2.1 Simultaneity and sequentiality in sign languages

A large body of research on sign languages has documented how form-meaning pairings are packaged into words using more simultaneous layering in signed languages than in spoken languages (Meier 1993; 2002; 2012; Brentari 2002; Aronoff et al. 2005; Morgan 2014). An explanation frequently proposed for this tendency is modality: a simultaneous signal may be easier to parse when the co-occurring morphemes are processed by the visual rather than auditory modality (Meier 2002; 2012 and references therein). However, constraints on simultaneity can also be shaped by phonology, semantics, cognition, and linguistic structure. For excellent current summaries of simultaneity and sequentiality in sign languages see Vermeerbergen et al. (2007) and Loos et al. (2022).

Multiple independent manual structures (e.g., movement, location, handshape) can be produced simultaneously, and indeed, signers and silent gesturers have been found to produce high levels of simultaneously layered information to express complex events (Meier 1993; 2002; 2012; Goldin-Meadow et al. 1996; Brentari 2002; Aronoff et al. 2005; Morgan 2014; Slonimska et al. 2020; 2022). This suggests that the visual/manual channel facilitates the use of simultaneous structures. This type of simultaneity is not as likely in spoken language, in part because articulators in the oral modality are less independent than articulators in the manual modality (Brentari 1998; 2002; 2019; van der Hulst & van der Kooij 2005). Silent gesturers have been found to combine up to 5 different movement components within a single complex gesture; however, on average, signers combine more components, use redundancy more frequently in those simultaneous structures, and are more adept at coordinating simultaneous components within a single expression than silent gesturers (Slonimska et al. 2020; 2022). In Figure 1, we see a structure with five simultaneous structures from American Sign Language that means two-frail-humans move-forward-carefully-side-by side: the meaning ‘two’ is represented by the two hands, ‘people’ by the index fingers, ‘frail’ by the two bent knuckles, ‘move-forward’ by the direction of movement away from the signer, and ‘carefully’ by the pressed lips.

Figure 1

A predicate structure from American Sign Language (ASL) that means two-frail-humans+move- forward+carefully+side-by-side containing five meaningful simultaneous components—’two’ is represented by the hands, ‘people’ by the two index fingers, ‘frail’ by the two bent knuckles, ‘move-forward’ by the direction of movement away from the signer, and ‘carefully’ by the pressed lips. [Reprinted with permission from Brentari, 1998].

A white man with short dark hair wearing a long-sleeved button-down shirt models a predicate structure from American Sign Language (ASL) that means two-frail-humans+move-forward+carefully+side-by-side containing five meaningful simultaneous components—'two’ is represented by the use of both hands, ‘people’ by the two index fingers, ‘frail’ by the two bent knuckles, ‘move-forward’ by the direction of movement away from the signer, and ‘carefully’ by the pressed lips. The image shows two time points of the sign as it is produced, with both hands moving forward in space.

Sign language morphology can also appear sequentially, however, and a sign language may tend toward sequential structures as it undergoes certain changes. There are numerous types of sequential structures in sign languages, and we provide a few examples here. Fischer and Janis (1990) analyzed sequential morphology in ‘verb sandwiches’ in ASL, separating a bare form of a verb from the a copy of the same verb with aspectual morphology in a single clause; see (5) from Irish Sign Language (Leeson 1996). Aronoff et al. (2005) suggest that simultaneous morphology is the initial state in an emerging sign language, and that sequential morphology requires time to emerge. Results from the current study suggest that time is not the only factor.

Some two-handed reciprocal forms in German Sign Language (DGS) are also sequential (Pfau & Steinbach 2003(6)). The verb direction moves from the first argument to the second in V1, and then from the second argument back to the first in V2; in this version of the sign the two hands move together in the same direction at the same time.

As a third example of sequential structure in sign language, (7) presents an example of a resultative structure in Hong Kong Sign Language (HKSL, Lau 2012) involving the spread of both dominant hand (H1) and the non-dominant hand (H2)—that is, both hands are held in place across subsequent signs.

