Formal and functional factors in classifier choice: Evidence from American Sign Language and Danish Sign Language

3.1 Participants

The data for the current study are drawn from narrative retellings of the Canary Row cartoon. The DTS narratives were elicited in 2002 from five native signers. Three of the signers were in their late 20s and two in their 50s at the time of recording. They all have deaf-signing parents, two of them are a mother and her son. Data collection took place at the school for the Deaf in Copenhagen.

Narratives from 15 ASL signers were collected in 3 different data collection periods 2003, 2006, and 2013 (24–67 years of age at the time of recording). The data from 2003 and 2013 were collected in the Brentari lab, and the 2006 data were collected in the signers’ homes. All of the signers were native signers with two Deaf parents, or early learners, who acquired ASL before age 5. Narratives from all 15 ASL signers are included in the analysis of volitionality, and from 9 of the signers elsewhere.

3.2 Procedure

Participants were shown the Canary Row cartoon in its entirety (6 min 21 s), and then again, episode by episode (8 episodes in total, between 24 s and 1 min 16 s per episode). Each episode has the narrative structure of the cat Sylvester figuring out a plan to try to get at the bird Tweety in its cage, Sylvester carries out the plan, but fails, and it has fatal consequences for Sylvester himself. The signers retold the narrative to another fluent signer, who was the experimenter and had also seen the cartoon.

3.3 Annotation

All verbs of motion and location in the data were annotated in ELAN, the classifier verbs according to a template established by the collaborative project (Kimmelman and Khristoforova 2025; the template is published at OSF, cf. the references). We also made the following additions.

Volitionality [±volitional]: Volitionality refers to the characteristic of a referent in a predicate, indicating whether the referent intended to perform the action or, in the case of stative classifier predicates, intended to be in a particular state or location. Sylvester climbing the inside of a drainpipe would be considered volitional, as this was his intended action; however, Sylvester falling down that same drainpipe would be considered non-volitional.

Sign type [T0, T1, T2, and T3]: Sign type refers to the categories as established by Battison (1978): Type 0 signs are one-handed signs (e.g., ASL think and DTS tænke ‘think’).^[3] Type 1 signs are two-handed signs in which the hands are identical in form, both hands are active, and they perform identical motor acts (ASL meet and DTS træffe ‘meet’ ). Type 2 signs are two-handed signs in which one hand is active and one hand is passive, but both hands have the same handshape (e.g., ASL sit and DTS arbejde ‘work’). Type 3 signs are two-handed signs in which one hand moves (active) and one hand is static (passive), and the two hands have different handshapes (e.g., ASL touch and DTS værdi ‘expensive’). Battison established this system for lexical signs, and the constraints were extended to classifier signs for ASL, Swiss German Sign Language (DSGS), and Hong Kong Sign Language (HKSL) and found to also apply in such forms when handshapes were analyzed using phonological features (Eccarius and Brentari 2007).

Handshape of the nondominant hand : If a sign is two-handed, Battison (1978) proposed the Symmetry and Dominance Conditions, which (among other things) restrict the shape of the nondominant hand (NDH) to a set commonly known as BASCO15 for the handshapes of the letters of the ASL manual alphabet and the number signs one and five (Battison 1978, 35). These handshapes were annotated according to the system developed by Eccarius and Brentari (2008).

Same or different event : We indicated whether or not a sign with a classifier described the same event as any other sign in the narrative (Section 7).

4.1 Phonological analysis: DTS

In the following comparisons, we ignore Type 1-signs and Type 2-signs as they had less than 5% of the total number of classifier signs. As Type 3-signs, we include all simultaneous combinations where at least one hand is used for 1-cl, 1-Parallel, or 2-cl, and the handshape of the other hand is meaningful (Figure 2b and d for examples). We find that DTS signers are more likely to use the 2-handshape (i.e., use 2-cl) than the 1-handshape (i.e., use 1-cl or 1-Parallel) in classifier signs, despite the increased complexity of the 2-handshape. Using a Mann–Whitney U test of ranked comparisons, we find that this difference is significant overall (U = 0.05; z = −272; p = 0.05). Note also that we find no significant differences between the proportion of the 1-handshape in two-handed (Type 3) signs vs overall proportions of use in the corpus (i.e., in all classifier signs in the DTS data irrespective of whether they are one-handed or two-handed). This suggests that the overall complexity of the two-handed form is not the main criterion for handshape selection.

Moreover, in terms of phonological sub-types, there is just one sub-type for each of the handshapes of 1-cl and 1-Parallel, while there are eight subtypes of the 2-handshape in 2-cl: 2, 2+mov, 2-stacked, 2-bent, 2-bent+mov, N, N-bent, and N-bent+mov. This indicates strongly that the purely phonological criteria are not very highly ranked when considering the distribution of these two classifier handshapes. Other factors must play a larger role in handshape selection. Figure 3 provides a breakdown from the DTS corpus of 1-handshape and 2-handshape forms across Type 0-signs, Type 3-signs, and overall.

Figure 3

Proportion of DTS use of the 1-handshape (of 1-cl or 1-Parallel) vs the 2-handshape (of 2-cl) in Type 0 (one-handed signs), Type 3 signs (two-handed with one active hand with a classifier and one passive, but meaningful hand), and overall.

4.2 Phonological analysis: ASL

We find that ASL signers are more equally balanced between their use of the 1-handshape (of 1-cl and 1-Parallel), and the 2-handshape (of 2-cl) at least in Type 3-signs, which make up 84% of all signs with the targeted classifiers in the ASL data. Using a Mann–Whitney U test of ranked comparisons we again (as in DTS) find no significant differences between the proportion of the 1-handshape or 2-handshape overall compared to their proportions in Type 3-signs (Figure 4). As in DTS, this suggests that the overall complexity of the two-handed form is not the main criterion for handshape selection.

Figure 4

Proportion of ASL use of the 1-handshape (of 1-cl/1-Parallel) vs the 2-handshape (of 2-cl) in Type 0 (one-handed) signs, Type 3 (two-handed with one active hand with a classifier and one passive, but meaningful hand) signs, and overall.

The ASL signers are also more balanced than the DTS signers in the number of subtypes used in the handshapes of 1-cl, 1-Parallel, and 2-cl. In ASL there are a total of seven subtypes of the 1-handshape: for 1-cl there are four (1, 1+mov, 1-bent, and 1-bent+mov), and for 1-Parallel, there are two (1-parallel, 1-parallel+mov). There are nine subtypes of the 2-handshape: 2, 2+mov, 2-bent, 2-bent+mov, 2-stacked, 2-stacked-bent, 2-stacked-bent+mov, N, and N-bent.

The phonological profiles of the two languages demonstrate several important things. The 1-handshape is less complex than the 2-handshape, and we predicted that (i) the 1-handshape would be more frequent, (ii) have more subtypes, and (iii) appear more frequently in Type 3 signs. None of the three phonological predictions were confirmed. The 1-handshape is not more frequent, does not have more sub-types, and does not appear more often in Type 3-signs than the 2-handshape. The fact that the 1-handshape is simpler phonologically is not predictive of its use in the classifier system of either language.

