2.1 Demonstrative semantics and phoric type

Demonstratives are traditionally divided into two main categories: exophoric and anaphoric (see e.g. Diessel 1999; Levinson 2018; Peeters & Krahmer & Maes 2021). A demonstrative’s status as exophoric or anaphoric is its phoric type (Botley & McEnery 2001). In some demonstrative systems, phoric type is exclusively contextual. For instance, in English, each demonstrative (this/that, here/there) displays both exophoric uses (Stand over there) and anaphoric ones (When Angel came into the room, Victoria was already there). This means that the phoric type of a given demonstrative token – whether it picks out a referent from the physical surroundings vs. from the discourse – must be inferred from context.

But in many other languages, including Korean (Ahn 2017: 41–42), Romanian (Ahn 2022), Yucatec Maya (Hanks 1990: 448–455), Tzeltal Maya (Brown & Levinson 2018), and several unrelated Amazonian languages (Guirardello-Damian 2018; Herrmann 2018; Meira 2018; Skilton 2019), phoric type is lexical. In these systems, some demonstratives are exclusively exophoric, always requiring a particular spatial relationship between the discourse participants and referent. Other demonstratives are exclusively anaphoric. They convey nothing about the spatial relationship between the participants and referent, only that the referent has been previously mentioned.

The Yucatec Maya discourse in (1), reproduced from Hanks (1990: 451), illustrates the alternation between lexically exophoric and anaphoric demonstratives over a discourse. In (1a), the speaker uses the exophoric demonstrative circumclitic téʔel…aʔ ‘here/there (exophoric immediate)’ to refer to his motion goal, a cave in the surroundings. In (1b), he refers to the same location again with the anaphoric demonstrative tíʔ…iʔ.

The Ticuna discourse in (2), from this study’s dataset, shows a parallel example of shift from lexically exophoric to anaphoric demonstratives. In (2a), the speaker uses the exophoric demonstrative ɟe⁵a² ‘there (distal),’ in a syntactic focus (cleft) construction, to introduce a new referent. In (2b), she refers to the same referent again. This time, instead of exophoric ɟe⁵a² ‘there (distal),’ she uses the anaphoric demonstrative ŋe⁵ma² ‘there (anaphoric),’ again in a cleft.¹

As the translations of (1) and (2) illustrate, the English demonstrative system does not make the distinction in phoric type that Ticuna and Yucatec do, instead using there for both distal exophoric functions and anaphoric ones. Thus, in English and other languages where phoric type is contextual, it is often ambiguous whether a given demonstrative token makes exophoric (especially distal) vs. anaphoric reference (Botley & McEnery 2001). But in demonstrative systems where phoric type is lexical, there is no such ambiguity. These systems therefore offer an ideal environment for analyzing the effects of phoric type on other properties of demonstrative reference.

2.2 Phoric type vs. information status

In demonstrative systems where phoric type is lexical, it is essential to distinguish the phoric type of a demonstrative from the information status of the demonstrative’s referent. In these languages, information status constrains the inventory of possible demonstratives for every reference, since lexically anaphoric demonstratives cannot index new referents. But this does not mean that information status deterministically controls demonstrative choice. All anaphoric demonstratives have previously mentioned referents, but exophoric demonstratives can have either discourse-new or previously mentioned referents.

The Ticuna example in (3) illustrates this. In line (3a), the speaker uses the lexically exophoric demonstrative nu⁵a² ‘here (near me)’ to make the first mention of the referent location. In (3b), she uses the same demonstrative again, this time in a cleft, for the same referent.

The token of the speaker-proximal nu⁵a² ‘here near me’ in (3a) introduces a new referent and the token in (3b) refers to an old one, but in both cases the referent is picked out from the surroundings rather than the discourse – that is, the demonstrative is exophoric, not anaphoric. Although the authors do not highlight the contrast between phoric type and information status, similar examples of multiple successive references with exophoric demonstratives also occur in Yucatec (Hanks 1990: 419) and Tzeltal (Brown & Levinson 2018: 167). Another example is the English translation of (3). In (3b), here has a discourse-old referent, but it is still exophoric: this discourse would be infelicitous in English if the referent was not near the speaker and did not meet the spatial requirements of here.

2.3 Co-demonstrative gesture

Following McNeill (1992), gesture researchers classify co-speech gestures into five main categories: pointing gestures, iconics, metaphorics, emblems, and beats. Pointing gestures identify a referent by projecting a vector from the speaker’s body to the referent. Many pointing gestures occur with demonstratives. I refer to these as co-demonstrative pointing gestures (Cooperrider 2023), and I refer to the combination of a pointing gesture and a demonstrative as a demonstrative composite utterance (Enfield 2009). To define the other types, iconic gestures depict the appearance of a referent; metaphorics represent an abstract concept by depicting an associated referent; emblems are conventional forms such as shaking the head for negation; and beats are non-referential gestures aligned with prosodic boundaries. These other gesture types are not completely mutually exclusive with pointing. For example, pointing gestures can have iconic handshapes (Cooperrider & Mesh 2022).

