The revolutionary discovery of multiscale neural coding in bat brains
For decades, neuroscientists believed the mammalian brain navigated space using a simple neural "GPS": hippocampal place cells that fire when an animal occupies specific small locations. This model emerged primarily from studies of rodents scurrying in laboratory boxes rarely exceeding a few meters in size. But how does the brain handle navigation across kilometer-scale distances—the kind bats traverse nightly while foraging? A paradigm-shifting study led by researchers at the Weizmann Institute of Science reveals that bat brains employ a radical multiscale neural code unlike anything previously described. This discovery not only rewrites our understanding of spatial cognition but hints at how brains might efficiently represent complex real-world environments 1 5 .
The hippocampus contains specialized neurons called place cells. Each activates when an animal enters a discrete zone ("place field"), collectively forming a neural map. Traditionally observed in small enclosures:
Natural environments dwarf laboratory settings. A bat flying 20 km nightly would require ~66,000 place cells with rodent-like 30 cm fields—an implausible biological burden. Theoretical models suggested two solutions:
Bats challenged these models. When researchers recorded from flying Rousettus aegyptiacus (Egyptian fruit bats), they found:
In 2021, Tamir Eliav, Nachum Ulanovsky, and colleagues published a landmark study in Science. They wirelessly recorded dorsal CA1 neurons in wild-born bats flying through a 200-meter tunnel—orders of magnitude larger than standard rodent arenas. The results defied expectations 1 :
| Feature | Observation | Significance |
|---|---|---|
| Field size range | 0.6 – 32 meters | Largest directly measured fields in mammals |
| Max intra-neuron variance | 20-fold difference in field sizes | Violates classical "single-scale" neuron models |
| Fields per neuron | 1.7 (mean) | Supports combinatorial efficiency |
| Theta oscillation role | Absent during flight | Proves oscillations unnecessary for multiscale coding |
Theoretical decoding analyses revealed why this code excels in large spaces:
Achieved via small fields (~1 m)
Large fields (~30 m) cover expansive areas with fewer neurons
Combining scales reduces neuron count by >50% versus single-scale models 1 .
| Neural Code Type | Position Error (meters) | Neurons Required |
|---|---|---|
| Classical (single-scale) | 3.2 ± 0.5 | 100% (reference) |
| Multiscale (multifield) | 1.1 ± 0.3 | 42% |
| Rescaled fields | 5.8 ± 1.2 | 75% |
The team overcame formidable technical hurdles to study naturalistic flight:
Essential tools enabling this research and their functions:
| Research Tool | Function | Innovation |
|---|---|---|
| Wireless Neuropixels | Records hundreds of neurons in freely moving animals | Eliminates cables enabling flight studies |
| Motion Capture System | Tracks 3D position with millimeter precision | Correlates neural activity with exact location |
| Bayesian Decoding Models | Reconstructs spatial position from neural activity | Quantifies map accuracy and efficiency |
| Large-Scale Environments | 200-m tunnels, flight rooms (6×6×3 m) | Enables ethologically relevant navigation |
| Calcium Imaging (GCaMP6f) | Longitudinal neuron activity tracking in the same bats over weeks | Confirms coding stability across time 6 |
The multiscale code's discovery reshapes multiple fields:
"Our findings suggest the brain evolved a fundamentally different strategy for large-scale navigation—not just rescaling the map, but inventing a new kind of map altogether."
Bat navigation research exemplifies how studying brains in ethologically relevant contexts reveals fundamental principles invisible in constrained lab settings. The multiscale hippocampal code solves a core problem in spatial cognition: achieving high resolution and high capacity without biological implausibility. As wireless recording technologies advance, studying animals in naturalistic group settings—from bats to birds—promises even deeper insights into how brains build usable models of the world. This work reminds us that sometimes, to understand the brain, we must let it do what it evolved to do: navigate the real world in all its complexity 3 9 .
Eliav et al. (2021) Science 372: eabg4020
Ulanovsky et al. (2023) Nature 621: 796–803