Even events that occur simultaneously can be pulled apart and represented sequentially within a predicate, as shown in (5) and below in (8). In British Sign Language (Morgan et al. 2002), we see sequential structures representing two perspectives. Some 2-participant events require the signer to locate the two referents in signing space through spatial indexing, and then articulate the main verb (V) from two shifting perspectives. V1 is produced with the perspective of the signer as the subject, and V2 is produced with the signer as the object. These structures are often used when depicting an action that takes place on a specific body part.

Forms such as these are also observed in LSN (Senghas et al. 1997; Flaherty 2014). Senghas et al. (1997) found that, when both the grammatical subject and object are animate, an MVP is often required in the clause and SVPs are rare. Senghas et al. (1997) has also documented a change in word order of the MVPs over LSN’s first two decades of emergence, from an NVNV structure (e.g., man push woman be-pushed), to an NNVV structure (e.g., man woman push be-pushed).¹

Note that there are different types of redundancy in each of the examples (5)–(8); The verb stems are the same in V1 and V2 in (5), (6) and (8), and in (7) the articulation of the V1 is held across V2 so that both hands are in the signing space at the same time.

In the current study, we examine variation in the sign vs. sign+sign structures associated with meanings of agency (agent/no-agent) and number (singular/plural) in iconic signs that are found in most known sign languages––classifier constructions (sometimes referred to as “depicting constructions”; see Supalla 1982; Zwitserlood 2012, and references therein). We chose these forms because they have been observed in homesign (Coppola et al. 2013) and LSN (Goldin-Meadow et al. 2015a), and because we could document them even in the absence of a stable lexicon or “core” vocabulary in homesigners (Richie et al. 2014). The meanings in a classifier construction are conveyed iconically by the individual parameters of the sign. Each parameter in a classifier construction—handshape, movement, location, orientation, and non-manual behaviors—can convey a discrete meaning, which is productively combined with the meanings of the other parameters. Parameters thus function as individual morphemes with movement functioning as a ‘light’ verbal root (either move or be-located), and handshape functioning as a classifier affix (Supalla 1982; Kegl 1990; Schick 1990; Brentari & Padden 2001; Emmorey 2003; Zwitserlood 2003; 2012; Benedicto & Brentari 2004). As noted earlier, the phonological shape of a specific form-meaning pairing in handshape, movement or location is the same whether it is produced simultaneously or sequentially with other forms. The difference lies in their temporal organization.

In classifiers, agency is associated with both the handshape parameter and movement axis, and number is associated with movement repetition. In order to introduce the relevant forms, Figure 2 presents SVP examples of classifier predicates illustrating the handshape types and movement properties that are analyzed in this work. Combinations involving MVPs are illustrated in Figure 3.

Figure 2

Still images drawn from classifier constructions in ASL with the following glosses in English (descriptions of the scene described follow each gloss): (a) Pen-on-horizontal-surface (scene: no-agent/single object/no movement); (b) Pens-on-horizontal-surface (scene: no-agent/multiple objects/no movement); (c) Someone-put-pen-on-horizontal-surface (scene: agent/single event/movement); (d) Someone-put-pens-on-horizontal-surface (scene: agent/multiple event/movement) [Images reprinted from Brentari et al. 2020 with permission].

An Asian American woman with shoulder-length dark hair wearing a long-sleeved black sweater is featured in still images drawn from videos in which she models four classifier constructions in ASL. Red arrows indicate the path and manner of movement of the hands. The examples are arranged in a 2x2 grid, with the events featuring No Agent on the top row, the events with an Agent on the bottom row, the first column showing examples with a single object or event, and the second column showing examples with multiple objects or events.

Figure 3

Single verb predicates (SVPs), repeated to express plurality, illustrating: (a) redundant marking for agent on the handshape (handlingHS) and movement axis (midsagittal); (b) agent marking on handshape only (handlingHS); (c) agent marking on the movement axis only (midsagittal).