4.1 Predictions about gesture rate

Semantics and pragmatics research on demonstratives argues that gesture contributes to reference for exophoric demonstratives, but not for anaphoric ones. As a result, authors in this literature argue, anaphoric demonstratives “do not usually occur” or do not overlap with pointing, while exophoric demonstratives do (Diessel 1999; 2006; Ahn 2022). Thus, this claim predicts that Ticuna speakers will point less often when using the anaphoric demonstratives ŋe⁵ma² and ɟe⁵ma² than when using any of the four exophoric demonstratives (Prediction 1.1). Additionally, while Diessel and Ahn’s claims are not expressed in quantitative terms, they suggest that the gesture rate with anaphoric demonstratives will approach zero (Prediction 1.2). While Ahn’s claims specifically concern entity-referring demonstratives, Diessel’s claims do not distinguish between entity and location reference; therefore, these predictions apply to both types of reference.

Gesture studies research on information status also offers predictions about gesture rate. This literature indicates that familiar and/or discourse-old referents are associated with lower gesture rates, whether we consider iconic gestures (Debreslioska & Gullberg 2019), object-referring pointing gestures (Azar et al. 2019), or all gesture types together (Debreslioska & Gullberg 2022). These studies are specific to entity reference rather than place reference. However, if place reference patterns the same, these findings predict that referent information status will affect the co-organization of demonstratives and pointing: people will be more likely to point when introducing discourse-new referents than when referring back to previously mentioned ones (Prediction 1.3). This is a prediction about information status, not (only) about phoric type. It therefore requires us to distinguish between demonstratives with new vs. previously mentioned referents.

4.2 Predictions about gesture form

Gesture studies literature shows that, when people point at objects or places that are previously mentioned and/or otherwise identifiable to the addressee, they reduce the visual salience of their pointing gestures (Enfield et al. 2007; Peeters et al. 2015; Cooperrider et al. 2021). This reduction – perhaps the behavior that Ahn (2022) describes as pointing “in a casual manner” – affects many different articulatory features, including stroke duration, head orientation, and arm extension. For methodological reasons, I analyze only arm extension. For this variable, I predict, following Enfield et al. (2007) and Cooperrider et al. (2021), that speakers will be less likely to extend the arm when pointing at previously mentioned referents (Prediction 2.1).

While the reduction studies just cited are all quantitative, qualitative research also makes predictions about the relationship between information status and gesture form, specifically handshape. Ethnographers of pointing have argued that when people point at previously mentioned referents or make “anaphoric” mentions, they are less likely to use index-finger handshapes (Kendon & Versante 2003; Wilkins 2003); Flack and colleagues’ (2018) quantitative study shares this conclusion. In a possibly related pattern, people are also less likely to use index handshapes when pointing out directions vs. when pointing at locations (e.g., Haviland 2003). Thus, I predict that Ticuna speakers will be less likely to use index-finger pointing handshapes (a) when pointing at previously mentioned referents (Prediction 2.2) and (b) when pointing indexes a direction, rather than a location (Prediction 2.3). While I express the first of these predictions in terms of information status, Wilkins’ (2003) description actually contrasts “anaphoric” mentions with “first” mentions, meaning that it can also be understood as a prediction about phoric type.

5.1 Participants and procedure

Six Ticuna speakers from the town of Cushillococha, Peru, aged 35 to 72 years, participated in 30-minute monolingual interviews about the town’s landscape. Three participants – Sotil Suárez González (SSG), Deoclesio Guerrero Gómez (DGG), and Angel Bittancourt Serra (ABS) – were male. The other three – Ortencia Coello Guerrero (HCG), Lucinda Gómez Cordero (LGC), and Yaneth Cándido Guerrero (YCG) – were female. All of them were hearing, had no exposure to sign language, and spoke Ticuna as their sole first language. SSG, DGG, ABS and YCG spoke Spanish as sequential bilinguals. HCG and LGC understood some Spanish but did not speak it.

Interview questions were adapted from Kita’s (2001) landscape description task. Acting as a research assistant as well as a participant, Angel Bittancourt Serra translated the interview guide into Ticuna and assisted me in adapting it for the area. The interview questions prompted participants to describe the current and historical locations of landmarks; describe locations where they had lived; give route directions; and provide eyewitness accounts of a flood.

