Oksana Tkachman research seminar: Semantics-based spontaneous compounding emergence in artificial sign languages

Compounding is one of the simplest and most widespread word-formation processes in human languages: it creates novel words with already existing ones, and has a simple hierarchical structure of a head and a modifier. In spoken languages, compounding is nearly universal (Bauer 2017), and so far, it has also been attested in all signed languages where word-formation processes have been investigated (Meir, Aronoff, Sandler and Padden 2010; Tkachman and Meir 2018). Some researchers claim compounding to be one of the earliest linguistic processes to emerge in language evolution (Jackendoff 1999, 2002; Heine and Kuteva 2008).

We report on a study where sign-naïve English speakers, asked to create manual labels for various familiar objects, employed compounding as one of their ways of creating their novel lexicons. 50 native speakers of English with no knowledge of any sign language, ages 19-72, were asked to create sign names for 100 objects, appropriate to be used in an artificial sign language (note: “sign” throughout this abstract is a short name for  “artificial sign”). The data were coded by two independent coders.

We were primarily interested in the influence of various kinds of iconicity in signs, but we also noted that semantics of the objects motivated our participants to reuse already created signs in compound names for objects that are semantically or conceptually related. For example, a few participants reused the sign they coined for CAT much later in a compound sign for LION . We therefore analyzed such families of signs to see what kind of semantic motivations motivated their creation as well as what the basic properties of these compounds were. We classified as compounds all instances of multi-sign responses that were used as a label for an object, and that did not include poses, hesitations, body shifts, etc. Overall, we elicited 4975 responses, of which 2385 consisted of two or more signs. We excluded all compounds where one of the signs indicated only size or shape, as it could lead to overestimation of sign families. Additionally, we excluded all compounds were all the signs were unique, that is, we only included compounds were at least one of the signs used have been coined for an earlier referent. We then grouped responses with at least one sign reused into sign families, which were further analyzed for the underlying motivation of the sign family.

All 50 participants created multi-sign responses, and 44 out of 50 reused signs in compounds. Overall, we identified 266 sign families (~6 per participant in those who created them), with 3.9 as the average size of a sign family (range 2-20). Of 2385 multi-sign responses, 1042 were members of a sign family (~23.7 responses per participant, out of 100). Semantically, all sign families were classified as belonging to 35 categories, with the majority belonging to vehicles (30 sign families), water-related entities (25), trees (23), hard/solid entities (21), and animals (19). We speculate that these particular categories all have a prototypical member which can be used as part of a compound for other members of the category. This is especially interesting, since the order of presentation of pictures for sign creation was random; that is, we did not manipulate the data to have more prototypical members to appear before less prototypical members.

    The coding and analysis of this dataset is ongoing. We are looking forward to hearing your opinions and comments on this project!