An Asian American woman with shoulder-length dark hair wearing a long-sleeved black sweater is featured in still images drawn from videos in which she models three predicates in ASL that express agentivity and plurality in the handshape and/or movement axis. Red arrows indicate the path and manner of movement of the hand. The examples are arranged in a row.

Figure 2 gives examples of a no-agent meaning, pen-on-horizontal-surface (Figure 2a, 2b, intransitive/stative/no-agent) and a contrasting agent meaning Someone-put-pen-on-horizontal-surface (Figure 2c, 2d transitive/agent). The figure also gives examples of a singular meaning, pen-on-horizontal-surface (Figure 2a, 2c; singular) and a contrasting plural meaning, pens-on- horizontal-surface (Figure 2b, 2d; plural). The following three phonological properties that express agency and number in these types of sign language classifier predicates will be analyzed.

Agency/Handshape. Across sign languages, object handshapes are used to express intransitive, non-agentive clauses, and handling handshapes are used to express transitive, agentive clauses (Zwitserlood 2003; Benedicto & Brentari 2004; Brentari et al. 2012; 2015a; 2015b; 2016; Mazzoni 2008; Goldin-Meadow et al. 2015a; Rissman et al. 2020).² The two classes of handshape are defined by their iconic properties. Object handshapes use hand-as-object iconicity, and handling handshapes use hand-as-hand iconicity.³ Accordingly, Figures 2a and 2b show an object handshape representing a long, thin object in a location without an agent; Figures 2c and 2d show a handling handshape representing how a long, thin object is moved by an agent.

Using the hand to represent the hand when there is an agent present and using the hand to represent the object when an agent is not present may seem like an obvious choice. However, hearing speakers do not produce this distinctive pattern when asked to describe these types of events using gesture without speech. Gesturers readily produce handling handshapes, which use hand-as-hand iconicity, but they produce few object handshapes in the non-agentive predicate context (Brentari et al. 2012; 2015a; 2016).⁴

Agency/Movement axis. In Figures 2a and 2b, the signer produces a downward movement in neutral space without reference to the body. This is a lateral axis of movement, in this case a straight vertical movement (in the y-plane) at a perpendicular angle to a horizonal plane in front of the signer with no reference to the signer in the movement itself. Lateral movements are associated with a no-agent, intransitive meaning. In Figures 2c and 2d, the signer produces a movement anchored at the signer’s body, which then moves away from the body. This is a midsagittal axis of movement (z-axis, toward/away from the signer), which is associated with an agentive, transitive meaning. In this way, the movement axis, along with object and handling handshapes, is used to mark the agent/no-agent distinction (Horton et al. 2015). In Figure 2, we see agency expressed redundantly on the handshape and movement axis, but it is possible for the handshape and movement axes to represent different meanings (e.g., a lateral [no-agent] axis combined with a handling [agent] handshape).

Movement repetition. In Figures 2a and 2c the signer produces one movement with a single trajectory. In Figures 2b and 2d, the signer produces repetitions of the movement trajectory. Prior crosslinguistic work on number marking in sign languages has shown that movement repetition often iconically maps onto objects or multiple events—also referred to as pluractional verbs (Kuhn & Aristodemo 2017; Kuhn 2019)—both in classifier constructions (which express spatial events, e.g., Coppola et al. 2013) and in other vocabulary. Expressing multiple nouns or verbs via repetition can be restricted depending on the base lexical item (Fischer 1973; Fischer & Gough 1978; Zwitserlood & Nijhoff 1999; Pfau & Steinbach 2003; 2005; 2006; Hou 2013; Börstell et al. 2016; van Boven 2021). We focus on plural forms marked with movement repetition in the light verbs move and be-located; in such cases the repetition can also include information about the spatial arrangement of the objects under discussion (Pfau & Steinbach 2005)

2.2 Previous work on variation in sequential and simultaneous structures in predicates

Studies on LSN (Senghas et al. 2004, 2013) and other sign languages (Brentari et al. 2020) lay the groundwork for the current study of diachronic variation in an emerging language. Senghas et al. (2004) found that one group of LSN signers was more likely to split manner and path into two signs—V1 and V2, as in (9b)—than Spanish-speakers, who tended to combine manner and path simultaneously in a single conflated gesture produced along with speech (9a). This type of sequential structure is also documented in ASL (Supalla 1990). The iconicity of the time of the event is lost since both manner and path occur at the same time in the event being described. These results highlight a contrast between gesturers’ (simultaneous) and signers’ (sequential) preferences, showing that the visual-manual modality does not always favor simultaneous over sequential forms (see also Özçalışkan et al. 2016).