I interviewed all six participants in Ticuna, which I speak well as a second language. I conducted the interviews, rather than a native-speaker interviewer, because many of the Kita locality description questions would be pragmatically odd if asked by a person from the same location as the interviewee. To avoid priming effects, prior to debrief I did not inform participants that the study concerned pointing gestures; however, all of the participants were aware of my interest in demonstratives. Interviews took place in 2017 and 2018 in Cushillococha or the neighboring town of Caballococha. They were held outdoors, or in spaces with half-walls, in and near participants’ homes. Interviews were recorded in HD with one camera (2017: Sony PJR540, 2018: Canon XA30).

5.2 Speech coding

Angel Bittancourt Serra and Lilia Witancort Guerrero (also an L1 Ticuna speaker) worked with me to transcribe all speech in the interviews and translate it into Spanish. The first five minutes of each interview were transcribed, but treated as a warmup period and excluded from all further analysis.

I identified all locative demonstrative tokens in the transcripts and coded each token for phoric type, information status, and referent type (location vs. direction), all treated as binary. Speaker-proximal, dyad-proximal, speaker-distal, and regional demonstratives were coded as exophoric; anaphoric and past anaphoric demonstratives were coded as anaphoric. To code information status and referent type, I identified the referent of each demonstrative, then determined (a) whether it had been mentioned previously in the interview and (b) whether it was a location (point or region) vs. a direction (path or bearing). The discourse in (6) includes demonstratives of each phoric type, information status, and referent type.

I analyze only locative demonstratives (equivalent to here/there), not nominal demonstratives (this/that). This is because the nominal and locative demonstrative tokens were not comparable in count, distribution, or referent type. First, the transcripts contained many more locative demonstratives (724 tokens) than nominal ones (395 tokens). Second, while most locative demonstratives in the transcripts were exophoric (Table 5), most nominals were anaphoric (64.5%, 256 of 395 tokens). Third, a large fraction of the anaphoric nominal demonstratives had abstract referents such as time periods (that year) or propositions (they liked that). These kinds of referents are not comparable to places or concrete objects because they do not have locations in space. I also did not analyze non-deictic location references, which generally employed absolute or intrinsic frames of reference.

5.3 Gesture coding

Seven US-based research assistants coded the manual gestures in the footage. The coders were aware of the study hypotheses, but did not understand Ticuna and were not provided with translations of the audio. Coders identified all manual gestures in the footage. They were also trained to identify non-manual gestures, such as lip pointing. Coders did not identify any exclusively non-manual gestures. They did identify head and lip movements occurring with manual gestures, but these were not analyzed for reasons of interrater reliability (discussed later in this section).

Based on visual criteria, each manual gesture was coded for handshape, arm extension, and orientation of the speaker’s head relative to the pointing vector. Handshapes were coded according to a controlled vocabulary with seven possible values, shown in Table 3.

For arm extension, gestures were coded as displaying full arm extension if the participant’s forearm attained a 180° angle with their upper arm during the movement. Gestures where the elbow was already fully extended at the beginning of the movement were not coded as including full arm extension. As a proxy for eye gaze, coders also annotated the orientation of the speaker’s head as toward vs. away from the pointing target at the gesture peak. Arm extension and head orientation were treated as binary. All speech and gesture coding was performed in ELAN using Transcription Mode (Wittenburg et al. 2006; Dingemanse et al. 2012).

To assess reliability, a secondary coder re-coded the footage for 25% of the primary coder’s annotations, blinded to their codes. Primary and secondary codes were compared in R 4.3.1 (R Core Team 2023) using the irr package (Gamer et al. 2019). For handshape, the research questions were concerned mostly with the contrast between index and non-index pointing. I therefore transformed the handshape codes from Table 3 into a binary variable which contrasted handshapes with an extended index finger (index, index+1 and index+2) with all other handshapes. Inter-rater agreement on the binary handshape variable was 93.1% (κ = 0.831), indicating almost perfect agreement. For arm extension, inter-rater agreement was 92.9% (κ = 0.734), indicating substantial agreement, which I considered sufficient. For head orientation, inter-rater agreement was 77.8% (κ = 0.535), indicating only moderate agreement. Due to this low level of agreement, head orientation/eye gaze data was not further analyzed. Only primary coders’ results were used in the analyses.

Following gesture coding, I combined the speech and gesture transcripts using the fuzzyjoin and tidyverse packages (Wickham et al. 2019; Robinson 2020) and identified all demonstrative tokens that overlapped with gestures for >100ms. I reviewed the video of each gesture which overlapped with a demonstrative and coded the gesture type, taking into account both the gesture’s form and the co-occurring speech. The interview questionnaire, fully coded speech transcripts, fully coded gesture transcripts, and analysis code are included as Supplementary Materials.

