Everyday input variability affects word recognition at 11 but not 16 months

Emerging results point to a relationship between environmental variability and early language acquisition on various levels. Most prominently, the amount of talk infants hear has been positively linked to their vocabulary development, and follow-up work showed that input from not just the main caregiver but all talkers is a better predictor for language development. It is yet unclear whether multiple talkers in the environment lead to more input and thus better vocabulary outcomes. At the same time more talkers also add more pronunciation variability. This might be problematic, because infants might have to first develop robust phonological representations before they can learn different pronunciations of one word efficiently. Thus, it is an open question whether being habitually exposed to multiple talkers, possibly with accented speech, leads to a beneficial effect of speaker variability in lexical development or not. To assess how general talker variability in infants’ daily life as well as whether or not they hear accented speech influences their budding lexicon, we tested to what extent these two variables predict word recognition. We tested two age groups, 11-month-olds (n=51) who are just beginning to build a lexicon and 16-month-olds (n=47), whose “phonological constancy” is beginning to emerge. We measured French-learning infants’ recognition of 2 known words (“biberon” [babybottle] and “chien” [dog]) in terms of looks to the correct picture. Parents additionally reported details of their household size, daycare arrangements, and social engagements to estimate the number of talkers in their child’s input. We also asked whether talkers had a noticeable accent. Infants’ receptive vocabulary size served as a proxy for linguistic development and was measured via the MacArthur Bates communicative development inventory (French version). We analyzed the eye tracking data following standard protocols: Our dependent variable is empirical logit of target looks in a 1s time-window following target word onset, determined in a consistent way across age-groups as the 1000 ms time-window surrounding the peak of the word recognition curve. This resulted in a time-window of 300-1300 ms after word onset for 11-month-olds, and 1000-2000 ms for 16-month-olds. The dependent variable and the predictors (log normalized number of speakers, presence of accented speech, vocabulary) were entered into a linear mixed effects model using R. Significance was determined with normal approximation.
For 11-month-olds, the recognition response is significantly modulated by the presence of accented speech (p = .031) and this factor interacts with the number of talkers in infants’ input (p = .016). The main effect of number of talkers (p = .070) and of vocabulary (p = .078) were marginally significant . At 16-months, we do not observe any significant effects of our predictors. Taken together, our data indicate that less experienced language learners rely on few talkers to learn words, whereas advanced learners can harness variability in their input and thus are seemingly not negatively affected by it.