We also see that there are crosslinguistic differences between the DTS signers’ and the ASL signers’ use of the targeted handshapes (Figure 5); there is a marginally significant difference between the use of the 1-handshape (of 1-cl and 1-Parallel) across the two languages (Mann–Whitney U = 12; z = 1.53; p = 0.06). Table 1 shows a summary of crosslinguistic similarities and differences between DTS and ASL, and for further analysis of the differences between the two languages, check Section 5.

Figure 5

Differences between DTS and ASL in their overall proportion of use of the 1-handshape (of 1-cl and 1-Parallel) and the 2-handshape (of 2-cl).

5.1 An earlier analysis of the difference between 1-cl and 2-cl in DTS

Engberg-Pedersen (1993, 247–51) described the meaning of what she called Index-Pm (∼1-cl) and V-Pm (∼2-cl) (Pm for polymorphemic) in DTS. For DTS, she observed that 1-cl is used in verbs about “an individual’s approaching or passing the holder of the point of view” and “an individual’s going away from the holder of the point of view” (Engberg-Pedersen 1993, 248–49) if the individual stays within the same area. It can also be used in signs about someone’s motion toward a goal or along a predestined route such as in a stadium, but not complex routes like a maze. Finally, it is used about two people meeting each other, especially in a place not set out for a meeting, such as the street or a corridor, and in signs about their separating after having met. But it cannot be used when the manner of motion or location is focused, such as in stative signs with a focus on the manner (‘sit’, ‘lie’, and ‘stand’), in signs about jumping or walking in contrast to bicycling or driving, walking backward (here a sign with the BPCL for legs would be used instead), or moving somewhere in a cumbersome way or on a difficult surface. In contrast to 2-cl, 1-cl is also not used if the sign mainly describes a change of location such as ‘go into the kitchen to get something’.

In its use in signs about motion toward a goal or along a route, 1-cl overlaps with the use of 2-cl. The latter is a more obvious choice in signs about motion or location when the focus is on the being’s walking or on someone’s motion in a certain direction, to a goal, or within an area with no relation to a holder of the point of view. 2-cl is also used in signs about the manner of a stative verb. The selected fingers of 2-cl can be bent and move independently of each other.

5.2 The difference between 1-cl and 1-Parallel in the two languages

As mentioned in the overview of classifier types in Section 1, 1-cl differs from 1-Parallel in both form and meaning. The index finger of 1-cl does not bend in any joint in DTS in the current data or any other DTS data that the first author has ever seen, while the index finger of 1-Parallel can bend in all joints when it is moved along a path with an angle of ninety degrees or in a curve. The finger of 1-cl is perpendicular to the plane of motion, i.e., it is held horizontally about someone moving up or down (Figure 6) and held vertically about someone moving horizontally (Figure 7). By contrast, the finger of 1-Parallel is held parallel to the plane of motion: in descriptions of someone moving up, the finger is held vertically (Figure 8a), and in descriptions of someone moving parallel to the ground, the finger is held horizontally (Figure 8b).

Figure 6

1-cl in a sign about vertical motion (Sylvester climbing up to Tweety Bird on the outside of a pipe as seen from the latter’s perspective). The finger of 1-cl is held horizontally and perpendicular to the plane of motion. The example is from DTS.

Figure 7

1-cl in a sign about horizontal motion (Sylvester approaching Tweety Bird from behind). The finger is held vertically and perpendicular to the ground. The example is from DTS.

Figure 8

(a) 1-Parallel in a sign about a vertical motion (Sylvester climbing up to Tweety Bird). The finger is held vertically and parallel to the plane of motion. (b) 1-Parallel in a sign about horizontal motion (Sylvester crossing a street). The finger is held horizontally and parallel to the ground. The examples are from DTS.

We are not aware of any widely accepted way of distinguishing 1-cl and 1-Parallel in linguistic descriptions of ASL. Besides the difference in orientation in relation to the plane of the articulator’s movement in the signs (Table 1), a formal difference between what could be analyzed as signs with 1-cl and signs with 1-Parallel in ASL is that the articulators of the two types of signs show different forms of hand-internal movement in descriptions of someone moving along a path: the finger of 1-cl may bend repeatedly in the two outer joints as the articulator is moved (Figure 9). When what may be seen as 1-Parallel is used in ASL in a sign about horizontal motion along a path, the articulator may bend repeatedly in the knuckle joint during the movement (Figure 10). These types of hand-internal movements are not seen in 1-cl and 1-Parallel in DTS, and 1-cl seems to be a more comprehensive classifier for anthropomorphic beings’ walking in ASL than in DTS.^[5]

Figure 9

Bending of the articulator at the Distal joint in a sign with 1-cl in ASL, the two extreme positions.

Figure 10

Bending of the articulator at the Proximal joint in a sign with 1-Parallel in ASL, the two extreme positions.

5.3 Comparison of 1-cl and 2-cl in DTS and ASL

The first event that we have picked out for comparison between the two languages is repeated in three episodes of the full cartoon. It shows Sylvester walking back and forth speculating about how to catch Tweety Bird. This is an atelic event without a goal. Given the focus on walking and lack of a goal, we would expect the DTS signers to use signs with a variant of 2-cl for these clips, and so they do. They present Sylvester walking back and forth from the same perspective as seen in the cartoon, i.e., seen from an observer’s viewpoint either from side to side or forward and backward, but in some cases, the signers have a facial expression indicating Sylvester’s concentration. All five DTS signers mention Sylvester’s walking back and forth in at least two narratives, and they do so one to three times per narrative. Out of the 14 scene descriptions, 13 include variants of 2-cl. One signer uses a sign with 2-cl and, after that, and in the same narrative, two signs with 1-cl, in the other two narratives she uses only signs with 2-cl.

In the ASL data, Sylvester’s walking back and forth gives rise to more variation between signers, but not within signers. Altogether, they describe 18 of these scenes with variants of either 1-cl (eleven scenes) or 2-cl (seven scenes), also from an observer’s viewpoint as Sylvester’s walking is seen in the cartoons. Four signers use only signs with 1-cl, and three only signs with 2-cl, and no one uses both signs with 1-cl and signs with 2-cl. In both languages, almost all variants are made with hand-internal movement (in ASL both 1-cl and 2-cl, in DTS only 2-cl), consistent with the fact that Sylvester here moves voluntarily (Section 6).