Senghas et al. (2013) later discovered a transitional state in LSN’s trajectory. They analyzed descriptions of manner and path events in child homesigners in Turkey, the signers who created LSN (i.e., LSN1), as well as the signers who learned LSN from Cohort 1 a decade later (i.e., LSN2). They found a “mixed” form, in which one verb conflates manner and path, combined with a second verb that represents either manner or path (but not both; see also Goldin-Meadow et al., 2015b). They first appear in homesign, peak in LSN1, and are rarely found in LSN2. (The criteria for assigning LSN signers to cohorts [LSN1 and LSN2] are described in detail in Section 2.4.)

Brentari et al. (2020) analyzed SVPs and MVPs in classifier predicates in four sign languages—ASL, BSL, HKSL, and Italian Sign Language (LIS)—to examine variation in the use of SVPs and MVPs expressing agency and number. They found that object handshapes and lateral movements did not always express non-agentive meanings; in contrast, handling handshapes and midsagittal movements were primarily restricted to agentive meanings.⁵ Object handshapes and lateral axis thus appear to be default forms, such that the distinction is [+agent] /ø agent (Pfau & Steinbach 2006).

Handshape type (handling) and movement axis (midsagittal) redundantly mark agency. However, as Pfau and Steinbach (2006) have noted, the midsagittal movement axis can be blocked when movement is repeated to indicate plurality. Figure 3 illustrates this point. All three forms in the figure are SVPs, repeated to express plurality. But the three forms express agency differently. In Figure 3a, both handshape (handling) and axis (midsagittal) redundantly express [+agent]. In Figure 3b, only handshape (handling) expresses [+agent], and axis (lateral) is a default form. In Figure 3c, only axis (midsagittal) expresses [+agent] and handshape (object) is a default form. Brentari et al. (2020) found that in BSL, HKSL and LIS, signers tend to produce single verb forms as in Figure 3a, combining repetition and midsagittal axis (no blocking); however, ASL signers tend to produce single verb forms as in Figure 3b, blocking the use of midsagittal with repetition. Verb forms like Figure 3c were much less frequent, suggesting that repetition blocks movement marking (midsagittal) more frequently than handshape marking (handling) for agency.

More relevant for the current study is redundancy across verbs in MVPs. Figure 4a illustrates an MVP that is partially redundant across V1 and V2: both V1 and V2 are marked with repetition for plural, but only V1 is marked for [+agent] (on both handshape and movement axis). We call these forms “mixed,” as in Senghas et al. (2013) and Goldin-Meadow et al. (2015b). Figure 4b illustrates an MVP that is not redundant: V1 is marked for agent (again on both handshape and movement axis) but has no marking for plural; V2 has repetition and is thus marked for [+plural] but has no marking for agent (it instead uses two default forms, object handshape and lateral movement). We call these forms “split.”

Figure 4

Multiple-verb predicates (MVPs) illustrating: (a) a “mixed” form with partial redundancy on V1 and V2: V1: plural (repetition) and agent (handshape and axis) and V2: plural (repetition) and no-agent (handshape and axis); (b) a “split” form with no redundancy between V1 and V2: V1: agent handshape (handshape and axis) and V2: plural (repetition) and no-agent (handshape and axis).

An Asian American woman with shoulder-length dark hair wearing a long-sleeved black sweater is featured in two sets of still images drawn from videos in which she models two multiple-verb predicates in ASL that express agentivity and plurality. Red arrows indicate the path and manner of movement of the hand. The examples are arranged in two sets of two images each.