6.1 Speech

Participants produced 742 total locative demonstrative tokens. As described in §3, the Ticuna Regional, Anaphoric, and Past Anaphoric demonstratives can appear in non-demonstrative functions, for example as relative pronouns. All tokens of these types were reviewed to identify any non-demonstrative uses. Twenty-two demonstrative tokens (3.0%) were used in non-demonstrative functions: 12 were temporal connectives, eight were relative pronouns, and two indexed the size of a referent. All 22 non-demonstrative tokens were excluded from further analyses.

After this, 720 demonstrative tokens remained in the dataset. Table 4 shows the number of tokens of each lexical item there. Table 5, grouping together all exophoric and all anaphoric forms, shows the number of tokens of each demonstrative lexical item which occurred with new vs. previously mentioned referents.

As Table 4 indicates, the majority of all exophoric demonstratives in the data were Speaker-Proximals and the majority of anaphoric demonstratives were the temporally unmarked Anaphoric item. Further, as the Exophoric Total rows of Table 5 show, exophoric demonstratives both introduced new referents and indexed previously mentioned referents, in line with the theoretical predictions from §2.2. In particular, the Speaker-Proximal exophoric demonstrative indexed previously mentioned referents more than three times as often as it indexed new referents. The other exophoric demonstratives were about equally likely to index new or previously mentioned referents.

While I predicted above that anaphoric demonstratives would index only previously mentioned referents, the Anaphoric Total rows in Table 5 show that two tokens of anaphoric demonstratives actually introduced new referents. (7) shows one of the two new-referent anaphoric tokens. This example is biclausal. In the first clause, the speaker describes an event; in the second clause, he refers to the location of the event with the past anaphor ɟe⁵ma².

The token of ɟe⁵ma² in the second clause of (7) can be seen as introducing a new referent, since it is not coreferential with any noun phrase in the subordinate clause or earlier in the discourse. On the other hand, this location can also be seen as previously mentioned, since it is the location of a previously mentioned event (the landing) and referent (the airplane). The other new-referent token of an anaphoric demonstrative in the dataset involves a similar bridging use, indexing the location of a previously mentioned event. As the data included only two anaphoric tokens which displayed this type of ambiguous information status, I excluded them from all analyses, leaving 718 demonstratives in the dataset.

Finally, turning to referent type, Table 6 reports the total number of demonstratives which indexed locations vs. directions, by phoric type and information status.

As Table 6 shows, almost all demonstratives indexing directions – 179 of 187 (95.7%) – were exophoric. This reflects that participants frequently indexed the direction of a place using an exophoric demonstrative, then referred back to the place itself with an anaphoric term. (8) is an example of this structure. In (8a), the participant describes the direction in which the town of Erené lies, relative to another town that she has already mentioned, Bellavista. Her direction reference uses the speaker-distal exophoric demonstrative. In (8b), she states that her brother lives in Erené. This is a location reference, not a direction one (the man lives in Erené, not in its direction); it uses an anaphoric demonstrative.

6.2 Gesture

Of the 718 demonstrative tokens analyzed, 512 co-occurred with gestures. In 475 of these composite utterances, the gesture was classified – based on its form and the content of the co-occurring speech – as a pointing gesture coreferential with the demonstrative. In the other 37, the gesture either was not a point or was not coreferential with the demonstrative. Table 7 classifies all of the co-demonstrative gestures in the data.

Since the 33 iconic, beat, and emblem gestures do not index places, they are not comparable to the place-referring pointing gestures that occurred with other demonstratives. The points at the speaker or addressee may index places (e.g. on their bodies), but are not comparable to points that index places and are reinforced with a coreferential demonstrative. These composite utterances were therefore excluded from all analyses, leaving 475 demonstrative composite utterances and 681 total demonstrative tokens in the dataset.

7.1 Gesture rate

On average, participants pointed with 66.5% (range: 30.0–87.7%) of all demonstrative tokens. Specifically, pointing occurred with 88.8% (range: 61.1%–100%) of exophoric demonstratives indexing new referents, 70.2% (range: 33.3%–89.9%) of exophoric demonstratives indexing previously mentioned referents, and 43.3% (range: 4.0%–62.5%) of anaphoric demonstratives indexing previously mentioned referents. Figure 1 displays the proportion of exophoric vs. anaphoric demonstratives, with new vs. previously mentioned referents, occurring with pointing gestures for each participant. Figures 2 and 3 provide video still examples of two pointing gestures occurring with exophoric (Figure 2) and anaphoric (Figure 3) demonstratives.

Figure 1

Proportion of demonstratives occurring with a pointing gesture, by phoric type of the demonstrative and information status of the referent.

Figure 2

Pointing gesture accompanying nu⁵a²ta²ã⁴ ‘right here (near me; exophoric).’

(ABS, CLA 2015-06.039, tca_20170825_abs_ahs_elicit_video_002_archive.mp4, 2:09).

Figure 3

Pointing gesture accompanying ɟe⁵ma²ta²ã⁴ ‘right there (where I mentioned; anaphoric).’