In the next set of scenes, the protagonists go to a specific goal, all in the horizontal plane: (1) Tweety Bird goes to get a bowling ball, (2) and (3) Sylvester goes outside twice to get two objects for a seesaw, (4) he goes to hide behind a corner, (5) a monkey joins him there, (6) he appears onstage from behind the corner, (7) he goes to specific places in the apartment to search for Tweety Bird, (8) he goes into an apartment to get some pieces of luggage, and (9) a bellboy goes to pick up a phone, altogether nine scenes. All the actions are telic. Altogether the 9 ASL signers contribute 39 scene descriptions of these events. Here, the ASL signers are more consistent than the DTS signers. Five ASL signers use only signs with variants of 1-cl to describe these scenes, and three use both variants of 1-cl and variants of 2-cl. One signer describes only one of the scenes and here uses a sign with a variant of 2-cl. In 34 descriptions, the ASL signers use signs with variants of 1-cl, in descriptions of three scenes they use signs with variants of 2-cl, and in descriptions of two scenes, one signer uses only signs with 1-Parallel.

The 5 DTS signers contribute 33 descriptions of the scenes and use 8 signs with 1-cl and 15 signs with variants of 2-cl. All four signers who describe the monkey’s approaching Sylvester use signs with 1-cl, as we would expect since Sylvester is the holder of the point of view here. In one case, the event involves boundary crossing, i.e., Sylvester entering the apartment. Here, three DTS signers use a variant of 2-cl, one ASL signer uses a sign with 1-cl, and all others who describe the scene find alternative solutions.

In two narratives Sylvester moves on a surface that makes motion difficult or awkward: he climbs (1) into a pipe and then (2) upwards on its inside, and he (3) walks, and later (4) runs and (5) jumps, on an electric wire above the street, altogether five events with focus on legs. The DTS signers here contribute 23 scene descriptions. They all use signs with some variant of 2-cl about Sylvester walking and running on the wire; three also use signs with 2-cl about his climbing into the pipe and about his climbing up inside the pipe. The three signers who mention his jumping on the wire also use signs with variants of 2-cl about jumping. Otherwise, they use signs with 1-Parallel, especially about his climbing into (horizontal index finger) and up inside (vertical index finger) the pipe. One signer uses signs with variants of 2-cl as well as one sign with 1-cl about his running on the wire; this is the same signer as the only one who used 1-cl about Sylvester’s atelic walking back and forth.

To sum up, the classifier uses descriptions of motion on the difficult ground in DTS:

• In 19 out of 23 scene descriptions (83%), the DTS signers used a sign with a variant of 2-cl.

• In 9 out of 23 scenes, the DTS signers used a sign with 1-Parallel.

• In 1 scene out of 23 scenes, a DTS signer used a sign with 1-cl.

The widespread use of signs with variants of 2-cl complies with the description of 2-cl as being used about difficult motion by means of the legs and feet in DTS (Section 5.1 and Engberg-Pedersen 1993).

The 9 ASL signers contribute altogether 33 scene descriptions. In 20 out of the 33 scene descriptions (62%), the signers use signs with variants of 2-cl. Specifically in the descriptions of Sylvester’s walking, running, and jumping on the electric wire, the ASL signers use signs with variants of 2-cl in 13 out of 17 scene descriptions (76%), and signs with variants of 1-cl in 8 scene descriptions (47%). Four of the nine signers use both signs with 1-cl and signs with 2-cl, four use only signs with variants of 2-cl, and one uses only signs with variants of 1-cl.

5.4 Discussion of the uses of 1-cl and 2-cl in the two languages

In spite of the obvious iconicity of 1-cl and 2-cl, we see some similarities and clear differences between their uses and thus their meanings in the two languages. 1-cl has variants with hand-internal movements in ASL, but not in DTS in the data. By contrast, there do not appear to be differences in form between the variants of 2-cl in the two languages: the index and middle fingers may be bent or extended, and in both cases, the fingers may change their relative position (Section 4).

The analysis of which classifier is used in descriptions of the groups of scenes reveals semantic differences. For the atelic events (Sylvester’s walking back and forth) the DTS signers use almost exclusively signs with 2-cl, the ASL signers sign with both classifiers. For the telic events (someone going somewhere) the ASL signers use primarily signs with 1-cl, the DTS signers sign with either one or the other classifier with a preference for 2-cl. Finally, for the events with difficult motion, the DTS signers are more unanimous in preferring signs with 2-cl than the ASL signers, who nevertheless use relatively more instances of 2-cl than for the other two types of events. Specifically for describing Sylvester’s walking, running, and jumping on the wire, the ASL signers describe 76% of the scenes by means of signs with 2-cl, but still also describe almost half (47%) of these scenes by means of signs with 1-cl. That is, one and the same signer may use a sign with 1-cl and a sign with 2-cl about the same scene (Figure 11).

Figure 11

ASL signer using a sign with 2-cl about Sylvester standing on the tram wire and a sign with 1-cl about Sylvester balancing on the wire (represented by the index hand).

Judged from these cartoon descriptions, there is a preference for 2-cl in DTS and a preference for 1-cl in ASL. But the choice between them is not arbitrary: atelic motion or motion on difficult ground draws the DTS signers toward 2-cl, with telic motion leaving room for more variation; telic motion draws the ASL signers towards 1-cl, whereas atelic motion and, to a higher degree, motion on difficult ground make them use also 2-cl. Moreover, the ASL signers appear somewhat more consistent within themselves than the DTS signers.

An explanation for the ASL preference for 1-cl and the DTS preference for 2-cl may be that 1-cl has variants in ASL, not in DTS (Section 4). The form variants in ASL indicate differences in meaning. For example, the ASL signer who only used signs with 1-cl to describe Sylvester’s behavior on the wire describes his walking awkwardly and trying to keep his balance by a rocking movement to the sides of the index finger of 1-cl. The alternating bending and extension of the outer joints of the finger of 1-cl in verbs about someone going somewhere probably indicates determinate walking (Section 6) and thus allows the signers to indicate this type of walking. By contrast, the index finger of 1-cl in DTS does not bend at any joint during the course of the sign. The only possible variation is a movement up and down of the whole hand to indicate determinate walking, a form not seen in the cartoon descriptions, however. Other data from DTS show how 1-cl is used to describe someone bumping into someone else and falling, where the articulator changes its orientation, but otherwise, the index hand of 1-cl is used with no variation from the rigid orientation perpendicular to the plane of motion.

Whether or not ASL and DTS inherited the classifiers 1-cl and 2-cl from a common ancestor or they are iconic inventions in historically separate situations, they have different meanings in the current states of the two languages, differences that are reflected in their use in signs about events with different characteristics. In this sense, the classifiers have unpredictable, lexicalized, meanings.

6.1 Volitionality, hand-internal movement, and handshape – statistical results

The following computational resources were used for statistical processing and analysis of the role of volition and its interaction with 1-cl and 2-cl: the R packages dplyr (Wickham et al. 2023), tidyverse (Wickham et al. 2019), readr (Wickham et al. 2024), esquisse (Meyer and Perrier 2024), report (Makowski et al. 2023), and ggplot2 (Wickham 2016). The Large Language Model ChatGPT (OpenAI 2024) was used to debug and simplify code throughout. The output revealed that the pairs of features demonstrating statistically significant correlations were similar between ASL and DTS, suggesting that both DTS and ASL share underlying patterns concerning hand-internal movement and volitionality (Table 2).