BSL, HKSL, and LIS signers’ MVPs containing [+agent]/[+plural] resemble those in Figure 4a; that is, “mixed” forms where one verb uses handshape (handling) and movement axis (midsagittal) to mark agent, and movement repetition to mark plural; the second verb redundantly marks plural using the default handshape and motion forms with movement repetition. In contrast, ASL signers’ MVPs containing [+agent]/[+plural] meanings resemble those in Figure 4b, that is, “split” forms where one verb uses handshape (handling) and movement axis (midsagittal) to mark agent (no repetition); the second verb uniquely marks plural using the default handshape and motion forms. Thus structures for agency and number appear to vary systematically across sign languages.

To summarize this section, we see that sign languages use simultaneous layering of meaning within words more than spoken languages do, as seen in Figure 1, and producing complex predicates as SVPs is very common in sign languages, more so than in spoken languages. However, not all complex predicates are SVPs, as we see in as we see in examples (5)–(8).

Since new languages are found only among sign languages, we analyze single-verb predicates (SVPs) and multiple-verb predicates (MVPs) in three groups of signers in Nicaragua—sign vs. sign+sign structures in classifier predicates. We analyze the emergence of form-meaning pairings to express agency (agent/no-agent) and number (singular/plural) in this new language. These data allow us to track the process of change over a shorter time frame, in a more granular manner, and at a different moment in historical time, than is typically possible in older languages.

2.3 The emergence of LSN

The creation of a new sign language in Nicaragua over the past five decades offers us the opportunity to observe the earliest stages of agent and plural structures as they emerge. Before the 1970s, deaf Nicaraguan individuals had little contact with each other (Kegl & Iwata 1989; Polich 2005; Senghas et al. 2005; Coppola 2006; 2020). There were periods when various classrooms and clinics were available to young children, but intergenerational contact and formation of a deaf community were hindered by the lack of a unifying national educational system, societal attitudes that isolated deaf individuals, and marital patterns that generally precluded hereditary deafness. One school, founded in Managua in 1974 with 25 deaf students, became publicly accessible in 1979 and expanded to include approximately 50 deaf students. In 1981, the school had a total enrollment of 500, with 125 deaf students, and by 1983, the school served 251 deaf students (Polich 2005). For the first time, a social community of deaf signers existed, with continuity from childhood through early adulthood. Today, LSN signers in Managua continue to socialize frequently at the National Nicaraguan Deaf Association and in each other’s homes.

In projects documenting LSN, researchers have divided signers into cohorts based on their year of entry into the signing community. The first cohort of signers began life as homesigners, the term used for deaf individuals who do not have access to spoken language and are not exposed to a sign language (Goldin-Meadow 2003). They were brought together as children, and they formed a deaf community in the late-1970s and early-1980s. Although the teachers at the school initially used spoken Spanish exclusively, these early signers produced and saw the signing of the other students during free periods outside of class. The resulting language is called the “initial contact variety,” and the signers are considered the first cohort of LSN signers (LSN1; Senghas et al. 2005). LSN1 signers have had the opportunity to sign with each other, but there was no pre-existing sign language model when they entered school. A second cohort of signers (referred to here as LSN2 signers) includes deaf individuals who entered the school in its second decade, from the mid-1980s to the early-1990s. In their early school years, they interacted with teenagers already at the school who were LSN1 signers. LSN2 and subsequent cohorts (LSN3, LSN4, etc.) developed a “sustained contact variety” (Senghas et al. 2005); in addition to a shared, deaf signing community, they have grown up with access to the signing produced by the previous cohorts as a language model.

