Editorial: Neural Mechanisms of Perceptual Categorization as Precursors to Speech Perception

This research topic describes recent advances in understanding the brain functional organization for sensory categorization along with its implications for speech perception. Among the 14 papers, one theme is how neural representations of auditory and visual input are transformed across different scales of neural organization to enable speech perception, and another is the neural mechanisms of category learning. In the first theme, several animal and human studies delve into the complex hierarchical organization of auditory ventral pathways for speech perception. Prior work has established an important role for the auditory ventral stream in complex sound categorization (Rauschecker and Scott, 2009; Romanski and Averbeck, 2009). In humans, a preference has convincingly been demonstrated for phonemic over non-phonemic sounds in non-primary auditory fields in the middle of the ventrolateral superior temporal cortex (mSTG/S) (Liebenthal et al., 2005; Leaver and Rauschecker, 2010). The present papers contribute novel insights about the function of dorsal areas in and near the auditory core, the functional specificity of the mSTG/S, and the role of non-auditory areas, for phonemic perception. Collectively, they suggest that multiple stages of abstraction from the original form of speech occur in low-level sensory cortices. In the mSTG/S, the neural representations are highly specific to phonemic categories. Tsunada and Cohen's review of research in the monkey suggests that single neurons in the auditory core encode categories for simple sounds (e.g., direction of spectral changes), whereas neurons in the auditory belt encode more complex categories (including speech phonemes) based on input from the entire population of core neurons. At the cellular level, they report the intriguing finding that different classes of neurons within the auditory belt may have different sensitivity to category information: The more common pyramidal neurons encode auditory categories with less sensitivity than the less common interneurons (Tsunada et al., 2012). Astikainen et al. also show that in anesthetized rats' primary auditory cortex, neurons automatically encode structural patterns (order of syllable repetition) from a fast paced speech stream and generalize to novel patterns. Based on intracranial high-gamma electrophysiological recordings in subjects with intractable epilepsy, Steinschneider et al. propose that within 200 ms, activity in the human primary and non-primary auditory cortices reflects non-categorical spectrotemporal sound attributes. Only later, activity in non-primary auditory areas receiving modulatory input from higher-order, lexico-semantic associative cortex represents phoneme categories.