A statistically significant relationship between Volitionality and Handshape (1-cl and 2-cl) was not found in either ASL or DTS. As seen in Table 2, Fisher’s exact test confirmed a p-value of 0.6483 for the relationship between Volitionality and Handshape in ASL, and a p-value of 0.1098 for that of DTS. A statistically significant relationship was found between Volitionality and Hand-internal Movement and between Hand-internal Movement and Handshape. A Fisher’s exact test confirmed a ***p-value of 1.493 × 10⁻⁸ for the relationship between Volitionality and Hand-internal Movement in ASL, and a ***p-value of 0.004361 for that of DTS. For the relationship between Hand-internal Movement and Handshape, again using a Fisher’s exact test, a ***p-value of 4.496 × 10⁻⁷ was confirmed for ASL, and a ***p-value of 3.737 × 10⁻¹⁴ was confirmed for DTS (Table 2).

These statistical connections lend support to several points raised in this article. Hand-internal movement was observed at a much higher rate in volitional classifier predicates than in non-volitional ones. This provides strong statistical support for the semantic notion of volition being linked to the handshape-internal motion of a classifier predicate. Additionally, movement, in general, was observed at a significantly higher rate in 2-cl than 1-cl handshapes. This may be due to the increased mechanical possibilities in 2-cl vs 1-cl (e.g., stacking and spreading). The results are illustrated in Figure 13.

Figure 13

Percentage of volitional and non-volitional signs incorporating movement in (a) ASL and (b) DTS.

Figure 14 shows a statistical preference from signers of both languages, with DTS signers avoiding all hand-internal movement in classifier predicates with 1-cl.

Figure 14

Percentage of 1-cl- and 2-cl-type signs incorporating movement in (a) ASL and (b) DTS.

6.2 Discussion of volitionality, hand-internal movement, and handshape

The findings of this study offer insights into the relationship between volitionality, hand-internal movement, and classifier handshape in ASL and DTS, lending support to the notion that meaning can play an important role in determining classifier form.

First, the Canary Row narratives provide ample opportunity to produce an abundance of both volitional and non-volitional classifier predicates. Interestingly, the relationship between volitionality and handshape did not show statistical significance in either language.

Second, despite the absence of an apparent genetic relationship between the two languages, signers of both languages exhibit parallel tendencies in how volitionality, hand-internal movement, and handshape interact – that is, an increase in hand-internal movement associated with volitionality (for a discussion of the relationship between control of the motion and hand-internal movement in what corresponds to 2-cl in RSL, refer to Kimmelman et al. 2020). This suggests that there may be underlying modality-level preferences influencing the organization of classifier systems, as well as common cognitive or linguistic strategies for encoding actions and intentions within their classifier systems. These results reinforce the idea that it is not only iconicity but volitionality, along with other semantic notions, that play a meaningful role in how actions are depicted through classifiers.

Third, a similar statistically significant relationship was found between hand-internal movement and classifier handshape. The increased number of selected fingers in the 2-handshape compared to the 1-handshape supports a greater range of possible configurations for those fingers. It is likely that this mechanical reality would lead to a similar link between movement and selected fingers in other sign languages. ASL signers were willing to produce hand-internal movement in 1-cl when it suited their communicative needs. This distinction between the two languages suggests a possible language-specific setting, which warrants further exploration in additional sign languages.

8.1 Serial verbs of motion

Typologically, Aikhenvald (2018) defines serial verb constructions as constructions with two or more verbs with no overt marker of dependency; the verbs together conceptualize a single event. They tend to share subjects and objects. The verbs cannot be questioned or negated separately, and each verb can function as the main verb of a clause by itself. Generally, a serial verb construction falls within one clausal intonation contour.

One type of sequence of same-event verbs was first described by Supalla (1990) for ASL as ‘serial verbs of motion’. Supalla distinguishes manner of locomotion from manner of motion along the path. Manner of locomotion may be, for instance, limping, running, or coasting a toy wagon with one leg. He illustrates these different manners of locomotion by means of verbs with a BPCL for legs, i.e., two index hands. If a path is added as in the representation of, for instance, limping in a circle, the path will be described by means of a separate sign, the drawing of a circle in space. The two signs – one representing the limping manner of locomotion, the other the path – constitute the serial verb construction. According to Supalla, they occur in the order manner followed by path with no intervening signs.

Since Supalla’s description of manner and path verbs of motion, several researchers have used his framework for similar constructions in other sign languages. Slobin and Hoiting (1994) found serial manner and path verbs in Sign Language of the Netherlands (NGT) and linked the construction with Talmy’s (1985) theory about verb-framed and satellite-framed languages and with Aske’s (1989) theory about two types of path phrases: ‘locative’ path phrases and ‘telic’ path phrases. Slobin and Hoiting (1994) find that the significant difference between the two types of path phrases is whether the path involves boundary crossing or not. If a path involves the crossing of a boundary (‘enter’, ‘exit’, and ‘cross’), it is telic, and the verb describing the path cannot also include manner description in verb-framed languages such as NGT, they claim.

In an article originally from 1996, Bos (2016) compares a construction from NGT, which she calls double-verb construction, with serial-verb constructions in spoken languages. Although these constructions may include the verb go, they are not serial verbs of motion. Couvee and Pfau (2018) describe go in these constructions as a generic verb that does not give detailed information about a path and does not include a w/e classifier.

Benedicto et al. (2008) compare sequences of classifier verbs in ASL, Catalan Sign Language (LSC), and Argentinian Sign Language (LSA) and find that LSA follows the order constraint of ASL, i.e., manner before path, but LSC accepts both orders. Moreover, both LSA and LSC have the order manner–path–manner, and LSC also the order path–manner–path. Clark (2017) describes serial verb constructions with the order manner-path in Sivia Sign Language (LSSiv), a sign language of Peru. Serial verb constructions with the order manner-path are described as Motion-Directional SVCs by Lau (2010) for HKSL. In her data from Korean Sign Language, Hong (2003) finds that manner verbs with a BPCL precede other verbs in only 57% of the serial verb sequences. But she also finds sequences of manner-path-manner verbs as in LSA and LSC and sequences of up to seven classifier verbs. Nyst (2007) describes serial-verb constructions in the village sign language Adamorobe Sign Language (AdaSL) of Ghana, a sign language in which w/e classifiers are rare. In retellings of the Canary Row cartoons, the signers did not use a single example of a sign with 2-cl. In descriptions of Sylvester walking back and forth, the AdaSL signers used a sign with a BPCL for the whole body, leaning the body alternately to the left and the right. In serial verb constructions, they used so-called directionals with very general meanings after manner verbs with BPCLs or handling classifiers. Thus, there may be a preference for the order manner before path in sign languages, but it is not universal. (See also Davis and Brentari, this volume.)