(ABS, CLA 2015-06.039, tca_20170825_abs_ahs_elicit_video_001_archive.mp4, 9:55).

Figure 1 visually suggests that both phoric type and information status affected participants’ rate of pointing. However, because the dataset contains no anaphoric demonstratives with new referents, only three of the four possible cells of the phoric type by information status interaction are defined. This makes it impossible to construct models which include both of these fixed effects and their interaction. Instead, I analyzed the effects of phoric type and information status by dividing the data into two subsets and fitting a separate model to each subset, as described in detail in the following paragraphs.³ Because referent type (location vs. direction) is only relevant to my predictions for handshape, not for gesture rate, I did not include it in these models.

First, in order to investigate differences in gesture rate motivated by phoric type, I constructed a subset of the data which included all and only the demonstratives – exophoric or anaphoric – with previously mentioned referents (n = 512). This subsetting holds information status constant, allowing us to isolate the effect of phoric type. Using mixed-effects logistic regression, I modelled the outcome of the presence of a pointing gesture with phoric type as the sole predictor variable; random intercepts for participants; and by-participant random slopes for phoric type. In this and all subsequent models, phoric type was coded as binary with exophoric as the reference level. I observed a significant effect of phoric type on the presence of co-demonstrative pointing gestures. When participants indexed previously mentioned referents, they were less likely to point if they used anaphoric demonstratives than if they used exophoric ones (β = –1.37, SE = 0.23, p < 0.001).

Next, to investigate the effect of information status on gesture rate, I constructed a subset of the data which included all and only the exophoric demonstratives (n = 497) – whether they indexed new or previously mentioned referents. As in the previous analysis, this subsetting holds phoric type constant, allowing us to isolate the effect of information status. Using mixed-effects logistic regression over this subset of the data, the binary outcome of the presence of a pointing gesture was modelled with information status, spatial deictic content, and their interaction as predictors, and random intercepts for participants.⁴ Since this model includes random intercepts but not random slopes, it assumes that participants vary in baseline gesture rate, but not in the effects of information status or spatial deictic content on gesture rate. To create the spatial deictic content variable, Speaker-Proximal and Regional demonstratives were coded as proximal, Dyad-Proximal demonstratives were coded as medial, and Speaker-Distal demonstratives were coded as distal. The variable was dummy-coded with proximal as the reference level. Information status was coded as binary with new as the reference level. This model observed a significant effect of information status. When participants used exophoric demonstratives, they were less likely to point if the referent was previously mentioned than if it was new (β = –1.22, SE = 0.40, p = 0.002). For exophoric demonstratives, there was no evidence for an effect on gesture rate from use of medial or distal demonstratives, or from interactions between medial/distal demonstrative use and information status (all p-values > 0.1).

7.2 Gesture form: arm extension

Next, I analyzed the relationship between gesture form, demonstrative phoric type, and referent information status in the 475 demonstrative composite utterances. My first form analysis concerned arm extension. On average, participants extended the arm completely in 46.2% (range: 27.3%–75.7%) of gestures occurring with exophoric demonstratives indexing new referents; 25.7% (range: 9.1%–56.5%) of gestures with exophoric demonstratives indexing previously mentioned referents; and 16.8% (range: 0%–35.3%) of gestures with anaphoric demonstratives indexing previously mentioned referents.

Figure 4 displays the proportion of co-exophoric vs. co-anaphoric gestures, with new vs. previously mentioned referents, occurring with full arm extension for each participant. Figures 5 and 6 show two example gestures with vs. without full arm extension, both occurring with (different tokens of) the exophoric, speaker-distal demonstrative ɟe⁵a².

Figure 4

Proportion of co-demonstrative pointing gestures with full arm extension, by phoric type of the demonstrative and information status of the referent.

Figure 5

Pointing gesture with full arm extension, accompanying ɟe⁵a²ã⁴ma⁴ma³ ‘toward there (far away; exophoric).’

(ABS, CLA 2015-06.039, tca_20170825_abs_ahs_elicit_video_001_archive.mp4, 6:44).

Figure 6

Pointing gesture without full arm extension, also accompanying ɟe⁵a²ã⁴ma⁴ma³ ‘toward there (far away; exophoric).’

(ABS, CLA 2015-06.039, tca _20170825_abs_ahs_elicit_video_001_archive.mp4, 7:49).

My first analysis of this data examined phoric type. To control for the effects of information status in this analysis, I subsetted the data to include only composite utterances with previously mentioned referents (n = 324). I then modelled the outcome of the presence of full arm extension with phoric type as the sole predictor variable and random intercepts for participants, but no random slopes.⁵ The model indicated no significant effect of phoric type on the presence of full arm extension (β = –0.52, SE = 0.32, p = 0.11).

