Gaze Propagates Predator Location in Flocks, While Kinematic Cues Dominate Schools

Question: How do fish schools and bird flocks use gaze direction or body orientation cues, rather than overt escape movements, to propagate predator detection and collective vigilance?

Answer

Taken together, the retrieved papers support a taxon-asymmetric answer. In birds, gaze direction is now experimentally established as a pre-escape channel in its own right: pigeons transmit predator location through gaze-following, and receivers weight the duration of conspecific gaze toward the predator rather than locomotion or head-up vigilance [1]. That channel is graded by demonstrator number and socially verified, which is precisely what makes it useful for collective vigilance before anyone flees [5]. In fish, the same claim is not supported by the direct evidence. Where detector identity was experimentally controlled, viewing angle and relative body orientation to the informed fish did not predict the order in which naive schoolmates responded, while speed changes did [4]; the best-characterised neural mechanism for socially transmitted threat in a schooling fish is detection of partners' rapid escape and sudden disappearance ("social-off"), not their orientation [7]. Across these papers, gaze direction functions as an anticipatory, location-bearing cue in flocks, whereas in schools the anticipatory information appears to be carried largely by the absence of ordinary motion — neighbours staying still — plus abrupt kinematic events, with body orientation acting at most as a weak or redundant input [16].

Gaze toward a predator is a distinct channel from vigilance posture in birds

In foraging pigeon flocks, high-resolution head tracking with reconstructed visual fields showed that individuals aligned foveal orientation with conspecific gaze toward a predator stimulus before initiating evasion [1]. The critical dissociation is that receivers responded to the duration of conspecific gaze directed at the predator, not to locomotion or to head-up vigilance [1]. Gaze therefore carries something head-up posture does not: a direction, and hence a candidate location, which allows verification before a costly escape [1]. This measurement is only interpretable because pigeons hold eye position fixed in the head during flight, so head orientation is a faithful proxy for gaze [11].

Gaze-following scales with demonstrator number without saturating

Observer pigeons followed demonstrators' gaze specifically toward the target object rather than a perceptually similar distractor, and both the likelihood and frequency of following increased with the number of demonstrators looking at the target [5]. No nonlinearity — no saturation or quorum threshold — appeared in that increase under the tested conditions [5]. This makes convergent gaze a continuously graded reliability signal, which is the property a collective-vigilance system needs if it is to trade false alarms against detection speed [5].

In fish, orientation relative to the detector does not predict who responds

When single aversively conditioned fish triggered escapes in schools of five Puntigrus tetrazona, naive fish sequentially increased speed and followed, yielding a hierarchical leader–follower structure [4]. Neither initial distance, viewing angle, nor relative orientation to the informed fish predicted the order of the escape sequence [4]. A model with local alignment and distance regulation reproduced the observations, and suggested the informed individual remains partly attentive to neighbours during escape [4]. Effective collective evasion can arise from attraction and repulsion alone, without alignment interactions at all [9].

The fish channel is kinematic and "social-off", not orientational

In Danionella cerebrum, midbrain optic tectum and dorsal thalamus populations were activated by the rapid escape of social partners, and by the sudden disappearance of virtual partners, but not by disappearing stimuli lacking social relevance [7]. Virtual schools that escaped or disappeared were sufficient to make observers escape in the absence of any direct threat [7]. In reef fish presented with looming stimuli, escape thresholds tracked predator attack speeds rather than cruising speeds, and social context did not modulate responses in that study [6]. Schooling fish also weight a single most influential neighbour rather than averaging the group [13], and interaction intensity rather than network structure governs collective state transitions under stress [12].

Non-escape — stillness — is itself the informative cue that sustains vigilance

Modelling shoals as noisy evidence accumulators shows that a neighbour's flight signals danger while its stillness signals safety, and that social discounting of threat while neighbours remain still allows groups to be both faster and more accurate than solitary individuals [16]. Wild sulphur molly shoals under bird attack were best described by discounting rates above what individually Bayesian updating supplies, and the model predicts a false-alarm rate that stays constant as shoals grow [16]. Collective vigilance strategy is not fixed: many-eyes and sentinel solutions are alternative optima of the same cost-balancing problem, selected by how vigilance costs scale [15]. In foraging cardinals, vigilance showed a nonlinear relationship with inferred risk, rising at intermediate threat and declining at the highest threat in favour of departure [14].

