Uncovering the representational geometry of durations

This study challenges the traditional unidimensional mental timeline by demonstrating that human duration perception is better explained by a multidimensional, helical geometry involving magnitude, contextual encoding, and periodicity, which is dynamically reflected in neural processing and constrained by individual endogenous oscillations.

The Gist

Imagine your brain doesn't just have a single ruler for measuring time. Instead, it has a complex, 3D playground where time lives. That's the big discovery from this new study.

For a long time, scientists thought our brains treated time like a simple straight line: a short second is at the start, and a long minute is at the end. It was a "mental timeline." But this paper suggests that reality is much more interesting, like a spiral staircase or a corkscrew rather than a flat line.

Here is a simple breakdown of how they found this and what it means.

The Experiment: Guessing the Shape of Time

The researchers wanted to map out this "mental playground." They did two things with 33 volunteers:

1. The "Feeling" Test: They played pairs of sounds (beeps) of different lengths (from 0.4 seconds to 2.2 seconds) and asked people: "How similar do these two feel?" If two beeps were close in length, people said they were similar. If they were far apart, they said they were different.
2. The "Brain Scan" Test: They put the same people in an EEG cap (a helmet that reads brain waves) and played the same sounds. This let them see what the brain was actually doing when it heard those durations.

The Discovery: Time is a 3D Spiral

When they analyzed the "feeling" data, they expected to see a straight line. Instead, they found three hidden dimensions that shape how we experience time:

1. The Magnitude (The Height): This is the obvious one. A 2-second beep feels longer than a 1-second beep. This is the "up and down" of the spiral.
2. The Context (The Center): Our brains are weirdly sensitive to the "average" of what we've just heard. If you hear a bunch of short beeps, a medium beep feels long. If you hear long beeps, that same medium beep feels short. The brain maps time based on how far it is from the "center" of the group.
3. The Rhythm (The Twist): This was the surprise. The data showed a wavy, repeating pattern. It's as if our brains have an internal clock that ticks in a rhythm, making certain time intervals feel slightly more "familiar" or "connected" to each other, even if they aren't the same length.

The Analogy: Imagine a spring (like a Slinky toy).

• If you stretch it out, the length represents the magnitude (short to long).
• If you look at the coils, the distance from the middle of the spring represents the context.
• The fact that it twists around and around represents the rhythmic component.

The Brain's "Two-Stage" Processing

The EEG data (the brain waves) revealed that this 3D shape isn't built all at once. It happens in two stages, like a camera focusing:

• Stage 1 (The Quick Snap - 150ms): Immediately after the sound stops, the brain does a quick, compressed calculation. It's like a rough sketch. It's very good at noticing if a sound is "weird" compared to the others (the oddball effect) and uses a logarithmic scale (where big differences feel smaller than they are).
• Stage 2 (The Full Picture - 300ms): A split second later, the brain refines the image. The "spring" shape fully emerges. This is when the brain's internal map matches what the people said they felt. The brain has now integrated the length, the context, and the rhythm into a complete 3D picture.

Why Does This Matter?

The study also found a link between your brain's "idle" state and how you see time. People who had stronger natural brain rhythms (specifically in the alpha and beta waves) when they were just resting had a "tighter" spring in their time perception.

Think of it this way:
If your brain's internal metronome is very steady and strong, your mental timeline is a tight, coiled spring. If your internal metronome is weaker, your timeline is a stretched-out, cone-shaped spring. This suggests that our personal "time sense" is partly wired into our brain's natural electrical hum.

The Big Takeaway

Time isn't just a straight line in our heads. It's a dynamic, multi-dimensional landscape.

• It's not just about "how long" something is.
• It's about "where it sits" in the group of sounds we just heard.
• And it's influenced by the rhythmic hum of our own brains.

This changes how we understand the mind: we don't just count time; we experience it as a rich, twisting geometry.

Technical Summary

1. Problem Statement

The fundamental question addressed by this research is how the human brain represents time intervals (durations). While traditional psychological frameworks often posit a unidimensional, linear "mental timeline" where durations are ordered monotonically from short to long, the authors argue this view is insufficient. They question whether the internal representation of duration is truly one-dimensional or if it possesses a more complex, multidimensional geometry. Specifically, they seek to determine:

• How many latent dimensions organize the psychological space of durations?
• What is the structure of this space (e.g., linear, logarithmic, helical)?
• How does this representational geometry emerge dynamically in neural activity?
• Is this geometry constrained by intrinsic neural dynamics?

2. Methodology

The study employed a multi-modal approach combining behavioral psychophysics and electroencephalography (EEG) within a Representational Similarity Analysis (RSA) framework.

Participants:

• 33 healthy adults (mean age ~27) completed two sessions.