The majority of deaf individuals in Nicaragua are not part of this signing community and do not know LSN. Due to a variety of social, geographic, and financial obstacles, they do not go to school or regularly interact with other deaf people. These deaf individuals are considered “homesigners”; see Coppola (2020) for a brief review. The homesigners included in this study have hearing losses significant enough to prevent the acquisition of a spoken language; they have not experienced regular exposure to LSN or to formal education; and none has successfully learned written or spoken Spanish. In addition, the homesigners do not interact with each other, do not have regular interactions with deaf or hearing signers of LSN, and have been using their individual homesign systems as their primary language for their entire lives (Coppola 2002). Each homesigner has a unique communication history with hearing family members and friends with whom they interact regularly (Coppola 2002). The hearing communication partners often engage with homesigners using signs and gestures, but the homesigners’ systems are not fully taken up by these hearing communication partners at either the lexical (Richie et al. 2014; Quam et al. 2022) or syntactic (Carrigan & Coppola 2017) levels.

The current study uses an apparent time approach (Labov 1994; Cukor-Avila & Bailey 2013), comparing different groups of signers in Nicaragua today to reconstruct the history of predicate constructions in LSN. Specifically, we examine whether morphemes are produced simultaneously on a single verb or are distributed across multiple sequential verbs in the predicate. We focus on meaning-form pairs associated with agency (no-agent/agent) and number (singular/plural). In cases where sequential forms are used, we also ask how the information is packaged, and the extent to which the information in an MVP is repeated in both verbs (and thus is redundant) or split between the two verbs (and thus is not redundant).

2.4 Research Questions

Because signers representing these early stages of language emergence continue to live and use their language today, we have the opportunity to reconstruct the emergence of specific structures within LSN by making comparisons in a controlled way across groups of present-day signers using the apparent time approach. Signers with different language experiences at different points in time and under different social circumstances reflect different historical time points in the development of LSN. In the current analysis, we ask whether there is a systematic diachronic pattern in how agency and number are expressed at three points in the emergence of LSN: Homesign, LSN1, LSN2. We ask two questions.

1. How are agency and number expressed?

   1. Are they expressed simultaneously in a single-verb predicate (SVP), or sequentially in a multiple-verb predicate (MVP)?

   2. Does simultaneity/sequentiality vary across these three stages of language emergence?

2. How are meanings packaged across the two verbs in an MVP?

   1. Does one verb in an MVP express both agency and number?

   2. Is the information expressed across the MVP partially redundant (“mixed”) or not redundant at all (“split”)?

   3. Does this pattern vary across the three stages of language emergence?

4.1 Which factors predict the probability of participants producing single-verb or multiple-verb predicate responses?

We now turn to the general analysis of factors predicting the occurrence of SVPs and MVPs in the data. We fitted a mixed-effects logistic regression model to the data, using the “glmer” function, from the package lme4 (version 1.1-27.1; Bates et al. 2015) in R (version 4.1.0; R Core Team 2021) to understand which stimulus conditions and participant groups are associated with the production of SVPs or MVPs, while accounting for nested and repeated measurements. 630 observations across participants and groups were analyzed. The random intercepts were Participant and Stimulus Object (8 object types shown in the stimulus vignettes). Model comparisons were performed using three models: M0, M1, and M2. We used M1 because it provided the best fit.

M0 is a baseline, unconditional model that included only the two random effects (Participant and Stimulus Object). M0 had an Intraclass Correlation Coefficient (ICC) of 0.32, indicating that the two random effects alone explained 32% of the variance in MVP responses.

M1 included three fixed effects in addition to the two random effects. The fixed effects were: Participant Group, with three levels: Homesigners, LSN1 (reference group), and LSN2; Agency, with two levels: no_agent, agent; and Number (of objects) also with two levels: singleOBJ and multipleOBJ. Significance of categorical fixed effects was assessed using Wald’s χ² implemented in the car package (version 3.0-12; Fox & Weisberg 2019). Table 2 shows the regression table for M1. A third model, M2, included an interaction term between Agency and Number. The added term did not indicate a significant interaction effect (Wald χ² = 0.54, df = 1, p = 0.46), nor did it improve model performance using standard model comparison techniques (e.g., ΔAIC = 1.5, ΔBIC = 5.9).