My next analysis considered the impact of information status on arm extension. To control for the effects of phoric type, I subsetted the data to include only observations with exophoric demonstratives (n = 396). I then modelled the outcome of full arm extension with information status, spatial deictic content, and their interaction as predictors, and random intercepts for participants.⁶ This model indicated a significant effect of information status on the presence of full arm extension. When participants produced an exophoric demonstrative and pointed, they were less likely to extend their arm completely if the referent was previously mentioned than if it was new (β = –0.94, SE = 0.33, p = 0.005). There was no significant effect of using medial or distal demonstratives, or of the interactions between medial/distal demonstrative use and information status (all p-values > 0.6), on the presence of full arm extension with exophoric terms.

7.3 Gesture form: index handshape

The second form analysis examined participants’ use of index-finger pointing handshapes. On average, participants pointed with the index finger in 43.2% (range: 18.2%–59.5%) of gestures accompanying exophoric demonstratives with new referents; also 43.2% (range: 27.3%–59.6%) of gestures accompanying exophoric demonstratives with previously mentioned referents; and 23.3% (range: 0.0%–47.1%) of gestures accompanying anaphoric demonstratives. Figure 7 summarizes participants’ use of index handshapes by phoric type and information status.

Figure 7

Proportion of co-demonstrative pointing gestures with index handshape, by phoric type of the demonstrative and information status of the referent.

As Figure 7 shows, these Ticuna participants do not share Western speakers’ (Cooperrider & Slotta & Núñez 2018; Flack & Naylor & Leavens 2018) strong preference for pointing with the index finger. Overall, only 210 (44.2%) of the 475 co-demonstrative pointing gestures used an index handshape. Because of reliability issues (§5.3), I did not conduct detailed analyses of the different non-index handshapes. However, based on the primary coders’ annotations, the most common non-index pointing handshapes in the dataset involve the entire hand: they are the flat (ASL B-like) and open (ASL 5-like) handshapes. Figures 8 and 9 provide examples of these handshapes occurring with exophoric (Figure 8) and anaphoric (Figure 9) demonstratives.

Figure 8

Pointing gesture with flat handshape, accompanying nu⁵a² ‘here (near me; exophoric).’

(LGC, CLA 2015-06.069, tca_20180705_lgc_ahs_haldi_archive.mp4, 10:20).

Figure 9

Pointing gesture with open handshape, accompanying ŋe⁵ma² ‘there (where I mentioned; anaphoric).’

(LGC, CLA 2015-06.069, tca_20180705_lgc_ahs_haldi_archive.mp4, 6:19).

In light of the theorized relationship between referent type (direction vs. location) and handshape (§2.3.2), I also calculated the proportion of gestures using the index handshape for each referent type. These figures do not immediately suggest a difference between locations and directions: on average, participants used index handshapes in 38.3% (range: 13.2%–62.0%) of points at locations and 40.0% (range: 27.3%–54.2%) of points at directions.

My first analysis of the handshape data examined the effects of phoric type and referent type. As in the arm extension analysis, I controlled for information status by subsetting the data to consider only observations with previously mentioned referents (n = 324, per above). I then modelled the outcome of use of the index-finger handshape with phoric type and referent type as predictors, random slopes, and random intercepts for participants.⁷ Referent type was coded as binary with location as the reference level. The model indicated a significant effect of phoric type on participants’ use of index-finger pointing handshapes: when participants used a demonstrative and a gesture to index a previously mentioned referent, the gesture was less likely to display an index handshape if the demonstrative was anaphoric than if it was exophoric (β = –0.95, SE = 0.46, p = 0.040). However, the model indicated no significant effect of referent type on handshape: for previously mentioned referents, participants were no less likely to use index handshapes when pointing at directions than when pointing at locations (β = –0.01, SE = 0.48, p = 0.98).

My second analysis of the index-finger pointing data considered information status. Again as in the arm extension analysis, I controlled for the effect of phoric type by subsetting the data to include only observations with exophoric demonstratives (n = 396, per above). I then modelled the outcome of use of an index-finger handshape with information status, spatial deictic content, and referent type as predictors and random intercepts for participants.⁸ The model indicated no significant effect of information status (β = –0.14, SE = 0.24, p = 0.55), referent type (β = –0.37, SE = 0.24, p = 0.12), or use of a medial demonstrative (β = –0.09, SE = 0.44, p = 0.84) on the use of index-finger pointing handshapes with exophoric terms. However, there was a significant effect of the use of distal demonstratives: people used index handshapes more often with distals (β = 0.54, SE = 0.23, p = 0.02) than with proximals.