Experimental Tasks:

1. Behavioral Similarity Judgment Task:
   • Participants rated the perceived similarity of all 45 possible pairs of 10 auditory durations (ranging from 400 ms to 2200 ms in 200 ms steps).
   • Stimuli were pure tones (440 Hz) with randomized loudness to prevent intensity-based cues.
   • Data was converted into Representational Dissimilarity Matrices (RDMs) for each participant.
2. EEG Oddball Detection Task:
   • Participants listened to a continuous stream of the same durations, interspersed with rare "oddball" durations (70 ms and 3400 ms).
   • They detected oddballs via button press.
   • EEG was recorded (64 channels) and analyzed time-locked to the offset of the standard durations.
   • Only trials without responses (standard durations) were used for analysis to isolate encoding processes.

Data Analysis Pipeline:

• Multidimensional Scaling (MDS): Applied to behavioral RDMs to visualize the latent geometry of the duration space.
• Theoretical Model Comparison: The authors constructed theoretical RDMs based on different hypotheses:
  • Linear: Absolute difference ($|d_i - d_j|$).
  • Logarithmic: Logarithmic difference ($|\log d_i - \log d_j|$).
  • Binary: Categorical (short vs. long).
  • Power: Stevens' power law ($|d_i^\alpha - d_j^\alpha|$).
  • Helix: A 3D generalized helical model incorporating monotonic, contextual, and periodic dimensions.
  • Random: Null model.
• Cross-Validation: Models were evaluated using leave-one-duration-out cross-validation to prevent overfitting.
• Neural RSA: Time-resolved EEG RDMs were computed using cross-validated Euclidean distances. These were correlated with theoretical RDMs and behavioral RDMs over time (0–600 ms post-offset).
• Commonality Analysis: Used to partition variance in EEG RDMs into components shared with behavioral similarity and unique components explained by specific theoretical models.
• Neural Correlates: Model parameters were correlated with resting-state EEG features (alpha/beta power) and task-related ERP components (P2 amplitude/latency).

3. Key Contributions

• Rejection of Unidimensionality: The study provides robust evidence that duration representation cannot be reduced to a single linear axis.
• Discovery of a 3D Helical Geometry: The authors propose and validate a generalized helical model that best explains the data, integrating three distinct dimensions.
• Dynamic Neural Unfolding: They demonstrate that the neural representation of duration is not static but unfolds in two distinct stages: an early logarithmic/contextual phase and a later behavioral-convergent phase.
• Link to Intrinsic Dynamics: They establish a correlation between individual differences in the geometry of duration space and endogenous resting-state oscillatory power (alpha/beta bands).

4. Key Results

A. Behavioral Geometry (Psychological Space)

• 3D Structure: MDS analysis revealed that the duration space is best described by three dimensions:
  1. Magnitude: A monotonic ordering of durations (short to long).
  2. Contextual Encoding: A U-shaped dimension representing the distance of a duration from the geometric mean of the set (central tendency).
  3. Periodic Component: A sinusoidal modulation suggesting a recurrent or cyclic structure.
• Model Fit: The Helix model (and the Power model) provided the best fit to behavioral data, significantly outperforming simple linear or logarithmic models, even after cross-validation. The helical model successfully captured the monotonic progression, the contextual compression, and the periodic modulation.

B. Neural Geometry (EEG Dynamics)

• Temporal Evolution: The neural geometry evolves in two stages post-stimulus offset:
  • Early Stage (~150 ms): Dominated by logarithmic encoding and sensitivity to "regularity" (distance from oddball anchors). This stage reflects compressed, task-relevant coding.
  • Late Stage (~300 ms): The neural RDMs converge with the behavioral similarity structure. The geometry becomes more complex, aligning with the linear, power, and helical models.
• Variance Partitioning: Commonality analysis showed that while behavioral similarity explains a large portion of neural variance at 300 ms, the Logarithmic and Regularness models explain unique variance at 150 ms, indicating the brain processes duration through multiple representational formats before settling into a stable behavioral-like state.

C. Link to Neural Dynamics

• Resting-State Correlation: Individual parameters of the fitted Helix model (specifically the $\beta$ parameter governing the exponential growth of the helix radius) correlated positively with resting-state alpha and beta power.
• Interpretation: Participants with higher alpha/beta power exhibited "spring-like" helical structures (slower expansion), while those with lower power showed "conical" structures (stronger compression of long durations). This suggests intrinsic neural oscillations constrain the geometry of time perception.

5. Significance and Implications

• Theoretical Shift: The findings challenge the "mental timeline" metaphor, proposing instead that time is represented on a manifold (a curved, multidimensional surface) rather than a line. This aligns with modern "manifold-based" accounts of neural representation seen in spatial cognition (e.g., toroidal manifolds).
• Mechanism of Timing: The results suggest that duration coding is not merely a scalar accumulation of time but involves:
  • Contextual integration: Relating intervals to the distribution of experienced stimuli.
  • Oscillatory constraints: The periodic component implies that endogenous neural rhythms may provide a temporal scaffold for interval timing.
• Methodological Advancement: By combining behavioral similarity judgments with time-resolved neural RSA, the study bridges the gap between subjective experience and neural population dynamics, showing that behavioral reports reflect a late-stage integration of earlier, more complex neural computations.
• Future Directions: The study opens new avenues for investigating how sensory modality, temporal scale, and specific neural oscillations shape the "geometry of time," moving beyond simple psychophysical scaling laws to a topological understanding of time perception.