Establishing ground truth about the development of episodic memory in infancy
Grantee: Yale University
Grant Details
| Project Lead | Nicholas Turk-Browne Ph.D. |
|---|---|
| Amount | $250,000 |
| Year Awarded | |
| Duration | 4 years |
| DOI | https://doi.org/10.37717/2020-1208 |
| Summary | Infants may hold the key for understanding human memory. They are prodigious learners, accumulating knowledge and abilities through experience that shape cognition and behavior throughout life. At the same time, memories for these formative infant experiences are eventually lost, a phenomenon known as infantile amnesia. Here we seek to address this conundrum by systematically characterizing the development of infant memory with theoretical ideas and methodological tools from cognitive neuroscience. We focus on a form of long-term memory known as episodic memory, the remarkable ability of humans to encode and retrieve autobiographical events that occurred even just once at a particular time and place. We test episodic memory for real-world events by providing each of 24 families with a phone to record a daily video of a distinctive event experienced by their infant, starting as early as three months old. Three months later and every three months thereafter until a total of five sessions, the family returns for a functional magnetic resonance imaging (fMRI) scan that assesses memory for preceding events. By staggering the starting age, we obtain continuous behavioral and fMRI data on memory retrieval from ages 6-20 months old. We focus on the function of the hippocampus, a critical brain region for episodic memory that fMRI has proven uniquely well suited to studying in adults. Although this technology is challenging to use in early development, over the past few years we have built a robust infrastructure for collecting and analyzing fMRI data from awake and behaving infants. In each fMRI session, infants will watch videos that they were not shown in a previous session spanning all preceding time periods. We will use cutting-edge computational and multivariate techniques from machine learning to quantify three component processes of episodic memory in eye-movement behavior and hippocampal activity patterns: relational binding (the ability to stitch different elements of a memory together across modalities and over space and time), pattern separation (the ability to store memories separately when they contain related elements to prevent interference), and pattern completion (the ability to reinstate missing elements of a memory from partial information). By examining how behavioral and neural markers of these processes decay as a function of the age of memories, we distinguish encoding, consolidation, and retrieval phases of infant memory formation. These decay functions can then be examined over the age of the child at retrieval, providing the first comprehensive dataset on the early development of episodic memory. The key advances of this work include: moving beyond traditional measures of adult episodic memory that rely on conscious verbal report and laboratory tests of infant memory that do not isolate episodic memory from reinforcement learning, conditioning, and priming; establishing the ground truth of what infants experience naturalistically and what they remember from these experiences; quantifying trajectories in how these experiences and memories change over early development; and providing the first data on the function, relationship to behavior, and early development of the hippocampus in human infants. |