Distributed Activity Patterns for Objects and Their Features: Decoding Perceptual and Conceptual Object Processing in Information Networks of the Human Brain
Neuroscience and Neurobiology
How are object features and knowledge-fragments represented and bound together in the human brain? Distributed patterns of activity within brain regions can encode distinctions between perceptual and cognitive phenomena with impressive specificity. The research reported here investigated how the information within regions' multi-voxel patterns is combined in object-concept networks. Chapter 2 investigated how memory-driven activity patterns for an object's specific shape, color, and identity become active at different stages of the visual hierarchy. Brain activity patterns were recorded with functional magnetic resonance imaging (fMRI) as participants searched for specific fruits or vegetables within visual noise. During time-points in which participants were searching for an object, but viewing pure noise, the targeted object's identity could be decoded in the left anterior temporal lobe (ATL). In contrast, top-down generated patterns for the object's specific shape and color were decoded in early visual regions. The emergence of object-identity information in the left ATL was predicted by concurrent shape and color information in their respective featural regions. These findings are consistent with theories proposing that feature-fragments in sensory cortices converge to higher-level identity representations in convergence zones. Chapter 3 investigated whether brain regions share fluctuations in multi-voxel information across time. A new analysis method was first developed, to measure dynamic changes in distributed pattern information. This method, termed "informational connectivity" (IC), was then applied to data collected as participants viewed different types of man-made objects. IC identified connectivity between object-processing regions that was not apparent from existing functional connectivity measures, which track fluctuating univariate signals. Collectively, this work suggests that networks of regions support perceptual and conceptual object processing through the convergence and synchrony of distributed pattern information.