8.2 Serial verbs of motion in DTS and ASL

In the analysis of serial verb constructions of motion verbs in DTS and ASL, we have chosen to categorize verbs with one of the w/e classifiers 1-cl, 2-cl, or 1-Parallel as path verbs, and verbs with BPCLs as manner verbs. The criterion for the distinction is that verbs of motion with w/e classifiers can show details of a path such as whether it is straight or meandering, while signs with BPCLs cannot describe details of a path. Signs with one type of BPCLs, i.e., legs classifiers (two index hands with the fingers representing an anthropomorphic being’s legs), can in fact indicate at least some details of the path in DTS: one of the DTS signers describes Sylvester’s being thrown out of a building by means of the legs classifier, one fingertip upwards and the other to the side, the hands moving in an arc from the signer’s left towards her right (Figure 17). At the top of the arc outside her forehead, she changes the classifier to 2-cl in the dominant hand.

Figure 17

A DTS signer’s description of Sylvester being thrown out of a building: (a) the sign kaste ‘throw’ based on a sign with a handling classifier, (b–c) a sign with the BPCL for legs, and (d–e) a sign with the w/e classifier 2-cl.

Some signs with handling classifiers may also show path to some extent. Before the description of how Sylvester flies through the air in Figure 17, the DTS signer uses the sign kaste ‘throw’, which is based on a sign with a handling classifier: the hands open from fists to open hands as they bend in the wrists. The signer places her hands at her left side and opens them in the rightwards direction. The orientation of the hands and the direction of their movement comply with the direction of movement of the sign with the BPCL (Figure 17) following kaste. In data elicited to test minimal pairs from RSL, Kimmelman et al. (2020) found that Russian signers might use a verb with a handling classifier even to describe complex paths. In our data, we do not see any examples of complex path descriptions in verbs with handling classifiers, but in another episode from the Canary Row cartoon, Sylvester swings on a rope from one building to another. Here, two ASL signers use a verb with a handling classifier (holding the rope) and an arc movement out from their body (Figure 18), one of them also a verb with the w/e classifier 2-cl. The other seven ASL signers use either a similar verb with a handling classifier and a short forward movement and then switch to a verb with a w/e classifier (1-cl or 2-cl), or they use only a verb with 1-cl or 2-cl. Of the DTS signers, one uses a verb with a handling classifier and a full arc movement followed by a verb with 2-cl and the same arc movement, the other four signers all start with a sign with the handling classifier and a very short movement and then switch to a verb with 2-cl and a full arc movement. Thus, it seems that, at least in these data, most signers hesitate to use a verb with the handling classifier and a full path movement, although not all. However, several signers use a verb with a handling classifier and a full path movement to describe how Sylvester floats down to the ground while holding the bird in his hands (Figure 19).

Figure 18

An ASL signer describes how Sylvester swings on a rope from one building to another.

Figure 19

A verb with 2-cl followed by the beginning and end positions of a verb with a handling classifier about Sylvester’s floating down while holding the bird.

Despite these complications, the main distinction here is between verbs with w/e classifiers, which may describe details of a path, and verbs with BPCL (especially the legs classifier) or with handling classifiers, which primarily describe manners of motion.

8.2.1 Serial verbs of motion in DTS

All the lexical verbs used by the DTS signers can indicate the direction of motion, but only følge ‘follow’ can indicate details about the path, e.g., ‘follow in a meandering pattern’. Følge is based on a sign with two w/e classifiers (expressed by A handshapes). Løbe(S) and løbe(5) ‘run’ are two-handed and based on signs with BPCLs, namely hands, and they can indicate the direction of motion by the orientation of the signer’s head, body, and hands. Ind ‘into, out of’ and flygte ‘escape’ are identical in form: the dominant index hand moves underneath a flat hand, both hands with the same palm orientation. They can indicate the direction of motion independently of the signer’s orientation, but they cannot indicate details of the path.

Making the dependence on the signer’s orientation for indicating the direction of motion a main criterion, we get two categories of predicates of motion. We base the categorization of the lexical signs on how they are used in the data.

(2)	Categories of lexical signs regarding manner and path in DTS
	A. Manner of motion verbs: signs with BPCLs, including handling classifiers, plus the signs gynge ‘swing on a swing’, løbe(S) ‘run’, løbe(5), rulle ‘roll’, and kaste ‘throw’
	B. Path of motion verbs: signs with w/e classifiers and the lexical signs ind ‘into, out of’, følge ‘follow’, flygte ‘escape’, tage-afsted ‘go off’, and ud ‘out of’

In the DTS data, there are altogether 60 sequences of two or more verbs that fulfill most of Aikhenvald’s (2018) criteria for serial verb constructions and include at least one token of a sign with 1-cl, 2-cl, or 1-Parallel. They occur within one clausal intonation contour, and they are not linked by any marker of dependency. They share one argument. It is the only argument of a verb with a w/e classifier or a BPCL, and, with a verb with a handling classifier, it may be either the agent argument as in the example in Figure 19 (example 3) or the patient argument as in the example in Figure 17 where a DTS signer describes Sylvester being thrown out by means of kaste ‘throw’ followed by a verb with a BPCL and a verb with a w/e classifier (see also the discussion of argument structure with classifier verbs in Kimmelman et al. 2020). Furthermore, in accordance with Aikhenvald’s (2018) criteria, the verbs conceptualize a single event by describing different aspects of it. All 60 tokens of verb sequences occur in positive clauses that function as statements, but based on the closed class of data, we cannot say whether the verbs might be questioned or negated individually.

Aikhenvald’s (2018) definition also requires each of the verbs to be able to occur as the sole verb of a clause. This criterion is not unproblematic for the DTS data of this study. There are altogether 221 solitary verbs in the data (Table 3). Almost all the solitary signs are verbs with w/e classifiers, which may describe paths, or lexical verbs such as the sign løbe(5) ‘run’, which is based on a BPCL for hands, and the sign gynge ‘swing on a swing’, which is based on a handling classifier. A few manner verbs with the legs classifier or with a handling classifier occur as the sole verb of a clause in the DTS data if they include a path.

Since the signers in these data use many same-event verbs, the lack of clear examples of signs with BPCLs and handling classifiers as the sole verbs of clauses may be a consequence of the narrative genre. It is worthwhile to find out in future studies to what extent verbs with classifiers that only describe manner with no path information can be the only verb of a full sentence in DTS. If they cannot, the serial verb constructions with manner verbs in DTS should be discussed in the light of symmetry in serial verb constructions (cf. Costello 2016, Aikhenvald 2018). For now, we will accept, as serial verb constructions, sequences of same-event verbs of motion that occur within one clausal intonation contour with no intervening constituents.

Out of the 60 such constructions, 33 sequences consist of a manner verb followed by a path verb with one of the classifiers 1-cl, 2-cl, or 1-Parallel as in (3) with a verb describing someone walking on a tightrope, followed by a verb with a variant of 2-cl (Figure 20).

(3)	DTS AA, Episode 8, ‘balance on wire’[8]
	DH: hands(B)+at-side-to-keep-balance 2-cl+straight-forwards/
	NDH: hands(B)+at-side-to-keep-balance Index+as-tram-wire
	“He walks on the tram wire as if on a tightrope.”

