The Vicsek model of flocking is deliberately minimal. Particles move at constant speed, align with their neighbors' average direction, and experience noise. The result: at low noise, global order — all particles moving in the same direction. At high noise, disorder. Between them, a sharp phase transition. The model captures the essential physics of collective motion in two ingredients and one parameter.
Gaur, Saha, and Paul (2602.21338) add one more ingredient: a vision cone. Each particle can only see neighbors within a forward-facing angular sector. This is realistic — birds see mostly forward, fish have lateral blind spots, even bacteria respond preferentially to stimuli ahead of them. The mathematical consequence is non-reciprocal interaction: particle A influences particle B, but B may not influence A, because B is behind A and outside its vision cone.
The result is a transition from global flocking to local clustering. As the vision angle narrows, global coherent motion breaks down. But it doesn't dissolve into random noise. Instead, small clusters form — groups of particles that maintain local alignment without global coordination. The clusters are internally ordered and externally independent. They persist, merge, split, and reform.
The physics of this transition is different from the standard order-disorder transition in the Vicsek model. The standard transition is driven by noise overwhelming alignment. This new transition is driven by information loss — narrow vision means each particle has access to fewer neighbors, and the alignment signal degrades not because of randomness but because of incompleteness. The particle is doing exactly what it should with the information it has. It just doesn't have enough.
The velocity-component distributions change character at the transition. In the globally ordered phase, one component of velocity (along the mean direction) has a narrow distribution while the perpendicular component is broad. In the clustered phase, both components are broad but show local correlations — within a cluster, particles agree on direction, but the directions of different clusters are uncorrelated. The correlation function reveals the cluster structure: strong short-range correlations decaying at a length scale that corresponds to the cluster size.
The relationship between density clustering and velocity coherence is causal, not coincidental. Velocity coherence creates density clustering, not the other way around. Particles that happen to face the same direction tend to collect in the same region because they're all moving the same way. This density enhancement then reinforces the velocity coherence because the particles in the cluster see mostly each other. It's a positive feedback loop, but one that saturates at the cluster scale — there's no mechanism to align clusters with each other.
High noise combined with narrow vision eliminates everything — no global order, no local clusters, just noise. This makes sense: noise destroys the alignment signal, and narrow vision means there isn't much signal to begin with. The two effects multiply rather than add. A system can tolerate high noise with wide vision, or narrow vision with low noise, but not both.
The ecological interpretation is clear. Real animal groups — flocks of starlings, schools of sardines — show exactly this clumpy, locally-ordered-but-globally-disordered structure. The standard Vicsek model predicts smooth global order that isn't observed. The vision cone modification produces the observed structure from a mechanistically realistic constraint. Animals don't have 360-degree awareness. The clumps are a consequence of this limitation, not of complex behavioral rules.