Effects of Neighborhood Density and Noise on Children's Word Learning

Abstract

Studies show that words are organized with similarity neighborhoods based on similar sound structure. Some words have many similar sounding words, while others have few. The number of neighbors a word has is called neighborhood density, which is known to influence word learning. Specifically, words with few neighbors are learned more accurately early in training perhaps because these words play a role in triggering the learning of a novel word. In contrast, words with many neighbors are learned more accurately later in training and post training perhaps because these words play a role in the construction of a new lexical representation in long-term memory and in the connection of the newly constructed lexical representation with existing representations (Storkel, Armbrüster, & Hogan, 2006; Storkel, Bontempor, Aschenbrenner, Maekawa, & Lee, 2013; Storkel & Lee, 2011). However, these findings were obtained in a quiet listening condition, providing little information about the effect of the environment where word learning typically takes place. The goal of this study was to examine whether noise alters the effect of neighborhood density on word learning. Seventy-seven typically developing 4- and 5-year-old preschool children were randomly assigned to one of three listening conditions: 0dB, +6dB, and +15dB signal-to-noise ratio (SNR). Sixteen consonant-vowel-consonant nonword-novel object referent pairs were embedded in two stories for training; neighborhood density for the nonwords varied from low to high. Nonword stimuli and audio narrative scripts for stories were digitally mixed with broadband white noise at 0dB, +6dB, and +15dB SNR. Learning was measured using a picture naming task and a referent identification task. Six cycles of story training-measures of learning were completed with two no training points each after the third and sixth measures of learning. Logistic multi-level modeling (MLM) revealed different patterns of word learning depending on the tasks. Only in the naming task, a significant effect of noise and an interaction between noise and neighborhood density were found at +6dB SNR compared to 0dB SNR. Specifically, results showed that (1) word learning was better at 6dB SNR than 0dB SNR; (2) no significant effect of density was found and this non-significance persisted over time. However, the high density advantage started to emerge at +6dB SNR and +15dB SNR; and (3) the difference between +6dB SNR and 0dB SNR was greater as density increased. In addition, in both naming and referent identification tasks, word learning increased over time with significant forgetting of words in the naming task and a trend of memory consolidation in the referent identification when no training was occurred. These results provide the evidence that word learning declines as listening environment worsen. The results indicate that noise hinders children's ability to use lexical representations, which adversely influences the whole process of word learning (i.e., triggering, configuration, and engagement). The results also imply that high density words are more sensitive to listening condition than low density words. In addition, the naming task that requires more detailed lexical representation is more sensitive to noise than the referent identification task.