Object-Based Attention in Naturalistic Auditory Streams

In each environment we experience in our daily life, we find ourselves surrounded by complex auditory soundscapes. These auditory scenes are often composed of many concurrent sound sources with both spatially and temporally overlapping spectral details, which humans are consistently able to parse seemingly effortlessly in order to navigate their surroundings. Two main aspects are involved in this capability: directing attention, first outlined by Cherry in 1953 in the form of a seminal paradigm, the so-called “cocktail party problem”, and parsing the sound scene, codified by Bergman in 1990 as “auditory scene analysis”. Attention, in every sensory domain, plays a fundamental role to efficiently select the relevant information and ignore the distracting inputs, with studies showing that it operates in the form of a “biased competition” between neural representation of perceptual objects. Objects were depicted in vision as the central “units”, on which non-spatial selective attention acts in many natural contexts, however in the auditory domain several questions remain to be addressed on that matter. Most notably it is indeed unclear how the attention mechanisms operate at both, the level of the auditory object formation and the level of the auditory object selection. More importantly where and how in the neural pathway an object-based representation of an attended sound emerges, is yet to be fully understood. In the present thesis we designed a novel paradigm to tackle the auditory selective attention at the level of object processing: a repetition target has been embedded in one of the two naturalistic streams in such a way that, after being cued, the participant necessarily had to build up representations of the auditory objects composing the scene across time in order to accomplish the task. This paradigm has been employed in both studies that compose the thesis. The first study was conducted to characterize the auditory selective attention system with a theoretical and empirical focus on high-level attentional modulations on the processing level of auditory objects. In line with studies of object-based attention in the visual domain, we reported a behavioral attentional facilitation effect in the validly cued trials, and an inhibition effect in the invalidly cued trials. In the second study, we focused on the neural activity during the repetition detection period by investigating the temporal dynamic of the cortical activity at the source level. These analyses were accomplished both in time (in form of evoked responses, ERFs) and frequency domain (in form of time-frequency spectrograms) providing not only insights into the time courses of events but also their exact spatial localization in cortex. We reported a significant stronger MEG response, mapped at the source cortical level, when the repetition segment was in the attended stream. The same has been found when the target repetition was detected, suggesting that the time course of the neural activity and its spatial distribution represent one aspect of the neural correlates of the object-based attention. Overall both studies show an attentional modulation effect operating indeed on the object-level in a naturalistic auditory scene supporting the biased competition theory.