Figure 20

The two verbs of example (3).

Toward the end of the manner-path construction, the signer establishes eye contact with the addressee, then she looks back down at her hands and repeats the sequence in a form of clause doubling (Section 8.3). Thus the example agrees with the criterion mentioned by Aikhenvald (2018) that serial verb constructions may be repeated in the same sequence.

In some instances of manner-path constructions, the signers use either more than one path verb after the manner verb or more than one manner verb before the path verb. For instance, one signer uses a variant of the sign løbe(5) ‘run’ followed by a sequence with first a verb with 2-cl describing how Sylvester runs down a spiral staircase, then a verb with 1-Parallel describing the same descent with a straight downward movement, and lastly a verb with 1-Parallel in a straight movement forwards about his running across the street. Here, all the path verbs do not strictly speaking describe the same motion event, but contribute parts of the total event, and the manner verb løbe(5) ‘run’ covers them all.

In the manner-path sequences, the path verb may be followed by a locative phrase about a goal, which could not follow directly after the manner verb. An example of a locative phrase after a path verb is the manner verb rulle ‘roll’ followed by the path verb 1-Parallel+arching-forwards and then the phrase bowling house ix _forwards, i.e., ‘rolls into a bowling hall’, which is the goal.

Slobin and Hoiting (1994) point to the importance of boundary-crossing in the motion scenario: if the path involves boundary-crossing, the verb describing the path cannot include a manner description in verb-framed languages, they claim. This generalization can be expanded to another case of telic path descriptions (cf. Aske 1989): if the path involves arrival at a point, a path verb is needed. In DTS mainly path verbs can describe arrival somewhere by means of an abrupt stop of the hand’s movement. That can be seen in examples where the signers describe how Sylvester climbs a pipe to Tweety Bird’s window: a manner verb with a BPCL (hands describing climbing) followed by a path verb with 1-Parallel or a w/e classifier (2-cl), the latter ending with an emphatic stop that indicates arrival. In cases of boundary-crossing, the signers use both the lexical path verbs ind ‘into, out of’ and out ‘out of’ and signs with a w/e classifier and a straight movement and especially the fairly neutral 1-Parallel or 2-cl, i.e., path verbs with no further specifications of manner or details of the path. By contrast, the signers do combine atelic manner verbs with locative phrases about the location, as, for instance, in a sequence with the manner verb meaning ‘climb’ followed by pipe and then a repetition of the manner verb: ‘Sylvester climbs the pipe’ (Section 8.3).

Besides the 33 sequences with manner before path, there are 9 examples with the path verb before the manner verb. Six of these are atelic, which makes it possible to end with a manner verb. Two of the telic path-manner constructions are descriptions of falls with a path verb with 2-cl followed by a manner verb with flat hands lifted to the sides and a forward movement of the body (Figure 21) (similar signs in ASL are described as gestures by Emmorey 1999). Here telecity is expressed in the path verbs and implied by the manner verb, but the manner verb is not followed by a locative phrase. The last telic path-manner construction describes Sylvester’s swinging into a wall, again with a manner verb with the BPCL for two hands. The path verb with 2-cl is clearly telic as it includes the NDH about the ground and there is contact between the articulator of 2-cl and the NDH; the manner verb is not followed by a locative phrase, but by a new path verb with 2-cl about Sylvester’s sliding down the wall. The last verb is not a same-event verb and is thus excluded from the analysis here. But it occurs within the same prosodic unit as the other two verbs and may be part of a different type of serial verb construction with verbs that describe temporal sequences.

Figure 21

A path-manner sequence: (a) falling described by a verb with 2-cl in the DH (and the NDH preparing for the following sign), and (b) a verb with two BPCLs about hands in a description of falling forwards.

Benedicto et al. (2008) found that LSC has both orders manner–path and path–manner and also both sequences of three verbs: manner–path–manner and path–manner–path. In the DTS data, there are no examples of manner–path–manner involving one of the classifiers 1-cl, 2-cl, or 1-Parallel. But there are 13 examples of path–manner–path sequences and 5 sequences with more than 3 verbs (up to 6) ending in either a path or a manner verb. An example of a path–manner–path sequence is a signer’s description of Sylvester’s climbing a pipe: the signer uses first a verb with 2-cl and an upward movement of the articulator, then a verb meaning ‘climb’ with a BPCL, and at the end, a less elaborate version of the first path verb with 2-cl.

In sum, we find that many of the sequences of motion verbs in the DTS data can be categorized as serial verb constructions.^[9] They tend to order the manner verbs before the path verbs, but we also find the orders path-manner and path-manner-path. In these data, there are no manner-path-manner sequences with the path verb including either 1-cl, 2-cl, or 1-parallel. But there are examples of longer sequences ending with either a manner or a path verb. The analysis shows that verbs with BPCLs may code path information in DTS, and verbs with w/e classifiers are preferred if the motion involves boundary-crossing or arrival somewhere.

8.2.2 Serial verbs of motion in ASL

As in the DTS data, we here include not only signs with the three w/e classifiers, but also some lexical verbs of motion and location, namely, chase, climb, escape, inside, into, roll, out, chase (two A handshapes), run(bC) (two bC hands), run(S) (two fists), and throw, but only if, somewhere in the data, they are same-event verbs with a verb with one of the classifiers 1-cl, 2-cl, or 1-Parallel. We base the categorization of the lexical signs on how they are used in the data (4).

(4)	Categories of lexical signs regarding manner and path in ASL
	A. Manner of motion verbs: signs with BPCLs, including handling classifiers, and the lexical signs climb, inside, run(S), roll, and throw
	B. Path of motion verbs: signs with w/e classifiers and the lexical signs chase, escape, into, off, out, and run(bC)

In the ASL data as in the DTS data, we see that verbs making up their own clause are mainly of the w/e type or lexical path verbs. But, in the ASL data, some lexical manner verbs (e.g., roll and climb) as well as manner signs with classifiers also occur as the sole verb of the clause (cf. Aikhenvald’s 2018 definition of serial verb constructions). For instance, one signer describes Sylvester’s running away with his catch by means of a manner verb with the BPCL legs classifier: the signer’s body represents Sylvester, and the index fingers making alternating circles at the signer’s sides represent his legs, there is no path verb.

We know from Supalla (1990) that ASL has serial verb constructions of the manner–path type. That analysis is confirmed by the data in this study. There are altogether 33 manner–path sequences in the data, including a few with either two manner verbs in the first position or two path verbs in the second position. By contrast, there are only two path–manner constructions. An example of a manner-path sequence is seen in (6) with a description of how Sylvester climbs a pipe and gets to Tweety Bird’s windowsill (Figure 22).

(5)	ASL DR, Episode 2, ‘Sylvester climbs up pipe’
	DH: climb 2-cl+straight-up-on-NDH 2-cl+straight-down-onto-NDH /
	NDH: climb flat-hand-vertical+hold‐‐‐‐flat-hand-horizontal+hold‐‐‐‐‐‐
	“(Sylvester) climbs (the pipe) and places himself on (Tweety Bird’s windowsill).”