Posture and eye-level measures are revising what counts as a vigilance cue

Blink rate decreased with group size in red and fallow deer, which the authors interpret as blinking being tied to social vigilance rather than anti-predator vigilance, and they question head position as a vigilance proxy [2]. In Arabian babblers, visual wing-lift postures altered the acoustics of mobbing calls and receivers signalled faster to a visual or joint visual-acoustic display than to an acoustic-only one, showing that body posture is a functional, receiver-directed component of collective antipredator signalling [8]. Guppies showed social conformity in eye-use asymmetry, with a left-eye bias in the presence of a live predator, indicating that lateralised visual monitoring has a group-level component [3].

Conflicts and gaps

The central conflict is between the pigeon and fish results: gaze duration is causal in flocks [1], while orientation and viewing angle fail as predictors in schools [4]. This may be a task artefact — the fish study used a diffuse light cue that has no location to look toward, so a location-bearing gaze cue could not exist — or a genuine sensory difference between mobile-eyed birds and laterally eyed fish; no retrieved paper separates these. A second conflict concerns mechanism: gaze-following implies verification prior to escape [1], whereas the glassfish result assigns the signal to post-hoc detection of others' escape and disappearance [7]. Scaling also conflicts: gaze-following rises monotonically with demonstrator number [5], while theory predicts discrete switching between many-eyes and sentinel regimes [15], and no retrieved study measures either in a fish school.

What would change the answer

A decisive experiment would combine conditioned detector identity [4] with head- and eye-tracking plus visual-field reconstruction in a schooling fish [11] during a spatially localised predator presentation. Orthogonally manipulating a virtual neighbour's orientation toward the threat and its speed in closed-loop VR — a facility already validated for schooling fish [10] and used to isolate neighbour number and influence [13] — would test whether oriented-but-stationary partners transmit threat at all. Recording tectal and thalamic populations in Danionella while partners orient without escaping would test whether "social-off" neurons also encode direction [7]. Finally, fitting a gaze-duration diffusion model to fish, as done for pigeons [1], would put the two taxa on a single quantitative footing.

References

[1] Delacoux, M., Chimento, M., Kano, F.. "Pigeons rapidly transmit information about predator location through gaze-following". bioRxiv. 2026-10-19T00:00:00.000Z. URL: https://doi.org/10.1101/2025.10.29.685277

[2] Howard, T. R., Rands, S. A.. "Video-recorded measures of blinking differ with group size in red deer Cervus elaphus and fallow deer Dama dama". bioRxiv. 2025-03-20T00:00:00.000Z. URL: https://doi.org/10.1101/2024.01.02.573958

[3] Penry-Williams, I. L., Brown, C., Ioannou, C. C.. "Visual lateralisation in female guppies demonstrates social conformity but is reduced when observing a live predator". bioRxiv. 2025-05-07T00:00:00.000Z. URL: https://doi.org/10.1101/2025.05.06.652411

[4] Jadhav, V., Aimon, C., Lamshana, F., Trendafilov, D., Escobedo, R., Sire, C., Theraulaz, G., Guttal, V.. "Tracing information propagation in fish schools with a conditioned escape response". bioRxiv. 2026-08-05T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.04.742699

[5] Delacoux, M., Itahara, A., Kano, F.. "Gaze Following in Pigeons Increases with the Number of Demonstrators". bioRxiv. 2024-12-09T00:00:00.000Z. URL: https://doi.org/10.1101/2024.12.05.627032

[6] Neven, S. L., Faber, L., Martin, B.. "Reef fish escape responses selectively match predator attack speeds". bioRxiv. 2026-03-25T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.21.713327