In the model shown in Table 2 (M1), all of the fixed effects were significant, and the random effects accounted for 29% of the variance (see the ICC in Table 2), with relatively equal amounts of variance explained by Participant and Object (τ values in Table 2). We observed the largest effect from the predictor Number (Wald χ² = 54.8, df = 1, p < 0.0001); MVP responses were produced more frequently in response to vignettes with multiple objects vs. vignettes with single objects. The model also indicated a large effect of Agency (Wald χ² = 5.11, df = 1, p = 0.02); MVPs were produced more frequently in response to vignettes with an agent vs. vignettes without an agent. Group also showed a significant but relatively weaker effect (Wald χ² = 8.9, df = 2, p = 0.01). There was a large confidence interval on the estimate (OR[95%CI] = 3.08[0.88–10.80], p = 0.078, Table 2). The difference between homesigners and LSN1 was not significant; however, LSN1 signers showed a trend toward fewer MVPs overall than either the homesigners or LSN2 signers.

Figure 8 shows the predicted probabilities of an MVP response in descriptions of scenes classified according to Agency (Agent, No agent) and Number of objects (Single Object, Multiple Objects) in homesigners, LSN1 signers, and LSN2 signers for the best-performing model.

Figure 8

Predicted probabilities of a multiple-verb VP response in descriptions of scenes classified according to Agency (Agent, red circles; No-Agent, blue circles), and Number of objects (Single Object; Multiple Objects) in Homesigners, LSN1 signers, and LSN2 signers. The data set is available in the project OSF repository.

A plot depicting the predicted probability of multiple-verb predicates produced on average by each participant group: Homesign, LSN1, and LSN2. Within each group, on the x-axis, results are shown separately for trials containing a single object or multiple objects, and within those categories red circles show the group means and confidence intervals for trials with an agent and blue circles show the mean and confidence interval for responses to trials without an agent. Each group on average produced multiple-verb predicates more frequently in the Multiple-Object conditions than in the Single-Object conditions, with the magnitude of the difference varying by group. Each group on average also produced multiple-verb predicates slightly more frequently in the Agent conditions than in the No-Agent conditions. The difference between Homesigners and LSN1 was not significant; however, LSN1 signers showed a trend toward fewer MVPs overall than either the homesigners or LSN2 signers.

4.2 How are agency and number information packaged in MVPs?

To assess how agency and number information is packaged, we narrow our focus to descriptions of scenes that are designed to elicit both agent and plural markings (i.e., agent_multipleOBJ items). We address two questions in this section: Are agency and number produced simultaneously or sequentially? If two verbs are used to convey agency and number, is the information distributed across the two verbs, or is the information in the second verb redundant with the information conveyed in the first verb?

For each response, we first needed to identify whether agency and plurality information was included. Signers could mark both the agent and the plural, only the plural, only the agent, or neither. Table 3 shows that all groups produced [+agent] [+plural] marking more frequently than any other type of predicate description. When information is not included in the description, it is typically the agent, not the plural, that is omitted. Very few descriptions omit both an agent and plural form.

To explore whether agent and plural marking are produced simultaneously or sequentially, we divided responses containing both agent and plural markings into three categories (85 data points). In simultaneous agent+plural SVPs, agent and plural are simultaneously produced on the same verb. In sequential split agent–plural MVPs, agent is produced on one verb and number is produced on the other. In mixed MVPs, agent and number are produced simultaneously on one verb, and either agent or number is produced separately on additional verbs. The results are presented in Figure 9.

Figure 9

The proportion of responses marking both agent and plural classified according to whether the two are: simultaneously produced on the same verb (simultaneous: V1: agent+plural (gray dots)); the two are produced simultaneously on one verb, along with either marker on a separate verb (mixed: V1:agent+plural, V2: agent: (black triangles)); or the two are produced sequentially, agent on one verb and number on another (sequential split: V1: agent, V2: plural (black squares)). Each dot represents a participant; the upper and lower bounds of the boxes indicate the first and third quartiles, and the horizontal bar in each box shows the group mean.