8.1 Both phoric type and information status affect gesture rate

I considered two sets of predictions about the relationship between the phoric type of demonstratives, the information status of demonstrative referents, and the frequency of co-demonstrative pointing gestures. First, linguistic research on demonstratives claims that anaphoric demonstratives rarely or never occur with pointing, while exophoric demonstratives often do (Diessel 1999; 2006; Ahn 2022). For this dataset, this claim predicts that phoric type will influence the rate at which pointing gestures co-occur with demonstratives, and more specifically, that the gesture rate with anaphoric demonstratives will be near zero. In line with this prediction (Prediction 1.1), I do observe effects of phoric type on gesture rate: participants point less with anaphoric demonstratives than with exophoric demonstratives. Contrary to the stronger claims in this literature, however, the effect of phoric type is not categorical (Prediction 1.2). Numerically, participants pointed relatively often with anaphoric demonstratives, and they quite often omitted pointing with exophoric demonstratives, especially when they indexed previously mentioned referents (Figure 1). Second, gesture studies research – by Azar and colleagues (2019), Debreslioska and Gullberg (2019; 2022), and many others – has shown that references to previously mentioned entities occur with fewer gestures than references to discourse-new entities. This predicts that people will point less with demonstratives indexing previously mentioned referents (Prediction 1.3). The results fell exactly in line with this prediction. Participants pointed more with demonstratives used for new referents than for previously mentioned referents, even when only exophoric terms were considered.

Thus, this analysis partly confirms the theoretical position that deixis and anaphora have different relationships with gesture, and partly challenges it. Because we observe an effect of phoric type, these findings do support the core argument of the literature on demonstratives and gesture – that gesture is more important to fixing reference for exophoric demonstratives than anaphoric ones (Ahn 2022). However, these findings empirically disprove the claim that anaphoric demonstratives do not occur, only rarely occur, or do not overlap with pointing (Diessel 1999; 2006; Ahn 2022).

Some readers may ask why participants so often pointed with anaphoric demonstratives, since the point in this type of composite utterance is redundant (i.e., repeats location information that was already present in earlier references). I offer two responses to this question. First, the locality description interview questions require participants to constantly introduce new locations into the discourse, as well as referring back to many previously introduced ones. Thus, location references in this task – whether exophoric or anaphoric – are highly ambiguous. This may lead participants to increase their rate of pointing in order to disambiguate between multiple recently mentioned referents (cf. Azar et al. 2019). For the same reason, location reference in this task is very often contrastive. Thus, many demonstrative tokens in the data appear in syntactic focus (cleft) constructions (e.g., 2a, b; 3b; 8b). Contrastive focus (as a semantic category) and/or clefting could conceivably also elevate gesture rates with anaphoric demonstratives.

Second, I offer a more general response to the idea that producing the same information multiple times in a discourse or utterance is redundant. If this is true, so-called redundancy is a core feature of both grammar and co-speech gesture. For instance, morphological agreement between a verb and an overt argument is arguably redundant. Similarly, the literature on iconic gesture observes that – while people sometimes express entirely different information in gesture and speech – they also sometimes express overlapping information in the two channels (Goldin-Meadow 2003: Chapter 7; Alibali et al. 2009). There is no reason to predict that demonstratives, or the gestures that accompany them, will be less redundant than other speech and gesture.

For gesture studies research, on the other hand, this study makes two contributions. First, it replicates in a less-studied language the finding that new referents are associated with higher gesture rates than previously mentioned referents (Foraker 2011; Debreslioska & Gullberg 2019; 2022, among others). Second, it shows that referring expressions can belong to the same syntactic category, but still be associated with different gesture rates. This finding has precedent in results showing that speakers of Turkish (Azar et al. 2019) and German (Debreslioska & Gullberg 2022) gesture more with noun phrases than with pronouns. Azar and colleagues, as well as Debreslioska and Gullberg, interpret this result as showing that gesture rate is sensitive to the semantic richness of the co-occurring speech. They argue that NPs are associated with higher gesture rates because they are “richer” – i.e., convey more information about the referent – than pronouns. In contrast, my findings about the association between exophoric demonstratives and higher gesture rates cannot be explained by the richness (i.e., informativity) of the co-occurring speech. Exophoric demonstratives and anaphoric demonstratives convey equal amounts, but different types, of information about their referents: exophoric demonstratives convey the referent’s location in space, while anaphoric demonstratives convey its information status. As such, these findings support the conclusion that gesture rate is sensitive not only to the quantity of information that the co-occurring speech conveys, but also to whether this information is spatial. Speakers gesture more when they produce referring expressions that convey spatial information – that is, exophoric demonstratives – than when they produce equally informative expressions, of the same syntactic category, which convey non-spatial information – that is, anaphoric demonstratives.