Figure 22

The three verbs of example (5). The first two are same-event verbs and constitute a manner–path sequence.

The manner verb climb is made with two claw-like hands. The signer places alternately one hand a little above the other, which gives the impression that Sylvester is climbing upward. But the verb does not show details of the path. By contrast, the following verb with 2-cl can show such details independently of the signer’s body, in the example the path is a straight line upward on a vertical surface (see the second picture in Figure 22) and onto the NDH representing the windowsill. The first two verbs are same-event verbs and constitute a serial verb construction, but the last verb adds further details about the path and does not belong to the serial verb construction.

The proportion between manner–path sequences and path–manner sequences is more in favor of manner–path sequences in ASL than in DTS. In DTS there were nine path–manner sequences to 33 manner-path sequences. In the ASL data, there are two path–manner examples to 33 manner-path sequences. The two path–manner sequences are made by two different ASL signers. One of the signers describes Sylvester’s climbing a pole, this time in the opposite order of example (5): first a path verb with 2-cl (without arrival) and then climb. The other signer describes Tweety Bird’s going offstage by means of a path sign with 1-cl followed by a manner sign with a BPCL representing wings. In the latter sign, the signer orientates his body to the left. Both sequences are atelic. There are also a few instances of three verbs: both manner–path–manner and path–manner–path sequences, with the second verb a repetition of the first (Section 8.3).

Although there are differences in the proportions of the types of sequences in the two languages, the languages are very similar with respect to serial verb constructions of motion and location. They both prefer manner-path constructions to path–manner constructions, possibly here reflecting a universal trend towards the manner-before-path order (cf. Slobin and Hoiting 1994). Another explanation may be that the occurrence of final path signs makes it possible to express telic motion, either boundary crossing or arrival somewhere. Manner signs in general express atelic motion unless they are accompanied by a path sign. In sequences of three verbs both languages prefer path–manner–path series, which again makes it possible to end with a sign that can express boundary crossing and arrival and can be followed by a locative phrase about the boundary or the goal. Thus, rather than describing the orders as influenced by a universal preference for manner before path, we might describe the order as a preference for path last, which reflects order iconicity: the boundary crossing or arrival comes toward or at the end of the motion. The two examples of path–manner in ASL are atelic, and of the nine examples in DTS, six are atelic. The last three path-manner sequences in DTS are made telic by the path verb preceding the manner verb, and they are not followed by any locative phrase. Moreover, the manner verb with raised flat hands about falling seems to imply telecity.

The examples of manner–path–manner and path–manner–path constructions that are described here as serial verb constructions are symptomatic of another shared feature of many sign languages, namely, the trend towards repetitions, which contributes to the large proportion of same-event verbs in the data from both languages. We will turn to examples of these constructions in the next section on verb doubling.

8.3 Verb doubling

Fischer and Janis (1990) described a type of repetition of verbs in ASL as verb sandwiches, i.e., constructions where two tokens of the same lexical verb share a subject, the last token has different inflections than the first one, for instance, continuous aspect or plural, and between the two tokens the object and/or an adjunct can occur. Fischer and Janis suggest that the reason for the construction is “a general constraint on the amount of information that can be carried by a single verb” (1990, 285), and, in that respect, they see a parallel between verb sandwiches and serial verb constructions: serial verb constructions avoid informational heaviness by splitting up manner and path on two verbs, verb sandwiches avoid heaviness by splitting up verb plus object or adjunct and verb plus inflections – or rather verb plus different inflections since both verbs may be inflected, as pointed out in a footnote by Fischer and Janis (1990, 282).

Liddell (2003, 64) provides further support for the explanation by pointing out that a verb inflected for aspect cannot occur with an overt object in ASL. But there is a general tendency to repeat signs in ASL and other sign languages even when the two versions of a sign do not contribute different information, as we already saw in the path-manner-path and manner-path-manner sequences of serial verb constructions where the first and the last verb may be identical (Section 8.2). Moreover, it is well-known that many sign languages repeat question words and/or pronouns (e.g., RSL, NGT [Kimmelman 2012, 2019], and Flemish Sign Language [VGT] [Vermeerbergen and Engberg-Pedersen 2024]). Fischer and Janis’ (1990) own article includes several examples of repetitions of identical forms, e.g., repetition of a negative modal and a pronoun in me won’t use next time, won’t, me ‘she swears not to use it again’, and repetition of finish in h-a-r-r-i-e-t finish wash-clothing finish… ‘When Harriet is done with the wash, […]’

Pinsonneault (1994) found as many repetitions of the same form of the verb as verb sandwiches proper in her data from Langue des signes québécoise (LSQ); she uses the term verb echo as a superordinate term for both types. Analyzing elicited data with single sentences from Croatian Sign Language, Milković et al. (2006) distinguish verb sandwiches and what they call verb echoes. They characterize verb sandwiches as constructions where the second verb is informationally richer than the first verb. In verb echo constructions, the two verbs are identical in form. In a study on word order in RSL, Kimmelman (2019) uses verb doubling as the superordinate term and distinguishes verb sandwiches and verb echoes in the same way as Milković et al. (2006). Kimmelman finds both verb echoes and verb sandwiches in RSL. In the verb sandwiches the second verb is always the one that is more marked in his data. Proske (2020) also finds both instances of identical verbs and instances where one verb is more marked than the other in German Sign Language (DGS); in contrast to Kimmelman’s results, the marked one is not necessarily the last one in DGS. In an appendix to a study on word order in Libras (Brazilian Sign Language), Napoli et al. (2017) mention one more instance of doubling: SVOV constructions where “the two Vs in the construction share (part of) their sense” (Napoli et al. 2017, 664). An example from Libras would be a verb meaning ‘build’ followed by a verb meaning ‘build with bricks’. They call the Fischer-Janis type of verb sandwich type-same and the last one type-different.

Koenders (2024) raises the possibility that the prominence of doubling of heads and phrases of all sorts in HKSL indicates that HKSL is a focus-prominent language. Like Kimmelman (2019) she distinguishes clause-internal doubling from other types of doubling such as index doubling. What is relevant in the context of this article is primarily clause-internal doubling of verbs. Clause-internal doubling is the most frequent type in Koender’s data from HKSL. The doubled constituent “must refer to the same referent or event/action/state and must be essentially the same lexical word, though either of the two copies may show morphological/phonetic modification” (Koenders 2024, 141). However, modifications are infrequent in her data, and as in DGS, the modified constituent may be either first or last. In by far most cases of clause-internal doubling the pattern is XYX. All sorts of elements may be doubled clause-internally in HKSL, but the most frequent group of elements are predicative adjectives, verbs, and verb phrases, i.e., constituents in predicate function like the ones we focus on in this article.

