**1. Introduction**

What are the origins of the complex symbolic systems that we find in modern human languages? Languages arise out of an interaction between the human mind that represents and organizes information, the body that manifests that information in a physical form, and the social function of transmitting the information from one individual to another. By looking at a language at its inception we can explore the nature of this interaction in order to learn about the specific contributions of these three resources. For example, what communicative expressions are pulled from the environment and how are they reshaped and reallocated as a system develops? Are initially dominant forms adopted from the outset, increasing in use as an emerging system is passed from one generation to the next? Or are alternative forms favored according to their suitability for particular functions? How is a dynamically changing linguistic system shaped according to its learnability by younger learners? In the current study we track the development of non-manual grammatical markers of wh-questions in Nicaraguan Sign Language (LSN1), over the course of the language's emergence over the past 45 years. We ask whether these markers originate from co-speech facial gestures, and we test three possible determinants of their selection and systematization.

Non-manual markers in sign languages are facial and body movements, often coarticulated with manual signs, that serve linguistic functions across all levels of language, from phonology to discourse (see Sandler and Lillo-Martin 2006, for a review). Although the

**Citation:** Kocab, Annemarie, Ann Senghas, and Jennie Pyers. 2022. From Seed to System: The Emergence of Non-Manual Markers for Wh-Questions in Nicaraguan Sign Language. *Languages* 7: 137. https://doi.org/10.3390/ languages7020137

Academic Editors: Wendy Sandler, Mark Aronoff and Carol Padden

Received: 29 October 2021 Accepted: 28 April 2022 Published: 30 May 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

appearance of non-manual features in conjunction with manual signs may seem holistic to a naive eye, these non-manual elements are indeed combinatorial (Herrmann 2015; Sandler 2010). Children natively learning a sign language can readily separate non-manual features from manual features, as evidenced by their early sequential, instead of simultaneous, articulation of non-manual and manual elements (Anderson and Reilly 1997; Reilly 2006; Reilly et al. 1990).

Several researchers have speculated that the non-manuals observed in sign languages have their source in the facial gestures produced by non-signers while speaking (Benitez-Quiroz et al. 2016; Janzen and Shaffer 2002; McClave 2001; Pfau and Steinbach 2006, 2011). In at least one isolated village sign language, Kata Kolok, researchers have observed the adoption of a gestural negative headshake as a marker of negation that increases across generations of signers (Lutzenberger et al. 2022). If facial gestures are indeed a source of non-manual markers, what is the process that integrates such gestures into the grammatical system of a sign language? In examinations of other morphosyntactic structures, we have documented the emergence of simultaneous manual morphology in LSN (Senghas and Coppola 2001; Kocab et al. 2016), so we expected we might observe a similar trajectory of emergence with non-manual markers. Some evidence points to the integration of the systematic use of non-manuals over generational transmission in another recently emerging sign language, Al Sayyid Bedouin Sign Language (ABSL); only second-generation signers showed a consistent use of head and body movements that were aligned with clause boundaries (Sandler 2010; Sandler et al. 2011).

In this study, we empirically test the proposal that the facial gestures of non-signers are the source of non-manual markers in sign languages by examining the emergence of non-manual markers in LSN. We specifically focus on those non-manual markers that indicate wh-questions; that is, questions that query specific information, such as who, what, when, how many, and where. Across many different sign languages, wh-questions are accompanied by a non-manual facial gesture, most commonly either a brow furrow or a brow raise (Zeshan 2004). We extend the previous linguistic work on the emergence and grammaticalization of non-manual markers by adopting a quantitative approach to our study of the emergence of non-manuals and by explicitly looking at the rates and characteristics of the production of a variety of possible non-manuals in non-signers and in learners of an emerging sign language in Nicaragua.

The case of LSN offers a unique opportunity to test the robustness of co-speech facial gestures as a source for non-manual markers and to observe the process by which such forms are taken up and integrated into a rapidly changing linguistic system. LSN was created by deaf children and adolescents, starting with an initial cohort of 50 individuals who arrived in a new special education school in Managua in the 1970s. Although they were instructed entirely in Spanish, they communicated with each other primarily using gestures and homesigns (Kegl et al. 1999; Polich 2005; Senghas et al. 2005). Through peer interaction and intergenerational transmission, these gestures and homesigns transformed into a new, natural sign language, currently the primary, daily language of over 1500 deaf people. Because the language developed so recently, its originators are still alive today and are able to offer us a view into its origins.

