Lexical and context effects in children’s audiovisual speech recognition

Abstract

The Audiovisual Lexical Neighborhood Sentence Test (AVLNST), a new, recorded speech recognition test for children with sensory aids, was administered in multiple presentation modalities to children with normal hearing and vision. Each sentence consists of three key words whose lexical difficulty is controlled according to the Neighborhood Activation Model (NAM) of spoken word recognition. According to NAM, the recognition of spoken words is influenced by two lexical factors: the frequency of occurrence of individual words in a language, and how phonemically similar the target word is to other words in the listeners lexicon. These predictions are based on auditory similarity only, and thus do not take into account how visual information can influence the perception of speech. Data from the AVLNST, together with those from recorded audiovisual versions of isolated word recognition measures, the Lexical Neighborhood, and the Multisyllabic Lexical Neighborhood Tests, were used to examine the influence of visual information on speech perception in children. Further, the influence of top-down processing on speech recognition was examined by evaluating performance on the recognition of words in isolation versus words in sentences. [Work supported by the American Speech-Language-Hearing Foundation, the American Hearing Research Foundation, and the NIDCD, T32 DC00012 to Indiana University.]