Brain-imaging evidence for compression of binary sound sequences in human memory.

Area of Science:

• Cognitive Neuroscience
• Neuroscience of Memory
• Computational Linguistics

Background:

• The language-of-thought hypothesis posits that regular sequences are compressed in memory via recursive loops, functioning like a mental predictive program.
• Understanding the neural mechanisms underlying sequence memory and prediction is crucial for cognitive science.

Purpose of the Study:

• To test the language-of-thought hypothesis by examining how the human brain processes and predicts regular sequences of varying complexity.
• To identify the brain regions involved in encoding and predicting structured sequences and their relationship to mathematical processing.

Main Methods:

• Functional magnetic resonance imaging (fMRI) and magnetoencephalography (MEG) were used to record brain activity.
• Participants listened to 16-item sequences of two sounds with hierarchical structures of increasing complexity.
• Deviant sounds were introduced to probe participants' knowledge and predictive capabilities regarding the sequences.

Main Results:

• Brain activity and task difficulty correlated with the complexity of sequence structures, as defined by minimal description length.
• Neural activity increased with complexity for learned sequences and decreased for deviant sounds, supporting predictive coding.
• fMRI and MEG data revealed that sequence prediction accuracy decreases and latency increases with rising structural complexity.

Conclusions:

• The findings support the language-of-thought hypothesis, indicating that the brain encodes regular sequences using recursive structures.
• Bilateral superior temporal, precentral, anterior intraparietal, and cerebellar cortices are implicated in processing and predicting complex sequences.
• These brain regions show significant overlap with areas involved in mathematical calculation, suggesting a shared neural basis for structured sequence processing.