Home / State of the art
State of the artWhat works today — and where the gaps remain.
An evidence-based, cited map of stunning and slaughter practice: the methods in use, the species-specific electrical parameters from the primary literature, the welfare-indicator hierarchy, the real limits of machine vision, and the open questions we intend to answer. Every figure on this page is referenced.
The largest animal-welfare problem almost no one measures.
Farmed finfish are now killed in numbers that dwarf all farmed birds and mammals combined — yet most are killed by methods the welfare science classifies as poor.
Peer-reviewed estimate for 2019 (range 78–171 bn); up roughly nine-fold since 1990 and now exceeding all farmed birds and mammals combined.
Mood et al. (2023), Animal Welfare [1]By recent estimates, the large majority are killed without effective stunning. The common default is live chilling in ice slurry — death by hypothermia and asphyxia while still sensible.
Welfare-sector estimateIn a survey of Italian operations, none stunned fish before slaughter — the gap that motivates a Mediterranean sea bass / gilthead bream pilot.
Clemente et al. (2023), Front. Vet. Sci. [14]The scale almost no one measures
Farmed finfish killed per year versus the share reliably stunned before killing.
Six methods, very different welfare.
Two methods can be humane when correctly applied; the rest are listed by the WOAH Aquatic Animal Health Code among methods shown to result in poor fish welfare2. The default in much of the industry is the bottom three.
| Method | How it works | Welfare | Time to insensibility | Notes |
|---|---|---|---|---|
| Percussive / mechanical | A blow or captive bolt to the brain causes immediate concussion. | Recommended | Immediate | Irreversible if correctly placed. Best for large fish (>1 kg). Field audits still found ~8% of trout with signs of consciousness after stunning.1112 |
| Electrical (in-water) | Current passed through the fish in an electric field induces an epileptiform insult. | Recommended | <1 s onset | Usually reversible — fish must be killed before they recover. Parameters are highly sensitive to water conductivity.67 |
| Ice slurry / live chilling | Fish are immersed in iced water; death by hypothermia and asphyxia. | Poor welfare | ~10–20 min | The Mediterranean default. Loss of consciousness is delayed — reported ~10–20 min depending on temperature and species. Fish are immobilised but can remain sensible.413 |
| CO₂ narcosis | Fish held in CO₂-saturated water until they stop moving. | Poor welfare | ~6 min | Salmon can stay conscious ~6 min despite loss of movement — "not considered a stunning method." Strongly aversive; banned for fish in Norway since 2010.6 |
| Asphyxiation in air | Fish removed from water and left to suffocate. | Poor welfare | 2–25 min | Trout modelling estimates 1.9–21.7 min of moderate-to-extreme pain per fish (3.5–74 min/kg).15 |
| Exsanguination w/o stunning | Gill arches cut while the fish is fully conscious; death by blood loss. | Poor welfare | Not quantified | Listed by WOAH among poor-welfare methods to avoid where stunning is feasible.2 |
"Recommended" methods are humane only when correctly calibrated and verified — which, as the next two sections show, is exactly where practice breaks down.
Time to insensibility, by method
Log scale (seconds). The two recommended methods act in ~1 second; the poor-welfare defaults take minutes.
There is no single setting — it is physics, per species, per site.
