Dynamics of working memory process revealed by independent component analysis in an fMRI study

Magdalena Fąfrowicz, Anna Ceglarek, Justyna Olszewska, Anna Maria Sobczak, Bartosz Bohaterewicz, Monika Ostrogórska, Patricia A. Reuter‐Lorenz, Koryna Lewandowska, Barbara Sikora-Wachowicz, Halszka Ogińska, Magdalena Hubalewska‐Mazgaj, Tadeusz MarekScientific 2023 · DOI

TL;DR

This study uses fMRI and Independent Component Analysis (ICA) to explore how our brains form false memories in short-term memory tasks, finding that a less engaged medial visual network during encoding is linked to these errors. It also reveals that the time of day influences brain activity during memory retrieval, with distinct patterns observed in the evening.

What Problem Does This Paper Address?

We all make memory errors, from forgetting where we put our keys to misremembering foreign language words. This paper delves into the neural mechanisms behind these "false memories," specifically in short-term memory. The researchers are interested in how phonologically similar but meaningless words (like pseudo-words in a new language) can lead to errors, and how these processes might differ depending on whether the task is performed in the morning or evening. Understanding these dynamics is crucial for a more complete picture of working memory and memory distortion.

Methods

The study employed a modified Deese-Roediger-McDermott (DRM) paradigm, a well-established method for inducing false memories. Participants performed this task twice, once in the morning and once in the evening, while undergoing fMRI.

The core analytical techniques were:

  • Independent Component Analysis (ICA): This data-driven approach was used to identify distinct, functionally connected brain networks (Intrinsic Connectivity Networks or ICNs) that were active during different phases of the memory task (encoding and retrieval) and for different response types (hits, false alarms, correct rejections). ICA is valuable here because it can reveal complex, distributed patterns of brain activity that might be missed by traditional methods.
  • General Linear Model (GLM): This standard fMRI analysis technique was used to examine specific contrasts between conditions (e.g., correct rejection of lures vs. false alarms) and to investigate differences in brain activation between morning and evening sessions.

Datasets

The study did not introduce new datasets or benchmarks. The data used was collected from 58 healthy participants who performed a specifically designed experimental task.

Key Insights and Ideas

  • Medial Visual Network's Role in False Memory: A key finding is that reduced activity in the medial visual network during the encoding phase of the task was associated with increased false memory formation (false alarms). This suggests that this network is crucial for accurately representing stimuli, and its under-engagement leads to memory distortions.
  • Time of Day Matters for Retrieval: Diurnal rhythmicity significantly impacts memory retrieval. Specifically, the default mode network (DMN) and the medial visual network showed less deactivation in the evening compared to the morning during the retrieval of correct rejections of lure probes. This indicates that our internal biological clock influences how our brains process and retrieve memories.
  • Evening Brain Activity Differences: The GLM analysis revealed greater activation in the right lingual gyrus (part of the visual cortex) and the left cerebellum in the evening session, particularly when distinguishing between correct rejections of lure probes and false alarms. This highlights specific brain regions that are more active or differently engaged during evening memory processing.
  • Dynamic Nature of Working Memory: The findings support the idea that working memory is not static but dynamic, with different brain networks being recruited and modulated depending on the task demands (encoding vs. retrieval) and internal states (time of day).

Key Results

  • Encoding Phase: ICA revealed that the medial visual network was active before correct recognition of positive probes and correct rejection of lure probes. However, this network's engagement was not observed before false alarms. Executive, right fronto-parietal, and sensory-motor networks were generally activated, while anterior DMN, auditory, and dorsal attention networks were deactivated across conditions.
  • Retrieval Phase: Executive and right fronto-parietal networks were activated, and visual networks were deactivated. No distinguishing networks were found for false alarms during retrieval.
  • Diurnal Differences (MS vs. ES):
    • The anterior and posterior DMN, and the medial visual network, showed more deactivation in the morning session compared to the evening session during the correct rejection of lure probes.
  • GLM results showed greater activation in the right lingual gyrus and left cerebellum in the evening session for the contrast of correct rejection of lure > false alarms.
  • Behavioral Results: Reaction times varied significantly across different response types, with false alarms and correct rejection of lure probes taking longer than correct rejection of negative probes.

Limitations and Future Work

  • Fatigue Measurement: The authors acknowledge that the lack of objective fatigue measurement is a limitation, as fatigue could influence the observed effects of homeostatic processes.
  • FDR Correction: While standard, the use of False Discovery Rate (FDR) correction is noted as less conservative than Family-Wise Error (FWE) correction.
  • Unrelated Probes: The study couldn't analyze false alarms to unrelated probes due to their low occurrence rate, which limits comparisons to related vs. unrelated probes.
  • Sex Differences: The authors point to recent research suggesting sex differences in neural correlates of retrieval processes, which could be a direction for future studies.
  • Veridical vs. False Memory: Future work could explore if similar network patterns are observed with meaningful verbal material.
  • Stimulus Modality: The study used visual stimuli; future research could investigate if these findings generalize to auditory stimuli.

Summary updated

Abstract

Human memory is prone to errors in many everyday activities but also when cultivating hobbies such as traveling and/or learning a new language. For instance, while visiting foreign countries, people erroneously recall foreign language words that are meaningless to them. Our research simulated such errors in a modified Deese-Roediger-McDermott paradigm for short-term memory with phonologically related stimuli aimed at uncovering behavioral and neuronal indices of false memory formation with regard to time-of-day, a variable known to influence memory. Fifty-eight participants were tested in a magnetic resonance (MR) scanner twice. The results of an Independent Component Analysis revealed encoding-related activity of the medial visual network preceding correct recognition of positive probes and correct rejection of lure probes. The engagement of this network preceding false alarms was not observed. We also explored if diurnal rhythmicity influences working memory processes. Diurnal differences were seen in the default mode network and the medial visual network with lower deactivation in the evening hours. The GLM results showed greater activation of the right lingual gyrus, part of the visual cortex and the left cerebellum in the evening. The study offers new insight into the mechanisms associated with false memories, suggesting that deficient engagement of the medial visual network during the memorization phase of a task results in short-term memory distortions. The results shed new light on the dynamics of working memory processes by taking into account the effect of time-of-day on memory performance.
BibTeX
@inproceedings{fafrowicz2023dynamics,
  author = {Magdalena Fąfrowicz and Anna Ceglarek and Justyna Olszewska and Anna Maria Sobczak and Bartosz Bohaterewicz and Monika Ostrogórska and Patricia A. Reuter‐Lorenz and Koryna Lewandowska and Barbara Sikora-Wachowicz and Halszka Ogińska and Magdalena Hubalewska‐Mazgaj and Tadeusz Marek},
  title = {{Dynamics of working memory process revealed by independent component analysis in an fMRI study}},
  booktitle = {Scientific},
  year = {2023},
  doi = {10.1038/s41598-023-29869-2},
}
Figure