Real-world memory and the brain
How do we construct and retrieve memories of complex real-world episodes? In this research we use realistic stimuli (such as movies and narratives) and behaviors (such as spoken recall) that contain rich natural semantics and unfold continuously across multiple timescales. Employing between-brain temporal and pattern analysis methods, we ask how mnemonic and sensory systems operate together dynamically to create the present moment.
I am recruiting postdocs!
Contact me by email if interested.
In the mind, the present moment is a convergence point of two information streams: one, a continuous flow of sensory input from the outside world; and two, a series of elements from our past experiences, i.e., memories. Memories may be triggered by sensory stimuli, they may themselves cue more memories, and they may change the way incoming stimuli are interpreted, all of which become part and parcel of our current experience.
Past information casts an influence across multiple timescales: events that occurred a moment ago, a minute ago, and a day ago may all impact the present. In order to understand how the mind and brain work, we need an account of how memories of past events, across multiple timescales, continuously influence and merge with ongoing perception and behavior.
Studying real memory requires using real stimuli. Scientists often trade realism for control; we use lists or configurations of random items, attempting to isolate selected variables. However, this approach can strip away the very richness and complexity that made memory such a compelling topic in the first place, and cause us to neglect phenomena that emerge only when stimuli are as dynamic and detailed as the real world.
My work aims to understand how we construct and retrieve memories of complex real-world episodes. I use realistic stimuli (such as movies and narratives) and behaviors (such as spoken recall) that contain rich natural semantics and unfold continuously across multiple timescales. Using novel between-brain temporal and pattern analysis methods, I ask how mnemonic and sensory systems operate together dynamically to create the present moment.
Shared experience, shared memory. Patterns in the brain which emerge during perception are later reactivated during spoken recall, are robustly similar across different individuals, and transform systematically between perception and memory.
A hierarchy of processing timescales. In order to interpret a continuous stream of input from the world, the brain must integrate information over multiple timescales. We propose that stimulus processing is distributed across a hierarchy of cortical regions, with processing timescales increasing along a gradient from low-level sensory areas (e.g., visual cortex) up to high-level association areas (e.g., default network).
[Hierarchical process memory: memory as an integral component of information processing. U Hasson, J Chen, CJ Honey; Trends in cognitive sciences 19 (6), 304-313.] [Processing timescales as an organizing principle for primate cortex (commentary). J Chen, U Hasson, CJ Honey; Neuron 88 (2), 244-246.] [How long is now? The multiple timescales of language processing (commentary). CJ Honey, J Chen, K Müsch, U Hasson; The Behavioral and brain sciences 39, e77.]
Under natural conditions, memories can persist in high-order cortex for minutes. It is well known that formation of new episodic memories depends on the hippocampus, but in real- life settings (e.g., conversation), hippocampal amnesics can utilize information from minutes earlier. What neural systems outside the hippocampus support this minutes-long retention? My work using functional MRI in the healthy brain suggests that default network cortical regions can intrinsically retain information for several minutes during continuous, semantically rich natural stimulation.
Analyses of functional neuroimaging data
Naturalistic stimuli and behavior
Collect brain data as people watch movies and listen to stories, and as they describe their memories out loud
Compare activity between the brains of different people, both in the temporal domain and the spatial domain
Multi-voxel pattern classification
Identify neural patterns that are specific to certain periods in the stimulus, such as a particular movie scene
Semantic model construction
Predict patterns of neural activity given combinations of stimulus features
Currently hiring! Contact me by email for more information.
Applications are invited for a postdoctoral researcher to study real-world memory and perception in the human brain, using fMRI, computational, and behavioral methods. Specific projects are flexible and may focus on memory search and recollection, transformations between perceptual and mnemonic brain states, mapping dimensionality and stimulus representations in high-level cortex, or related topics.
Start date is flexible. Funding is available for three years. The initial appointment will be for one year, with the opportunity for renewal.
Experience with fMRI is required, as well as proficiency in MATLAB or Python.