The problem of perceiving meaning in speech is one of the most challenging problems in cognitive science. Even something as simple as distinguishing “b” from “p” requires listeners to combine dozens of sources of information and these cues are heavily context dependent and noisy. And all this needs to be done 5-6 times a second and in real-time as the auditory signal rolls in.
Lexical and Perceptual Development
While much of our work in the MAClab concerns auditory and perceptual aspects of words, the process of mapping sounds to meaning offers some conceptually similar problems.
Individual Differences & Clinical Populations
Much research on speech perception and word recognition focuses on “normal” individuals - people who hear well, and have normal language abilities. Such research has revealed a great deal about the general characteristics of speech processing. From the perspective of this group of listeners, speech appears to be processed effortlessly and accurately, despite the challenging nature of the problem.
The human brain has the remarkable ability to process speech and language cues that often occur in just fractions of a second. While our eye-tracking techniques can often measure very early processes, they leave open the neural mechanisms that may underlie speech perception, and they can really only measure representations that are relatively late in speech processing like words.
Early in children’s schooling, they are faced with the challenge of learning to read. One of the most important early skills that children must master is “decoding”, the ability to map spelling to sound, e.g., to pronounce a string of letters.
Traditionally, psychology has focused on observable behaviors and how they changes with experimental manipulations in order to understand the underlying mechanisms of cognition, perception and language. More recently, we've turned to neuroscience to enrich these explanations.
Phonetics and Laboratory Phonology
One of the most important advantages we have in our work is the fact, that we have a very good understanding of what is in the speech signal - we know how to measure cues like voice onset time (VOT) and formant frequencies that listeners are likely to use to identify phonemes and words.
Broader Mechanisms: Learning, Competition & Categorization
Ultimately our approach to language stresses mechanisms like associative learning, real-time competition and categorization. These are not specifically tuned to language, but representative of broader psychological processes that apply in diverse domains.