Spatially Periodic Cells Are Neither Formed From Grids Nor Poor Isolation

Julija Krupic, Neil Burgess, John O’KeefearXiv 2015 · DOI · arXiv

Abstract

Grid cells recorded in the parahippocampal formation of freely moving rodents provide a strikingly periodic representation of self-location whose underlying mechanism has been the subject of intense interest. Our previous work(1) showed that grid cells represent the most stable subset of a larger continuum of spatially periodic cells (SPCs) which deviate from the hexagonal symmetry observed in grid cells. Recently Navratilova et al(2) suggested that our findings reflected poor isolation of the spikes from multiple grid cells, rather than the existence of actual non-grid SPCs. Here we refute this suggestion by showing that: (i) most SPCs cannot be formed from hexagonal grids; (ii) all standard cluster isolation measures are similar between recorded grid cells and non-grid SPCs, and are comparable to those reported in other laboratories; (iii) the spikes from different fields of band-like SPCs do not differ. Thus the theoretical implications of the presence of cells with spatially periodic firing patterns that diverge from perfect hexagonality need to be taken seriously, rather than explained away on the basis of hopeful but unjustified assumptions.

Introduction

The discovery of grid cells—neurons that fire in a precise hexagonal lattice—has profoundly influenced our understanding of the brain's spatial navigation system. Most theoretical frameworks, such as continuous attractor models, rely on this geometric perfection to explain how we track location. However, empirical data often reveals a more complex reality: a continuum of Spatially Periodic Cells (SPCs) that includes but is not limited to perfect hexagonal grids. These "irregular" SPCs can appear elliptical, irregular, or even band-like.

Critics have argued that these non-hexagonal patterns are merely artifacts of "poor isolation"—the accidental grouping of spikes from multiple distinct grid cells into a single recorded unit. This paper systematically refutes this "contamination hypothesis," demonstrating that irregular SPCs are genuine biological entities, not technical errors.

Geometric Impossibility of Grid Summation

A core claim of the contamination hypothesis is that irregular SPCs are formed by the overlapping signals of multiple hexagonal grid cells. To test this, we employed geometric modeling to see if summing offset hexagonal grids could recreate observed SPC patterns.

Geometric modeling demonstrates that irregular SPC patterns cannot be simply explained as the sum of multiple hexagonal grid cells. Attempts to fit hexagonal templates (circles and triangles) to recorded SPC rate maps result in significant discrepancies, often predicting firing fields where the actual cell remains silent.

As shown in Figure 3, the geometry does not support the contamination theory. Summing grids frequently predicts "ghost fields"—areas where the model expects firing that is absent in the real data. This suggests that the underlying spatial representation in these cells is inherently non-hexagonal, rather than a composite of hexagonal inputs.

Waveform Consistency and Cluster Isolation

If SPCs were indeed contaminated by multiple units, we would expect to see inconsistencies in spike waveforms across different spatial locations or poor statistical separation in the recording's feature space. We analyzed the action potential "fingerprints" of these cells across their various firing fields.

Spike waveform consistency across different spatial fields and robust cluster isolation confirm that SPCs are single-unit recordings. (A, B) Spatial firing maps with spikes colored by field. (C, D) Overlaid waveforms from different fields show near-perfect alignment. (E, F) Cluster plots show clear separation of the recorded units in feature space.

Figure 4 illustrates that for both grid cells and irregular SPCs, the spike waveforms remain remarkably consistent regardless of where the animal is in the environment. Furthermore, quantitative isolation metrics—such as the L-ratio and isolation distance—show no significant difference between "perfect" grid cells and "messy" SPCs. If irregular cells were the result of poor recording quality, they would consistently score lower on these metrics; instead, they are indistinguishable from grid cells in terms of recording quality.

Statistical Validation

Our analysis of 91 grid cells and 153 SPCs revealed several key findings that further debunk the contamination hypothesis:

  • Quality Parity: There was no significant correlation between a cell's "gridness" score and its isolation quality ($P=0.85$).
  • Refractory Periods: Refractory period violations (spikes occurring too close together to be from one neuron) were rare and occurred at similar rates in both groups (approx. 7% for grids, 3% for SPCs). Notably, none of the controversial "band-like" cells showed violations.
  • Sorting Robustness: Re-analysis using automated clustering algorithms (Cluster-Kwik) yielded a 93% overlap with manual sorting, proving that these cell classifications are not a result of subjective human bias.

Conclusion

The evidence strongly suggests that the "messy" GPS is a biological reality. Spatially periodic cells that deviate from hexagonal symmetry are not recording artifacts but represent a broader class of spatial signals in the parahippocampal formation. These findings challenge the exclusivity of hexagonal grid models and suggest that the brain's coordinate system may be more flexible and environmentally dependent than previously thought. Future theoretical models must move beyond the "perfect grid" to incorporate the full spectrum of spatially periodic firing.

Summary updated

Abstract

Grid cells recorded in the parahippocampal formation of freely moving rodents provide a strikingly periodic representation of self-location whose underlying mechanism has been the subject of intense interest. Our previous work(1) showed that grid cells represent the most stable subset of a larger continuum of spatially periodic cells (SPCs) which deviate from the hexagonal symmetry observed in grid cells. Recently Navratilova et al(2) suggested that our findings reflected poor isolation of the spikes from multiple grid cells, rather than the existence of actual non-grid SPCs. Here we refute this suggestion by showing that: (i) most SPCs cannot be formed from hexagonal grids; (ii) all standard cluster isolation measures are similar between recorded grid cells and non-grid SPCs, and are comparable to those reported in other laboratories; (iii) the spikes from different fields of band-like SPCs do not differ. Thus the theoretical implications of the presence of cells with spatially periodic firing patterns that diverge from perfect hexagonality need to be taken seriously, rather than explained away on the basis of hopeful but unjustified assumptions.
BibTeX
@misc{krupic2015spatially,
  author = {Julija Krupic and Neil Burgess and John O’Keefe},
  title = {{Spatially Periodic Cells Are Neither Formed From Grids Nor Poor Isolation}},
  year = {2015},
  eprint = {1512.06248},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1512.06248},
}
Figure