What stuns a fish in water is the field strength inside its head, not total current — and that depends on the ratio of the fish's conductivity to the surrounding water's7. Seawater (~50,000 µS/cm) is ~200× more conductive than fish flesh, so seawater species need far higher current density than freshwater ones. Published, measured parameters:
| Species | Water | Field / current density | Frequency | Duration | Insensibility / recovery |
|---|---|---|---|---|---|
| Atlantic salmon | Seawater | ≈1.25–2 Vrms/cm | 50 Hz AC | ~0.8–1 s | >2 min unconscious; EEG-recovery signs can appear from ~44 s.6 |
| Rainbow trout | Freshwater | 2.5 V/cm ≈ 0.12–0.16 A/dm² | 1000 Hz AC | ~1–2 s | Onset <1 s; recovers within minutes if not killed. 1 kHz minimises haemorrhage.4 |
| European sea bass | Seawater | 3.3 Arms/dm² | 50 Hz AC | 1 s | Epileptiform insult ~48 s; reported effective in lab trials.8 |
| Gilthead sea bream | Seawater | — no published spec | — | — | Industry extrapolates from sea bass. Recovery reported from <1 to >20 min by parameters.513 |
| Common carp | Freshwater | ≈0.14 A/dm² (0.09–0.41; 15–68 V/cm) | 50 Hz AC | 1 s | VER at 30±8 s; behavioural recovery 1–9 min → immediate kill required.9 |
| Nile tilapia | Freshwater | ≈1.1 A/dm² (often insufficient — resistant) | 50 Hz AC | ~30 s | Unusually resistant to electrical stunning; recent work recommends percussion instead.10 |
The biggest single gap: for gilthead sea bream — one of the largest unstunned finfish sectors — there is no published, measured electrical-stunning specification at all. Parameters are borrowed from sea bass. Closing species- and site-specific calibration gaps like this is the core of the problem.
Which species have a published stun specification?
The biggest single gap, at a glance: gilthead sea bream has no measured specification.
Stunning a fish is not the hard part. Knowing it worked is.
Insensibility can be confirmed rigorously only with brain measures — but those are lab-only. On the line, operators rely on physical proxies, and the science is clear that those proxies are weakly validated and differ by species.
Insensibility read from an isoelectric trace, or a generalised seizure followed by suppression.
The EEG response to a light flash; its absence indicates loss of consciousness.
Gill movement — the field sign that aligns most closely with VER, but not equivalent to it.
A brainstem reflex; useful but reliability varies markedly between species.
Can both over- and under-estimate insensibility. CO₂, for one, stops movement while the fish is still conscious.
Why this is the open problem. Electrical stunning is reversible — salmon can begin EEG recovery from ~44 s after the current stops6, so the stun-to-kill window is short. EFSA notes welfare indicators have "only been validated in the laboratory for a few species"5, and there are currently no behavioural indicators that can fully differentiate paralysis from unconsciousness or death14. That is precisely why verification and audit — not new stunning hardware — is where we focus.
The stun-to-kill window is short
Electrical stunning is reversible — recovery can begin fast. RSPCA Assured's target is ≤15 s stun-to-bleed.
Reliability vs field-usability of welfare indicators
The gold-standard measures are lab-only; the field-usable proxies are weaker. That inversion is the open problem.
What AI and machine vision can — and cannot — do yet.
Computer vision in aquaculture is advancing fast, but the capabilities that exist are not the capability this problem needs.
Demonstrated today
Not yet solved
Our reading of the evidence. Vision alone cannot prove consciousness, and pretending otherwise would be dishonest. The opportunity is a conservative, multi-signal, explainable system that flags welfare risk and produces audit evidence — fusing process, sensor and visual signals, with experts kept in the loop. See the boundary we hold ourselves to →
The rules point the right way — but stop short on fish.
The direction of travel is clear; what is missing is species-specific technical detail and routine verification.
WOAH Aquatic Code (Ch. 7.3)
Fish should be stunned before killing, with "immediate and irreversible loss of consciousness" — or be killed before they recover. Effectiveness is to be verified by absence of opercular activity, VER and the eye-roll reflex.2
Gap: sea bass and bream are notably absent from its worked method examples.
EU Regulation 1099/2009
Animals "shall be spared any avoidable pain, distress or suffering during their killing" — and this does apply to fish.3
Gap: only this general principle applies — there are still no species-specific technical requirements for fish.
RSPCA Assured
Permits only percussive-plus-bleed or electrical stunning; prohibits CO₂, air and ice suffocation. From 22 July 2025, CCTV must cover the whole slaughter process, including post-stun assessment.17
Signal: certification is moving toward continuous, recorded verification — the niche StunAssure serves.
We begin with questions, not assumptions.
These shape the Phase 1 problem map, the welfare-indicator framework, and the pilot design.
Welfare validation
- Which proxies are reliable for sea bass and bream specifically?
- How can ventilation and eye-roll signs be validated against VER in the field?
