Speech Stream Tracking in 2D: Attention Differentially Enhances Acoustic and Phonemic Encoding Across Spatial Planes

Selective attention enables the flexible allocation of neural resources toward relevant stimuli. In audition, it allows listeners to track a target stream within acoustically complex environments. This raises the question of how strongly auditory attention depends on spatial cues to achieve stream segregation and maintain distinct auditory objects. Evidence shows that attention enhances neural encoding of sound features for streams separated in azimuth. However, it remains unclear whether the same mechanisms apply without binaural cues in elevation, and how attention prioritizes acoustic versus linguistic features under these conditions. To address this, participants listened to arrhythmic streams of digits spoken by the same voice and separated in azimuth or elevation. They attended one stream to detect target numbers while ignoring the other. Neural responses to attended and ignored streams were modelled using envelope and phoneme temporal response functions, allowing comparison of low-level envelope and higher-level phonemic encoding across spatial dimensions. Results revealed distinct feature-weighting profiles across spatial planes. In azimuth, selective attention was primarily supported by enhanced envelope encoding of the target stream at early and middle latencies. In elevation, envelope encoding was reduced, while phoneme encoding exhibited a more widespread attentional modulation. These findings suggest that phonemic representations support selective stream tracking when binaural spatial cues are unavailable, reflecting flexible weighting of acoustic and phonemic information depending on spatial cue availability.