New Words
Learning new words despite obstacles
Published
May 2020
Team Members
Regina Calloway, Natasha Tokowicz, Marc Coutanche
Context
LeNS Lab, University of Pittsburgh
Role
I acted as our team’s first author on this project. I helped with the conceptualization of the project idea and research methods, collected the data with participants at all stages of the project (behavioral, fMRI, EEG), cleaned and analyzed the data, and wrote the original draft of the published paper.
Learn More
Check out the full publication!
Summary
One challenge to learning is that each learning experience does not look exactly like the last. To make things harder, once information is learned, it also often needs to be remembered again much later. In this experiment, we looked at how people are able to learn new information even though the information looks a little different each time it is experienced. In a world that is everchanging, this research speaks to how humans can overcome variability to learn and remember new information.
THE PROBLEM
Before we ran this study, there were a few things we knew from previous literature.
1) Concepts exist at many levels, but we sometimes only remember some of them. You can remember your friend has a dog, but forget their name. You can be aware that your dog is a mammal but not consciously retrieve that information when you think about her.
What we didn’t know was how these levels uniquely support memory for these concepts. For the sake of simplicity, we focused our investigation of this question on “item” and “category” levels.
Special thanks to my girl, Gracie, for modeling this idea for us!
2) When people learn new information, people tend to remember that information better later on if neural patterns are consistent each time the information is repeated. We call these patterns “robust” - the is brain re-entering a similar state each presentation. In addition, consistent neural patterns between the moment of learning and remembering - what we call “encoding-retrieval similarity” - can also predict how well someone remembers the information.
For this study, we wanted to examine a) how effectively these robust and reinstated patterns could stand up to perceptual changes between each presentation of the new information, b) if the patterns would predict memory performance even when people remembered the information a month after learning, and c) whether these relationships differed depending on whether the patterns were measured at item and category levels.
THE PROCESS
We used a lab study to address our questions. While scanning participants’ brain activity using fMRI, we asked participants to try to learn the names for strange animals in session I. Each time the animal was presented with its name, it was in a different position, similar to how encounters in real life rarely match past encounters exactly. About a month later, in session III, we measured neural activity again while we tested participants’ memories for the names of the animals.
THE DATA ANALYSIS
I conducted multivariate analyses of neuroimaging data from our fMRI scan to test for similar brain patterns. These analyses were conducted so we could calculate 4 main factors:
Finally, I conducted linear mixed effects regression models to examine if the significant brain patterns we measured might be important for predicting how well people remembered the animal names. Change in AIC values were used to determine significance, where the null model contained all variables aside from the neural measure of interest.
THE TAKEAWAYS
Despite perceptual changes between presentations of animals during learning, the brain can create robust patterns to aid in category-level memory, even with a month’s delay.
Robust patterns are important for predicting memory performance - finding strategies for building or maintaining this robustness might be important for those looking to support memory.