Transmission from the original cohort to the learners that followed was a critical moment in the emergence of this new linguistic system. Research on artificial language emergence in the laboratory suggests that combinatoriality and systematicity can arise over repeated transmission of a system, through multiple iterations of learning (Kirby et al. 2008). In the case of LSN, the language was taken up by each cohort predominantly while they were children, who later, as adolescents, transmitted the language to a subsequent cohort of child learners. In many cases, the new arrivals introduced systematic changes to the language that the (by that time) adolescent and adult members of the community did not acquire (Senghas and Coppola 2001). We capture this change empirically, by dividing today's LSN signers into three roughly decade-long age cohorts, with each cohort having served as a language model for the next. We can then apply an "apparent time" approach, measuring diachronic change over LSN's first three decades with a cross-sectional comparison of these cohorts today (Bailey et al. 1991; Labov 1963; Sankoff 2006). This approach critically rests on the observation that beyond adolescence individuals do not significantly change their language, and their current language use reflects the language of their childhood (Labov 1963). In this way, we can read a layered, living record of LSN, with the first cohort's language revealing aspects of the initial form of LSN, and the later cohorts' representing more recent developments and changes to the system.

If the markers for wh-questions in a sign language originate in co-speech facial gestures, we should be able to find, in that seed, corresponding forms for any LSN non-manual markers we identify. Furthermore, if observable characteristics of the seed determine which forms will ultimately be taken up by the language, we should be able to capture those characteristics empirically from the outset and measure their cascading effects. We selected three candidate characteristics that might influence the adoption of a form. First, we determined how frequently the different forms are used. Previous research on older languages has shown that the frequency of a construction in learners' input and in languageusers' production predicts language acquisition and change (Bybee 2007, 2010; Diessel 2007). Accordingly, a non-manual that appears more frequently in the facial gestures that accompany spoken questions of non-signers might be more likely to be selected as a grammatical marker of questions in LSN. Second, we considered the duration of the forms, as a measure of their fitness for acquisition and a measure of their suitability to the function of marking questions. A feature or form that is produced for a longer period has greater salience than a shorter one, making it easier to perceive and acquire (e.g., Fridland et al. 2004). Additionally, a form that can be sustained has more potential to be leveraged for its grammatical affordances such as spreading over longer phrases (Sandler 1999) as a way to mark the scope of a query. Third, we examined the timing of the non-manual, specifically whether it was produced at the same time as the question word. Wh-question non-manuals in older sign languages are consistently produced simultaneously with the manual signs for a question (Benitez-Quiroz et al. 2014). The coarticulation of the non-manual with a wh-question word might strengthen the association between the non-manual and the function of asking a question, making the mapping between form and function more salient to learners. If these characteristics drive form selection, then the facial gestures and body movements of non-signers that are most frequent, are held the longest, and overlap the most frequently with a wh-question word should be the ones taken up as non-manual markers of wh-questions in LSN.

These factors should be relevant at each moment of transmission, with each cohort seeding the next, from the hearing Spanish speakers, to the first cohort, to the second, to the third. We don't expect the same outcome at every stage; rather, we expect the same mechanisms to apply to a dynamically changing system. As the input for each subsequent group changes, the application of these same mechanisms will yield a different, corresponding output. So, for example, we might see a tipping point where one or very few forms come to dominate. The pattern of change in the community today will give us clues to its origins. Changes that are due to adults and children using the language repeatedly, over time, should be evident in all of the groups today, since all of the members were adults at the time the recordings were made, with many years of experience using the language together. Changes that are the result of how language is learned initially by children will leave a different pattern, one of the differences between cohorts that persist in adulthood.

We might expect certain aspects of the nature of children's learning to leave a particular kind of imprint on their language. Of course, children are not exposed directly to the grammar that produces the language they observe; instead, they observe only the patterned output that is generated by their interlocutor's grammar. Even so, children are quite skilled at mastering the intricacies of their language. They are particularly sensitive to wordinternal patterns, and detect them with very little exposure (Saffran et al. 1996). Evidence from both signed and spoken languages has shown that child learners are better than adult learners at extracting the regularities of languages (Mayberry and Fischer 1989; Saffran et al.

1996). How might we expect child learners to respond to a system that lacks such intricate regularities? One case study followed a deaf child acquiring his language from deaf parents who had learned ASL only as older adolescents, and therefore did not have the fluency of native signers. The child's command of ASL eventually surpassed his parents'; he applied the morphology of the language, particularly the spatial morphology, more consistently than they did (Singleton and Newport 2004). In artificial language-learning experiments, when children are presented with miniature languages that are unsystematic and variable, they deviate from that input in predictable ways, increasing its regularity. Adults in the same experiments generally do not impose that kind of reorganization (Hudson Kam and Newport 2005). Given this kind of creative power in each individual learner, we were interested in discovering how repeated acquisition of LSN in its earliest years, over 1500 instances of learners, might build up the system of the language, in this case, the grammar for generating questions.

In the current study, we documented the array of facial expressions and body movements produced by Nicaraguan Spanish-speaking non-signers and computed their frequency, duration, and co-articulation with a wh-question word. We then compared nonsigners' use of these co-speech facial gestures with that of the first, second, and third cohorts of LSN signers. We hypothesized that the most frequently produced facial gestures produced by non-signers would be taken up as the preferred non-manual marker for wh-questions. We also hypothesized that facial gestures that were produced for a longer time and that were co-articulated with a wh-question word would also be more likely to be selected as a grammatical marker.
