How do we recognize pseudowords in an audio signal?

  • Author(s) / Creator(s)
  • A number of speech perception studies have been carried out to investigate how we process audio signals containing real words. However, comparatively fewer studies have been conducted looking at how listeners process audio signals containing phonotactically legal pseudowords. Some traditional metrics, such as lexical frequency, that are used as predictors in this kind of analysis are difficult or impossible to calculate for pseudowords, but other metrics like phonotactic probability, phonological neighborhood density, and uniqueness point can be. Phonotactic probability is the likelihood that a particular sequence of segments occurs in a particular language, and it can be calculated using existing pronunciation and lexical frequency data. The present study uses the CMU Pronouncing Dictionary and the Google Ngram datasets to calculate phonotactic probability measures for each stimulus in an existing set of English auditory lexical decision responses and then statistically models the influence of phonotactic probability, phonological neighborhood density, and uniqueness point on the participants' reaction times to the stimuli. This modeling shows general increasing trends for each of the predictors of interest, where a high value is correlated with longer response times. These results are then framed in the greater picture of speech perception and word recognition overall. Originally presented at the 173rd Meeting of the Acoustical Society of America.

  • Date created
  • Subjects / Keywords
  • Type of Item
    Conference/Workshop Poster
  • DOI
  • License
    Attribution-ShareAlike 4.0 International