8.2 Information status, but not phoric type, affects arm extension

In addition to my analysis of gesture rate, I also tested predictions about the effect of phoric type and information status on gesture form. Gesture studies literature (e.g., Enfield et al. 2007) has shown that when more information about a referent is in the common ground, people point at the referent using reduced, or less visually salient, gestures. Using arm extension as a measure of articulatory reduction, I predicted that Ticuna speakers would extend the arm less often when pointing to previously mentioned referents than when pointing to new ones (Prediction 2.1).

My results upheld this prediction: when speakers used a demonstrative and pointed, their pointing gesture was less likely to include full arm extension if the referent was previously mentioned. This effect strongly resembles Enfield and colleagues’ findings (2007) about arm extension for Lao speakers, as well as Cooperrider and colleagues’ (2021) results on arm extension for English speakers and ASL signers. The difference between these results and theirs concerns the content of the co-occurring speech. Previous authors compare composite utterances where the speech contains more vs. less location information, whether this is described in terms of focus (Enfield et al. 2007) or as sharing the informational load with the gesture (Cooperrider et al. 2021). In contrast, all composite utterances in my information status analysis contained exophoric demonstratives. Thus, all of these utterances include the same minimum degree of location information, showing – much as in the rate analysis – that reduction in gesture form can occur even in the absence of total reduction in the semantic richness of speech (e.g. in the zero location anaphora examined by Enfield and colleagues). In contrast to information status, phoric type had no effect on participants’ arm extension. This is conceptually in line with Mesh’s (2017) finding that, in co-demonstrative pointing gestures by speakers of San Juan Quiahije Chatino (Oto-Manguean; Mexico), there is no relationship between elbow height (a correlate of arm extension) and (exophoric) demonstrative lexical item.

My analysis of arm extension has the methodological limitation that arm extension was coded visually and as a binary variable. In reality, arm extension is continuous. Analyzing it as such, for example using computer vision (Pouw & Trujillo & Dixon 2020), could have produced different results. Another limitation is that I did not analyze non-deictic location information, such as place names or location descriptions, that occurred in the speech accompanying gestures. Future work should explore how the presence and content of other location information affects arm extension (Cooperrider et al. 2021).

8.3 Phoric type, but not information status or referent type, affects handshape

As a second dimension of gesture form, I also analyzed handshape. Wilkins (2003) and Kendon and Versante (2003) have suggested that people are more likely to use index-finger handshapes (compared to whole-hand handshapes) on first mention of a referent (cf. Flack & Naylor & Leavens 2018), for more important referents, and for emphasis. With this background, I predicted that Ticuna speakers would be more likely to use index handshapes when pointing at new referents (Prediction 2.2). Additionally, given the many observations in the literature about the relationship between index-finger pointing and location (vs. direction) reference, I predicted that speakers would be more likely to use index handshapes when pointing at locations (Prediction 2.3).

The results were not consistent with either Prediction 2.2 or 2.3. There was no effect of either information status or referent type (direction vs. location) on the use of index-finger pointing gestures accompanying demonstratives. There was, however, an effect of phoric type. Points that accompanied exophoric demonstratives were more likely to display index handshapes than points that accompanied anaphoric demonstratives. This finding cannot be readily explained in terms of articulatory reduction, since the flat and open handshapes that predominate in this data are not necessarily less effortful or visually salient than the index handshape. There was also an effect of spatial deictic content: among co-exophoric points, people were more likely to use index handshapes with distal demonstratives than with proximals. These findings on phoric type and spatial deictic content are surprising, because other research on pointing has not suggested associations between index handshapes and specific demonstrative lexical items. For example, Mesh (2017: 99–100) observed no effect of the contrast between demonstratives and other referring expression types, and no effect of any individual demonstrative, on Chatino speakers’ use of index vs. open pointing handshapes. Likewise, despite an extensive discussion of handshape, Wilkins (2003: 193) describes no influence of demonstrative lexical item on handshape for Arrernte speakers.

The finding that there was no effect of referent type on handshape also departs from many observations in the literature, as well as from my own previous qualitative impressions (and many individually clear video examples) of Ticuna speakers’ handshape use. One possible explanation is that other properties of the referents in this dataset, such as size or distance (Mesh 2017), favor the use of non-index handshapes so strongly that they overshadow effects of the location vs. direction contrast. For instance, unpublished data on child-caregiver interaction which I collected in 2019 shows that when pointing at objects in their immediate surroundings, Ticuna adults use >90% index handshapes. This suggests possible relationships between handshape and distance, or between handshape and the contrast between object and place reference. Future research should pursue these possibilities further. Last, and more general, these handshape results provide further evidence that whole-hand pointing is cross-culturally commonplace (cf. Wilkins 2003; Mesh 2017; Flack & Naylor & Leavens 2018; Fenlon et al. 2019). Researchers should therefore avoid excluding whole-hand gestures from analysis in studies of pointing (Enfield et al. 2007: 1725).