Comparing Different Measures of Bilingual Input Derived From Naturalistic Daylong Recordings

Abstract

Purpose: Measuring language input, especially for infants growing up in biligual environments, is challenging. Although the ways to measure input haexpanded rapidly in recent years, there are many unresolved issues. In tstudy, we compared different measurement units and sampling methods useto estimate bilingual input in naturalistic daylong recordings. Method: We used the Language Environment Analysis system to obtain anprocess naturalistic daylong recordings from 21 French–English bilingual families with an infant at 10 and 18 months of age. We examined global context-specific input estimates and their relation with infant vocal activene(i.e., volubility) when input was indexed by different units (adult word counspeech duration, 30-s segment counts) and using different sampling metho(every-other-segment, top-segment). Results: Input measures indexed by different units were strongly and positivecorrelated with each other and yielded similar results regarding their relatiowith infant volubility. As for sampling methods, sampling every other 30-s sment was representative of the entire corpus. However, sampling the top sments with the densest input was less representative and yielded differeresults regarding their relation with infant volubility. Conclusions: How well the input that a child receives throughout a day is ptrayed by a selected sample and correlates with the child’s vocal activeness depends on the choice of input units and sampling methods. Different inunits appear to generate consistent results, while caution should be taken whchoosing sampling methods.