Behavioral experiments: Compressed timeline of past and future
Behavioral experiments test whether people search the recent past and the near future along an ordered, compressed timeline.
Our brains maintain a memory of what happened when over the recent past (multiple seconds). Those memories are critical for learning temporal patterns and form the foundation of episodic memory. That temporal structure has to be represented in neural activity if the brain is going to organize recent experience into an ordered timeline, as the behavioral data suggests.
In this line of work, we analyze single-unit recordings to ask how neural populations keep track of elapsed time over behaviorally relevant scales, on the order of several seconds. Rather than relying on a single clock-like variable, the data point to a richer code: some neurons fire in ordered sequences after an event, while others change more gradually with a broad spectrum of time constants. Together, those signals provide a plausible neural basis for a temporally organized memory representation.
A central theme across these studies is compression. Temporal coding is densest near the present and becomes broader farther into the past, which matches behavioral evidence that memory for time is not linear. That compressed structure gives the system fine resolution for recent events while still preserving information over much longer intervals.
Stimulus-specific temporal sequences
One of the strongest results is that sequential temporal coding can also preserve content. In monkey recordings, different cue images evoke population responses that unfold over time in a similar sequential format, but with clear cue-dependent differences. That means the activity is not just signaling elapsed time, it is jointly representing what was seen and how long ago it occurred.
This kind of conjunctive code is a natural building block for episodic memory. A downstream system could, in principle, recover both the identity of a recent event and its temporal position from the same evolving neural state, rather than storing content and time in separate disconnected signals.
From neural dynamics to mental timelines
These results connect directly to our broader work on compressed mental maps. Sequentially activated cells provide an explicit record of recent experience, while slowly varying neurons with different time constants offer a candidate mechanism for generating those sequences.