Sound blending in the brain: a functional magnetic resonance imaging investigation

The presence of high levels of background noise is a serious concern for functional magnetic resonance imaging studies of phonological processing using conventional methods. As a result, many such studies have focused on phonological units larger than phonemes (e.g. syllables) or used stimuli presented in the visual (e.g. printed letters) rather than the auditory domain. We used a recently developed functional magnetic resonance imaging method to present spoken stimuli without the scanner's background noise. Young adult participants mentally blended phonemes in a series (e.g. /b/, /ae/, /t/), counted the number of discrete tones, or rested. Relative to tone counting, sound blending elicited activation in bilateral temporal and prefrontal cortices with left asymmetry. Activation within the dorsoposterior inferior frontal gyrus, a subregion of Broca's area, was negatively correlated with sound-blending accuracy. Our findings are consistent with prior studies ascribing a role of general sequencing, motor and articulatory programming, and vocal or subvocal articulatory rehearsal to this brain region.