How do we capture parents’ multilingual speech with their kids: an exploratory comparison of self-reported and speech-based metrics of translanguaging

Multilingual children’s language environments vary widely, yet they are often described by exposure proportions alone. In this exploratory paper, we describe how caregivers distribute and combine languages in speech during a child-directed picturebook narration task, using language-use visualizations, language proportions, language entropy, proportion of multilingual utterances, span entropy, and memory. We applied these measures to 146 narrations from caregivers in Singapore of children aged 0;8 to 3;4 and compared them to caregiver reports. Caregivers reported and used different linguistic varieties with their children, and reported language use and observed language-use during narrations showed moderate correspondence, indicating they characterise different facets of multilingual input. Among multilingual narrations, caregivers of older children used their languages less evenly and proportionally more English, whereas utterance-level mixing and span organisation showed no detectable age differences. We discuss methodological considerations, including segmentation and labelling decisions and provide code for researchers to compute these measures.