- What failure modes are most dangerous to miss?
Engineering
- Which sensors stay reliable in wet, saline, harsh environments?
- Can stunner and water parameters be logged without modifying existing equipment?
- Can the system run fully offline at a remote site?
Adoption
- What cost is acceptable for small and medium farms?
- What audit evidence would certifiers and buyers trust?
- What training do line operators actually need?
Pilot design
- Sea bass or bream first — and at what site type?
- Which Mediterranean operation is easiest and highest-impact?
- What 6-month result would justify expansion?
Grounded in public evidence.
Primary sources behind the figures on this page. Where the literature is silent — sea bream parameters, validated consciousness detection — we say so explicitly rather than fill the gap with a number.
- 01Mood, Lara, Boyland & Brooke (2023) — Numbers of farmed finfish killed, Animal Welfare 32:e12
Peer-reviewed estimate of ~124 billion farmed finfish slaughtered in 2019 (range 78–171 bn).
- 02WOAH — Aquatic Animal Health Code, Ch. 7.3: stunning & killing of farmed fish
Requires stun-before-kill, verification by absence of opercular/VER/eye-roll signs, and lists poor-welfare methods to avoid.
- 03EU Regulation (EC) No 1099/2009 — protection of animals at killing
General duty to spare avoidable suffering applies to fish; no species-specific technical rules for fish exist.
- 04EFSA AHAW (2009) — Welfare aspects of stunning & killing of farmed rainbow trout
Electrical parameters, frequency effects, and ice-slurry insensibility times for trout.
- 05EFSA AHAW (2009) — Stunning & killing of farmed sea bass & gilthead sea bream
Notes welfare indicators are validated only in the lab for a few species; treats the two species jointly.
- 06EFSA AHAW (2009) — Stunning & killing of farmed Atlantic salmon
Electrical-stun thresholds and the ~44–443 s EEG-recovery window; CO₂ not a stunning method.
- 07Lines & Kestin (2004) — Electrical stunning of fish: the relationship between water conductivity and body resistance, Aquaculture 241
Establishes that field strength at the fish, governed by conductivity, is what determines the stun.
- 08Lambooij et al. (2008) — Electrical stunning of European sea bass, Aquaculture Research 39(1)
Measured current density (~3.3 A/dm²) and epileptiform insult duration in seawater.
- 09Retter et al. (2018) — Electrical stunning of common carp (open access), BMC Veterinary Research 14:205
Field-confirmed carp parameters and recovery times across a range of conductivities.
- 10Sundell et al. (2024) — The quest for a humane protocol for stunning and killing Nile tilapia, Aquaculture 593:741317
Finds tilapia resistant to electrical stunning and recommends percussion / combined methods.
- 11Hjelmstedt et al. (2022) — Percussive vs electrical stunning of rainbow trout, Aquaculture 552
Captive-bolt stunning induced immediate, permanent brain failure — no epileptiform seizures or VERs — in the trout tested.
- 12Jung-Schroers et al. (2020) — Stunning on trout farms (open access), BMC Veterinary Research 16:197
Found ~8% of percussively stunned trout still showing signs of consciousness; worst farm 20%.
- 13de la Rosa et al. (2021) — Twenty years of research on sea bass & sea bream welfare at slaughter, Animals 11(8)
Reviews recovery times and the absence of bream-specific electrical specifications.
- 14Clemente et al. (2023) — Frontiers in Veterinary Science
"No behavioral indicators that can fully differentiate paralysis from unconsciousness or death"; Italian sea bass/bream stunning survey.
- 15Schuck-Paim et al. (2025) — Quantifying suffering in air-asphyxiated trout, Scientific Reports 15:19850
Estimates the duration and intensity of pain during asphyxiation slaughter.
- 16Cui et al. (2024) — Survey of machine vision in aquaculture (open access), arXiv:2406.17800
Documents strong controlled-setting performance alongside data, generalisation, and field-validation gaps.
- 17RSPCA — Welfare Standards for Farmed Atlantic Salmon (2024)
Permitted stunning methods and the CCTV-at-slaughter requirement effective 22 July 2025.