but in doing so the horse bumps on the fence, and falls on the other side, landing on its
back and hurting one of its legs. The cow and the bird then come to help the horse: the bird brings a
first-aid kit which is then used by the cow for bandaging the horse’s leg.
These pictured stories were presented to all signers by deaf interviewers who are fluent in
each of the SL examined, and with whom the signers were well familiar. All signers were allowed
to familiarize themselves with the stories, with no time constraints, and then asked to retell the
stories from memory.
For the present study we selected from the larger corpora mentioned above, for each SL, the
productions of three native signers of comparable ages (young deaf adults, age range: 19-23 years)
and socio-cultural background (middle-class, high school degree, or attending the first years of
university). We focused our analysis on text sequences of comparable content (functionally similar
episodes involving animate and inanimate referents), and duration in time (approximately 1’ of
signed productions). The ASL and LIS data consisted of portions of ‘Frog Story’ narratives
corresponding to the sequence going from the beginning of the story through the episode in which
the dog falls from the windowsill, then the boy takes the dog in his arms, and the dog licks the boy
on his cheek. The LSF data corresponded to complete narrations of the ‘Horse Story’.
The reader may wonder why we used portions of the ‘Frog Story’ for our ASL and LIS data,
and a different story, ‘The Horse’, for our LSF data. Our choice was in part motivated by
indications provided by previous work, but also influenced by practical reasons. Research
independently conducted on ASL and LIS ‘Frog stories’ (Wilkinson, 2002), and on LSF ‘Horse
stories’ (Sallandre, 2003) indicated the relevance that semantic features such as animacy vs.
inanimacy, and human vs. not-human may have for a clearer understanding of deictic and
anaphoric devices in signed narratives. Both the stories we analysed include animate and inanimate
referents, but slightly differ with respect to the ‘human’ vs. ‘non-human’ feature of their
protagonists: the ‘Frog’ story has a human character (the boy) as the main protagonist, while in the
‘Horse’ story all characters are animals, hence not-human. This difference between the two stories
can thus provide information on the role that the human vs. not-human feature may have (within
the category of reference to animates) in the choice of deictic-anaphoric reference devices in
When we planned the research reported here, we intended to expand our data base in order to
perform our analyses on both “Frog ” and “Horse” stories produced in the three SL we examined.
However, time and funding constraints did not allow us to pursue this objective. The present
comparative study was thus re-designed, and should be intended, as a first exploration of the topic
we focus upon. We acknowledge that a more comprehensive analysis will require more data based
on the same language elicitation materials across languages.
2.1. Data transcription, coding and analysis
The transcription methodology, and the analytic categories we used for coding were agreed upon
among all co-authors on the ground of theoretical work and empirical analyses of signed discourse
as described in Cuxac (2000), Sallandre (2003) Cuxac & Sallandre (2007), Pizzuto (2007), Pizzuto
et al (2005).
All signed productions were transcribed and coded in a common format on excel files by
signers that were fluent in each of the national SL we considered5. We used so-called ‘glosses’ for
annotating in, respectively, written English, French and Italian the basic meaning of the linguistic
units we identified in the ASL, LSF and LIS narratives6. This ‘gloss’ notation was integrated, in the
5 Erin Wilkinson transcribed the ASL data, Paolo Rossini the LIS data, Marie-Anne Sallandre the LSF data
6 We are keenly aware that a so-called gloss-based notation, albeit still commonly used by practically all sign language
researchers, imposes severe limitations to the analysis of signed utterances and discourse, and more appropriate systems
for annotating in written form signed productions are sorely needed. For some recent work on this topic, and indications