Inducing a mental context for associative memory formation with real-time fMRI neurofeedback

Silvy Collin, Philip van den Broek, Tim van Mourik, Peter Desain, Christian F. Doeller

Abstract

Memory, one of the hallmarks of human cognition, can be modified when humans voluntarily modulate single neuron or neural population activity using neurofeedback. However, it is currently unknown whether memory is facilitated or impaired after such neural perturbation. In this study, participants memorized objects while we trained them with abstract neurofeedback to modulate their brain activity patterns in the ventral visual stream to represent either faces or houses. The neurofeedback created an implicit face or house context in the brain while memorizing the objects. The results revealed that participants created associations between each memorized object and its implicit context solely due to the neurofeedback manipulation. Our findings shed light onto how memory formation can be influenced by synthetic memory tags with neurofeedback and advance our understanding of mnemonic processing.

Introduction

Memory is rarely a standalone recording of an event; it is fundamentally tied to the context in which it was formed. This principle, known as "encoding specificity," suggests that external environments—like a specific room or a background song—become part of the memory trace. However, the role of internal, purely mental context is more difficult to isolate and manipulate.

This study investigates whether technology can bypass external stimuli to directly engineer an internal mental state during memory formation. Using real-time fMRI (rt-fMRI) neurofeedback, researchers aimed to create a "synthetic" mental context—specifically, category-level representations of faces or houses—and determine if these internal states would implicitly bind to objects memorized during the session.

Methods and Experimental Design

The researchers employed a sophisticated closed-loop system involving 20 healthy human participants. The experimental protocol was divided into three distinct phases:

  1. Classifier Training: Participants viewed images of faces and houses while in the fMRI. A machine learning classifier was trained to identify the unique neural signatures of these categories within the ventral visual stream (fusiform and parahippocampal gyri).
  2. Neurofeedback Blocks: During these trials, participants were shown an image of an everyday object (e.g., a tool) followed by an abstract feedback signal (a thumb icon). Participants were instructed to make the thumb grow greener and more visible by "thinking the right way," without being told the target was a specific category. In real-time, the fMRI data was fed to the classifier; if the participant's brain activity matched the target category (face or house), the feedback became more positive.
  3. Behavioral Memory Session: Post-scan, participants underwent a context test to see if they could identify which category (face or house) was associated with the objects seen during neurofeedback, followed by an associative learning task.
Figure 1: The experimental timeline showing the progression from training to neurofeedback blocks and the final behavioral memory sessions.

Key Results

Successful Neural Modulation

The neurofeedback mechanism proved effective. By the end of the training blocks, participants demonstrated above-chance classifier evidence for the target category (Overall mean: $T(1,18) = 2.203, p = 0.041$). This indicates that participants successfully learned to voluntarily modulate their neural representations to match the "synthetic" context required by the feedback loop.

Implicit Context Binding

The most significant finding was the successful integration of this synthetic context into the participants' memories. In the post-scan behavioral test, participants were able to correctly identify which objects were associated with which neurofeedback-induced context ($T(1,18) = 3.007, p = 0.008$). Remarkably, this association was formed implicitly; participants were never explicitly informed about the face or house categories during the feedback phase.

Figure 5: Behavioral results indicating that participants successfully associated memorized objects with the implicit neurofeedback context (face vs. house).

The Associative Learning Null Result

While the synthetic context successfully "tagged" the objects, it did not significantly impact a subsequent explicit associative learning task ($p = 0.659$). Memory performance remained consistent regardless of whether a new face/house pair matched the previously induced synthetic context. The researchers attribute this to a "ceiling effect," where the task may have been too easy for the subtle influence of the neurofeedback context to manifest in the final scores.

Discussion and Conclusion

This study demonstrates that the neural underpinnings of memory context are remarkably malleable. By using non-invasive brain-computer interfaces, it is possible to "inject" a category-specific tag into a memory trace. The act of searching for the correct mental state to satisfy the neurofeedback loop appears to be a powerful mechanism for binding that state to concurrent information processing.

Future research may focus on increasing task difficulty to overcome ceiling effects and exploring clinical applications. For instance, this technique could potentially be used to "re-tag" traumatic memories with neutral contexts or to strengthen specific associations in patients with memory disorders. Ultimately, these findings advance our understanding of how synthetic memory tags can influence the architecture of human memory.

Summary updated

Abstract

Abstract Memory, one of the hallmarks of human cognition, can be modified when humans voluntarily modulate single neuron or neural population activity using neurofeedback. However, it is currently unknown whether memory is facilitated or impaired after such neural perturbation. In this study, participants memorized objects while we trained them with abstract neurofeedback to modulate their brain activity patterns in the ventral visual stream to represent either faces or houses. The neurofeedback created an implicit face or house context in the brain while memorizing the objects. The results revealed that participants created associations between each memorized object and its implicit context solely due to the neurofeedback manipulation. Our findings shed light onto how memory formation can be influenced by synthetic memory tags with neurofeedback and advance our understanding of mnemonic processing.
BibTeX
@inproceedings{collin2022inducing,
  author = {Silvy Collin and Philip van den Broek and Tim van Mourik and Peter Desain and Christian F. Doeller},
  title = {{Inducing a mental context for associative memory formation with real-time fMRI neurofeedback}},
  booktitle = {Scientific},
  year = {2022},
  doi = {10.1038/s41598-022-25799-7},
}
Figure