Abstract

Shughni, an Eastern Iranian language spoken in Afghanistan and Tajikistan, possesses a binary system of grammatical gender (feminine and masculine) and a predominantly semantic system of gender assignment. In Shughni, unlike in most familiar Indo-European languages with grammatical gender (e.g., French, Russian), a noun's semantics, rather than its phonological form, is the most reliable predictor of its gender. In this paper, I explore two semantic patterns proposed to underlie gender assignment in the language: (i) thematic categorization (e.g., liquids and technological tools are feminine; milk products and sicknesses are masculine); and (ii) the relation of meronymy, where conceptual wholes are said to be feminine while their parts are masculine.

After providing a short description of the Shughni gender system, I present experimental evidence aimed at determining whether conceptual meronymy is at the heart of a veritable gender assignment pattern in Shughni. The experiment asked native Shughni speakers to assign gender to technological nouns (e.g., carburetor) in two conditions—when presented as a conceptual "part" (e.g. part of a car) and as a "whole" (as a stand-alone item). Although we find a statistically significant, albeit weak, correlation between feminine gender and "whole" condition, we argue for an alternative interpretation of our results in which thematic categorization drives gender assignment, and the role of meronymy is simply an illusion created by the system.

This experimental study on Shughni gender assignment is the first of its kind, and our results are admittedly difficult to interpret in the absence of an established framework for studies of this kind. Therefore, an important contribution of this paper is an initial set of experimental results and discussion on a semantic gender-assignment rule; this will set the stage for interpreting future studies of this kind. Moreover, through our experimental setup—a drag-and-drop (Duolingo-style) language production task built with the javaScript library jsPsych—we contribute to the available infrastructure for conducting field-based experiments with under-described language, which are glaringly underrepresented in psycholinguistic research. With this in mind, we discuss several of the solutions we implemented to challenges often faced in conducting experiments with such languages, including the creation of stimuli, the choice of writing system, and the overall implementation of the experiment. It is our hope that other researchers conducting experiments with unwritten languages can take lessons from our design process and reflections.