Though much research has sought to better understand the process of language acquisition, many fundamental questions regarding how infants complete the enormous task of learning a language within just a few years remain without a definitive answer. So far, researchers have compiled a basic timeline for the acquisition of linguistic knowledge, and proposed two learning mechanisms: a bias leading infants to pay more attention to speech sounds that are easier to produce, and general statistical induction skills that are used to process various kinds of sensory input. Which mechanism is operative remains a continuing and fundamental debate from both theoretical and experimental perspectives.
Surprising results from a recent pilot study indicate that infants make use of certain information regarding sound sequencing as early as 4.5 months, 4 months earlier than the previously established age. The revised findings appear to be the result of using experiments that present words to infants in phonetically controlled comparisons, rather than as a random selection. The dissertation seeks not only to confirm the pilot findings, but also to uncover further evidence regarding which of the two proposed learning mechanisms primarily facilitates acquisition of this early knowledge.
The research will employ the well-established Head-turn Preference paradigm, in which words are played to an infant on a speaker, accompanied by a blinking light. The infant's attention to the words is indicated by turning toward the light. The experiment monitors the amount of time that infants attend to different kinds of stimuli, demonstrating the distinctions that are evident to them at that particular age. The innovation in this study lies in the stimuli the infants hear: the forms presented contrast specific sound patterns, in particular the distinction of voicing between fricative sounds (such as f, s, and sh versus v, z, and zh). This new approach allows for further exploration of fundamental questions regarding how infants learn language, a crucial aspect of human interaction.
Broader Impact: Though much research has been conducted in an effort to better understand the process of language acquisition, many fundamental questions remain without a definitive answer, among them the issues of the timeline of infant language ability development and the nature of the learning mechanism that allows infants to complete the enormous task of learning a complete language grammar from a continuous acoustic signal within just a few years. Investigating these questions will further our understanding of how humans learn language, a crucial aspect of human interaction. Intellectual Merit: Previous research has enabled the compilation of a basic timeline for the acquisition of various kinds of linguistic knowledge. We used corpus analysis and behavioral experiments with adults to get at which of two main learning mechanisms, proposed by researchers and supported by experimental evidence, might be aiding the infants at this early age: an unlearned bias in favor of phonetic principles, or statistical induction abilities. First, with corpus analyses of adult and infant-directed speech, we showed that in English, initial voiceless fricatives (e.g., s, f) are approximately 5-10 times more common at the beginning of words than voiced fricatives (e.g., z, v). Then we used the Headturn Preference Procedure, an experimental paradigm that has been successfully used by many researchers to probe infant reactions to different types of stimuli, without prior training. We showed that infants aged 4- to 10-months showed a preference for listening longer to voiceless fricatives. Further, infants do not extend this preference to Polish voiced and voiceless fricatives, indicating that the statistical knowledge is not generalized to the class of all fricatives. Finally, we show that this preference can be erased when we test infants on just the dental fricatives in English. In English, although fewer words start with dental fricatives, voiced dental fricatives are 7 times more frequent because they form onsets of function words. These results provide support for a role for statistical induction abilities in how infants learning of phonotactics.