[7] Yu, J.-H., Milan, J., Meyerhof, G. T., Napoli, J. L., Lovett-Barron, M.. "Neuronal detection of social actions directs collective escape behavior". bioRxiv. 2025-09-27T00:00:00.000Z. URL: https://doi.org/10.1101/2025.09.24.678087

[8] Guggenberger, M., Keynan, O., Yovel, Y.. "A two-signal, two-receiver signalling system: Snake mobbing in the Arabian babbler". bioRxiv. 2026-09-06T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.25.747098

[9] Daniel Strömbom, Catherine Futterman. "An Alignment-Free Explanation for Collective Predator Evasion in Moving Animal Groups". arxiv. 2025-09-16T00:00:00.000Z. URL: https://doi.org/10.3390/dynamics3040043

[10] Escobedo, R., Reynaud, J., Bastien, R., Sanchez, S., Sire, C., Theraulaz, G.. "Closed-loop real-virtual interactions validate 3D model of social coordination in fish". bioRxiv. 2025-12-10T00:00:00.000Z. URL: https://doi.org/10.64898/2025.12.07.692829

[11] Ros, I. G., Bhagavatula, P. S., Biewener, A. A.. "A view of bird's eyes -- Pigeons lock their eyes in place during flight". bioRxiv. 2026-01-08T00:00:00.000Z. URL: https://doi.org/10.64898/2026.01.07.698303

[12] Lin, G., Escobedo, R., Li, X., Xue, T., Han, Z., Sire, C., Guttal, V., Theraulaz, G.. "Experimental evidence of stress-induced critical state in schooling fish". bioRxiv. 2025-02-25T00:00:00.000Z. URL: https://doi.org/10.1101/2025.02.21.639527

[13] Kang, Z., Escobedo, R., Combe, M., Sanchez, S., Sire, C., Theraulaz, G.. "Decoding social integration in schooling fish using closed-loop real-virtual interactions". bioRxiv. 2026-03-18T00:00:00.000Z. URL: https://doi.org/10.64898/2026.03.12.711426

[14] Martin, N. W., Sieving, K. E.. "FEED, FLEE, OR FACT-FIND: MODELING LATENT COGNITIVE-MECHANISTIC RESPONSES TO HETEROSPECIFIC ALARM CALLS FOR FORAGING CARDINALS". bioRxiv. 2026-10-16T00:00:00.000Z. URL: https://doi.org/10.64898/2026.10.12.758620

[15] Charlie Pilgrim, Andrew M Bate, Anna Sigalou, Mélisande Aellen, Joe Morford, Elizabeth Warren, Christopher Krupenye, Dora Biro, Richard P Mann. "Many-Eyes and Sentinels in Selfish and Cooperative Groups". arxiv. 2025-11-25T00:00:00.000Z. URL: https://arxiv.org/abs/2511.19093

[16] Kilpatrick, Z. P.. "Social Discounting Enables Fast and Reliable Collective Escape". bioRxiv. 2026-08-19T00:00:00.000Z. URL: https://doi.org/10.64898/2026.08.14.744984

How this survey was made

This survey answers one question: “How do fish schools and bird flocks use gaze direction or body orientation cues, rather than overt escape movements, to propagate predator detection and collective vigilance?”

We keep our own collection of recent preprints and refresh it every day from bioRxiv, medRxiv, arXiv, PsyArXiv, SocArXiv and ChemRxiv. For each survey, we search that collection for recent work and keep the strongest matches: papers that use the same terms as the question, and papers on the same topic in different words.

The text was written from the titles and abstracts of those papers; the full texts were not read. Statements that draw on a particular paper are marked with a number in brackets. The reference list just above is the 16 papers whose work this text draws on, chosen from the 106 papers those searches returned; the rest of the shortlist is not printed. It is not a complete review of everything published on the question.

Treat this as a starting point and a reading list, not a peer-reviewed review. The searches behind this survey ran on October 21, 2026 and covered work from the previous two years.