Clinton Parker - A written experiment for an under-written language: Balancing ecological validity and experimental control with Shughni

 Psycholinguistic experiments typically rely on written stimuli and highly controlled tasks, presenting particular challenges for work on under-described languages with limited traditions of literacy.  In this talk, I discuss these challenges in the context of a recent experiment on grammatical gender assignment in Shughni, an Eastern Iranian language of Tajikistan and Afghanistan. The experiment tested a proposed gender assignment pattern involving the conceptual relation of meronymy, whereby conceptual wholes are grammatically feminine while their parts are masculine.  The validity of this pattern in the language's grammar was examined by eliciting the gender of borrowed Russian technological nouns in two conditions: whole (e.g., a carburetor) and part (e.g., the same carburetor presented as part of a car). To balance experimental control with ecological validity, we developed a hybrid drag-and-drop production task similar to a Duolingo activity, combining naturalistic recorded conversations and visual stimuli with a limited amount of written Shughni. The design required important decisions about orthography, participant recruitment, field-based implementation, and how to elicit the relevant grammatical contrasts without making the purpose of the experiment transparent to participants. In this talk, I discuss the reasoning behind these methodological choices, evaluate what worked and what did not, and briefly present the experiment’s results. More generally, I consider how controlled experimental methods can be adapted to languages for which standard laboratory-based paradigms may be poorly suited.