Kimmelman (2019) discusses clause–internal doubling in RSL and NGT, where also any type of constituent can be doubled, and as in HKSL, doubling can be continuous (a XX pattern) or discontinuous with something in between (a XYX pattern). The most frequent type of doubling in RSL is verb doubling, and in NGT it is clause doubling. In both languages the most frequent pattern is doubling with identical forms in the two positions except for possible differences in phonetic weight; the first occurrence is usually stronger than the copy. Kimmelman (2019) points out that since the two forms are morphologically identical in most cases, Fischer and Janis’ (1990) explanation for verb sandwiches – that the information is spread out over several signs – does not work for most of the cases of doubling in the data from RSL and NGT. Instead, Kimmelman suggests that the main function of doubling the XYX type is to foreground the doubled element and background the element in between.

As this short overview makes clear, verb doubling of several types may be part of a larger pattern of constituent repetition in sign languages. Moreover, doubling is seen in sign languages that are not historically related such as ASL and DGS. In what follows, we will examine how signs with the classifiers 1-cl, 2-cl, and 1-Parallel behave in such constructions; i.e., we include all examples of doubling of verbs with 1-cl, 2-cl, or 1-Parallel in the data from first DTS and then ASL. The constructions all include another constituent than a verb of motion or location between the two same-event verbs, i.e., examples of path–manner–path and manner–path–manner sequences are excluded here although they could also be seen as examples of verb doubling.

8.3.1 Verb doubling in DTS

The DTS signers use altogether fifty-three examples of verb doubling with signs with the same classifier – either 1-cl, 2-cl, or 1-Parallel – and the same movement on each side of a constituent (between eight and sixteen examples per signer). An example is seen in (6), which follows a description of how Sylvester flies up from a seesaw.

(6)	DTS HE, Episode 6, ‘catapult’
	DH: …2-cl+arc+loc	window board 2-cl+arc+loc /
	NDH: Flat-surface+hold	window board‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐
	“and lands on the windowsill.”

The prosody with a clear head nod at the end shows that the sentence finishes after the second version of the verb. There is a smaller nod on the first token of 2-cl+arc+loc; it is simultaneous with the verb and underlines +loc about getting to a stop. The constituent window board is the goal. In (6), the two versions of the verb are not totally identical: in the first combination of the dominant and the NDH, the handshape of the NDH represents the wall that Sylvester was passing on his way up from the catapult, i.e., a vertical wall, in the second combination, the handshape of the NDH represents the horizontal windowsill.

In (7), the middle constituent think describes the topical argument’s simultaneous activity.

(7)	DTS AA, Episode 3, ‘Sylvester climbs up pipe’
	DH: 2-cl+straight-back-and-forth think 2-cl+straight-back-and-forth /
	NDH: …Flat-surface+hold‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐
	“(Sylvester) goes back and forth thinking (about how to catch Tweety Bird).”

Based on a comparable corpus of retellings of the Canary Row cartoon, Kimmelman (2019, 190–1) describes similar examples from RSL and NGT as cases of clause doubling. Whether such examples are instances of clause-internal verb doubling or clause-doubling depends on which features are defined as constituting clause boundaries. In (7), simultaneously with the verbs with classifiers, the signer looks at her hands, and while signing think she looks up left in a direction that represents Tweety Bird’s window. But these changes in gaze direction are motivated by the content of the verbs and cannot be seen as arguments for separating the three verbs into three clauses. Moreover, the NDH, which represents the ground, was established before the sequence and is maintained when the three verb signs are made. We are inclined to see this as a case of clause-internal doubling.

The sequences may include three identical verbs; in one case, the signer mentions a simultaneous activity (think) between the first and the second version of a verb with 2-cl, and between the second and the third version, she mentions the location (ix corner ‘at the corner’). In the data, the location is particularly frequent between the two verbs as in example (6), but also the topical argument, a verb about a simultaneous activity as in example (7), a clause about the content of the referent’s thinking, and a clause about the reason why the event takes place (with an explanation of fifteen signs) occur between the two verbs. When there are clearly clause boundaries between the two instances of the verb, the doubling is of the clause-doubling type. A reason for the repetition of the sign in such constructions may be that it allows the signer to get back to talking about a sequence of events involving a specific referent in the continuation after the second verb.

The above examples are all of the echo type in that the second verb is made with the same classifier and the same movement as the first verb. There are no examples of verb sandwiches with different aspectual inflections on the two verbs in the data, but there are a few instances where the second verb repeats the movement of the first verb, but then stops, either abruptly to indicate that the protagonist has stopped the motion, or the movement of the hand stops as the focus shifts to the signer’s facial expression, indicating that the protagonist realizes something. Here the event described by the last verb is backgrounded, and the protagonist’s emotions as expressed by the signer’s face are foregrounded (cf. Engberg-Pedersen 2011).

In the DTS data, there are altogether eight examples where the two verbs describe the same event with a constituent in between, but with verbs that are not identical. One or both verbs include either 1-cl, 2-cl, or 1-Parallel, but the other may be a lexical verb of motion or location or a motion verb with a handling classifier. We will use semantic doubling about this phenomenon as the verbs share the semantic feature motion, but otherwise differ in meaning.

In (8) the first verb is a manner verb, climb, formed with a BPCL, the second verb is a path verb with 1-Parallel (Figure 23).

(8)	DTS HE, Episode 8, ‘electric wires’
	DH:	climb pipe pipe 1-Parallel+straight-up	line up …
	NDH:	climb pipe pipe point-buoy‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐	line up
	“He climbs a pipe up to wires […]”

Figure 23

The two verbs of example (8): climb and the verb with 1-Parallel, first position and a later position.

The path verb allows the signer to go on to describe the wires for the tram as the goal. In between the two verbs, the signer mentions the location of the climbing, the pipe. Thus, (8) is a construction with a manner–path sequence with the location inserted between the two verbs. We have not classified this as a serial verb construction although such constructions may include a constituent between the two verbs. It is indeed an area where semantic doubling overlaps with serial motion verbs. It is clearly a manner-path sequence. The example also shows that the atelic manner verb climb can combine with a location constituent in contrast to the goal line _up in the same example which occurs after the path verb 1+Parallel+straight-up.

In (9) the signer uses two path verbs, one lexical (ind) before the noun indicating the location and one with 2-cl after the noun.^[10]

(9)	DTS HO, Episode 1, ‘dogs and cats not allowed’
	DH:	ind door 2-cl+straight-forwards /
	NDH:	ind door‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐
	“He enters through the door.”

Signers may also vary between a path verb with 1-Parallel and one with 2-cl and between one with 2-cl and one with 1-cl. Most of the sequences of semantic doubling are made with signs with 2-cl and 1-Parallel. Thus, there is no indication that the construction semantic doubling in itself encourages either one or the other classifier. All examples but one have locatives in the middle position; the one exception is a verb about a simultaneous activity (think).

Whatever the explanation for the phenomenon, the existence of verb doubling of the echo type and the semantic type contributes to the high number of same-event verbs in signed discourse.