A beeswarm boxplot depicting the proportion of predicates using simultaneous, mixed, or sequential marking for agentivity and plurality, by participant group. The participant groups are on the x-axis (from left to right: Homesign, LSN1, and LSN2). Responses in which agent and number are marked simultaneously are indicated with light gray circles, responses with mixed simultaneous and sequential marking are indicated with dark gray triangles, and responses with sequential marking are indicated with black squares. Each participant in each group is represented by one light gray circle, one dark gray triangle, and one black square. Each group shows a distinct pattern. Homesigners on average produce simultaneous marking most often (.60), followed by mixed (.30), and three out of four Homesigners do not produce sequential marking, though one Homesigner uses it half the time. In contrast, LSN1 signers use simultaneous marking on average on .95 of responses, mixed marking on .05 of responses (with two LSN1 signers never using it, and one LSN1 signer using it on .70 of responses), and they do not use sequential marking at all. LSN2 signers show a third pattern, using each type of marking relatively equally, with one LSN2 signer producing no simultaneous marking and one producing no sequential marking.

The distribution of SVPs vs. MVPS differed across groups. A Mann-Whitney-U Test of Comparisons for small Ns) on SVPs revealed significant differences pairwise among the groups: homesigners vs. LSN1 signers, homesigners vs LSN2 signers, and LSN1 vs LSN2 signers all had significant differences (U = 1; p = .029)

Figures 10, 11, 12 provide illustrations of the [+agent] [+plural] response types (predicates only) shown in Figure 9. Complete transcriptions are given in (12)–(14).⁷ Figure 10 illustrates a “mixed” partially redundant MVP produced by a homesigner. The first verb is marked for agent (handling handshape + midsagittal axis) but not for plural; the second verb is marked for both agent (handling + midsagittal) and plural (repetition). Figure 11 illustrates an SVP produced by an LSN1 signer that is marked for both agent (handling + midsagittal) and plural (repetition). Figure 12 illustrates a sequential or “split” non-redundant form produced by an LSN2 signer. The first verb is marked for agent (handling + midsagittal) with a single, unrepeated movement (the default); the second verb is marked for plural (repetition), using the two default forms for no-agent (object + lateral).

Figure 10

A homesigner producing a partially redundant mixed MVP to describe an agent putting multiple planes on the table in a row (condition 7).

This example contains four images. The two larger images in the center show different time points of a Nicaraguan woman with long dark hair pulled back into a bun wearing a sleeveless blue and white patterned dress producing a partially redundant “mixed” multiple-verb predicate two handling handshapes. Red arrows indicate the path and manner of movement of the hands. The two smaller images on the far left and far right show models of the handshapes the woman uses in her response (they both show the same handshape, in which four extended fingers meet the thumb without any finger joints flexing).

Figure 11

An LSN1 signer producing an SVP to describe an agent putting multiple books on the table (condition 7).

This example contains three images. The larger image on the right shows a Nicaraguan man wearing a tight-fitting black cap and a short-sleeved white-and gray patterned T-shirt producing a single-verb predicate with two handling handshapes. Red arrows indicate the path and manner of movement of the hands. The two smaller images on the far left show models of the handshapes the man uses in his response (they show mirror images of the same handshape, in which four fingers are extended without any joints flexing, with a 1.5-inch gap between the fingers and the thumb).

Figure 12

An LSN2 signer producing a split, non-redundant MVP to describe an agent putting multiple lollipops on the table in a random arrangement (condition 8).

This example contains six images. The four larger images in the center show a Nicaraguan woman with long dark hair pulled back into a ponytail wearing a multi-colored printed short-sleeved blouse producing a sequential, multiple-verb predicate in which the first predicate shown in the first large image uses a handling handshape and the second predicate shown in the remaining three larger images uses an object handshape. Red arrows indicate the path and manner of movement of the hands, and black numbers 1 to 4 label the individual events expressed in her response. The two smaller images on the far left and the far right show models of the handshapes the woman uses in her response; the leftmost handshape shows the handling handshape of the first predicate, in which the tip of the index finger meets the top of the thumb, with the other fingers closed. The handshape on the right shows the object handshape, in which the index finger is extended and the other fingers are closed.