Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Is This the Real Life? Is This Just Fantasy? A Domain Adaptation Framework for Identifying Higher-Order Reality Representations

Ali Ekhlasi¹, Emil Olsson¹, Michaela Klímová², Xueyi Huang², Angela Shen¹, Nadine Dijkstra³, Jorge Morales², Megan AK Peters¹; ¹University of California, Irvine, ²Northeastern University, ³University College London

Presenter: Ali Ekhlasi

How does your brain know what’s real? Neural responses to perceived versus imagined content often seem to overlap in their neuroanatomical loci and representational structure, but the brain is nevertheless able to differentiate between imagery and perceptual input. How does the brain engage in this reality monitoring capacity, and what can this tell us about the mechanisms behind the conscious experience of seeing? A class of theories of consciousness---Higher-Order (HO) theories---posits that conscious experience depends on the formation of higher-order representations, which monitor and interpret first-order (perceptual and imagined) representations. Yet HO theories differ in their positioning on the nature of such HO representations. Here, we develop and validate a computational approach to interrogating the nature of HO representations by using domain adaptation (DA), a machine learning technique that can statistically align data distributions from different sources. We apply DA to ask when and where patterns of human fMRI BOLD responses can be aligned across perception and imagination, and across contents, to reveal areas that encode content-agnostic reality signals, reality-agnostic content signals, and signals of one type that are conditioned on the other, thereby providing differential support for specific aspects of competing HO theories. We find content-conditioned reality signals in the precuneus and posterior cingulate cortex, while content-agnostic reality signals appear more in the occipital fusiform and occipital pole. In contrast, content representations specific to perception versus imagination appeared in the precuneus and intracalcarine cortex, while the temporal occipital fusiform cortex appeared to encode content regardless of its source. These results suggest highly localized and separable computations may collectively drive the overall assessment of whether an internal representation is real or imagined. Further, these results demonstrate the viability, promise, and power of DA approaches to interrogating the inputs to HO neural representations.

Topic Area: Methods, Tools, Theory & Neural Coding