Farmed fish welfare is plausibly one of the effective animal advocacy community’s priorities because of the current neglectedness of the issue, the likelihood that farmed fish suffering is large in scale, and the potential tractability of interventions to improve farmed fish welfare.¹ With this in mind, ACE would like to reduce uncertainty about the impact of corporate campaigns aimed at improving farmed fish welfare.² We hope that the following analysis will help us move toward that goal.

This report covers considerations relevant to determining appropriate indicators of farmed fish welfare. It also discusses potential reforms to improve the welfare of farmed fishes. Following that, it outlines key resources and questions for further consideration. The appendix provides some further information about farmed fish production, our literature search, and our approach to analyzing the literature.³

Read the full report

---

1. For more information about ACE’s prioritization framework, see our report on cause prioritization.  

2. In the effective animal advocacy context, corporate campaigns usually involve some combination of professional relationships, petitions, media exposure, and other forms of public pressure to influence corporations to implement policies that aim to improve animal welfare. 

3. Much of the research cited on this page comes from studies using live fishes. ACE would like to acknowledge that this is problematic and that our engagement with this literature does not indicate that we support the use of animals in research. 

Table of Contents

• Background Information
  • Brief Overview of Fish Farming
  • Brief Overview of the Literature on Farmed Fish Welfare
• Indicators of Farmed Fish Welfare
  • Behavioral Indicators of Welfare
  • Cortisol Level Measurements
  • Mortality Rates
  • Environmental Indicators of Fish Welfare
    • Water quality and dissolved oxygen
    • Stocking density
  • Tentative Conclusions
• Reforms to Improve the Welfare of Farmed Fish
  • Changing Slaughter Methods
    • A brief overview of common slaughter practices
    • “Humane” stunning methods
      • Comparing promising stunning methods
      • Assessing insensibility
    • Direct financial costs of changing slaughter practices
    • Tentative conclusions for changing slaughter practices
  • Managing Dissolved Oxygen Levels
  • Environmental Enrichment
  • Reducing Triploidy
  • Reducing the Use of Wild-Caught Fishes in Farmed Fish Feed
  • Tentative Conclusions
• Key Questions for Further Consideration

Background Information

Our recommended charities currently direct a relatively small amount of resources to improving farmed fish welfare.¹ However, many interventions that are already in use for other farmed animals could be extended to farmed fishes,² and some charities have expressed interest in doing this.³ Numerous animal charities do advocate for fish sentience and highlight reasons to abstain from fish consumption.⁴ As far as we know, only four charities have completed undercover investigations of fish farms,⁵ though we’re aware of some organizations that direct significant resources to farmed fish welfare in other ways.⁶ Still, our understanding is that farmed fish welfare and corporate campaigns to improve it are relatively neglected.

Corporate campaigns on behalf of chickens seem to have a relatively strong track record,⁷ but it’s unclear whether corporate campaigns on behalf of farmed fishes will be as successful. Corporate campaigns are likely the most successful when they have strong public support, and farmed fish advocacy may have less public support than farmed land animal advocacy (e.g., people probably have lower levels of empathy for fishes than they do for land animals).⁸ In further analysis of the tractability of farmed fish corporate campaigns, Researchers may find it useful to factor in precedents set by earlier campaigns, and to consider how differences in the costs of implementing reforms (which may well differ significantly for chickens and fishes) could affect industry willingness to accept them. For now, ACE has a somewhat limited understanding of the tractability of corporate campaigns to improve farmed fish welfare. However, this could improve as a result of future (or further analysis of past) corporate campaigns.

Compared to other commonly farmed animals (such as birds and mammals), the scale⁹ of farmed fish suffering depends on three factors:

1. The capacity (or lack thereof) of fishes to suffer¹⁰
2. The degree to which farmed fishes are suffering in their living conditions¹¹
3. The number of fishes farmed and slaughtered each year¹²

It’s our view that these three factors combined indicate that farmed fish welfare is of considerable importance.

Brief Overview of Fish Farming

Humans farm several hundred species of fishes,¹³ some in much greater numbers than others. Mood & Brooke (2012) provide what appears to be the first detailed estimate of the global number of farmed fishes slaughtered annually.^{14, 15} They use FAO production tonnage data from 2010 (the FAO only measures this data by mass weight, not by individual fishes) combined with various other sources to estimate the number of farmed fishes of many different species.¹⁶ We now present updated estimates for these numbers by using FAO data from 2015 (the most recent year available at the time of drafting this report).¹⁷ The spreadsheets used for these calculations are available here.

Table 1: Ten Most Numerous Farmed Fish Species Slaughtered (2015)

Based on the 2010 FAO data, Mood & Brooke (2012) estimate that between 36.7–121.8 billion (midpoint approximately 80 billion) individual fishes were slaughtered that year.¹⁸ Using the same methodology, our results—based on the 2015 FAO data—indicate that 52–167 billion (midpoint 109.5 billion) farmed fishes were slaughtered that year.¹⁹ Please note that these estimates do not include aquatic invertebrates who some people may commonly classify as fishes (e.g., shellfish). The data also doesn’t account for fishes’ relatively high pre-slaughter mortality rates²⁰ or the fact that different species can have significantly different lifespans before their harvest. As a result, the reported slaughter totals may differ substantially from the total number of farmed fishes alive at any point in time for a given species.

There are significant differences as to which fishes are most commonly farmed in different countries. Aquaculture is currently largely dominated by carps, and the largest producer of carps is China,²¹ the apparent global leader in fish farming by volume.²² Europe mainly farms salmonids, such as rainbow trout (estimated midpoint of 1,889 million individuals slaughtered globally in 2015) and Atlantic salmon (estimated midpoint of 470 million individuals slaughtered globally in 2015).²³ In contrast, in the U.S. channel catfish (estimated midpoint of 908 million individuals slaughtered globally in 2015) reportedly make up over 80% of the nation’s farmed fishes by tonnage.²⁴ There are notable projected growth differences in the aquaculture industry between different countries—with the greatest growth expected in Asian countries.²⁵ Relatedly, there were significant differences in the change in production tonnage between 2010 and 2015 for the most produced species. Table 3 (in appendix) provides a brief overview of these differences. The appendix also elaborates on some limitations to our estimated numbers reported here.

In general, industrial fish farming is a young industry compared to other agriculture and livestock sectors.²⁶ The industry is dominated by freshwater fishes and inland aquaculture.²⁷ The farming practices across a variety of species can be generally categorized as follows: i) fertilization is induced, ii) sometime after birth young fishes are collected, iii) fishes are transferred to a nursery for a preset time, and iv) fishes are transferred from the nursery to a grow-out facility where they grow until reaching market size.²⁸ Our impression is that different systems (e.g., ponds, raceways, and cages)²⁹ seem to be used to varying extents for different species in different countries. We did not identify comprehensive information on the global prevalence of different systems during our literature search.

There are a number of farmed fish standards in the aquaculture industry. Some are voluntary private standards that include welfare guidelines alongside norms for sustainability and organic labeling. These include GlobalGAP, Global Aquaculture Alliance Best Aquaculture Practices (GAA BAP), and the Aquaculture Stewardship Council (ASC). Aquaculture that is certified according to these standards reportedly accounted for just over 6% of global production in 2015.³⁰ Roughly 3% of the world’s farmed fishes are GlobalGAP certified, and roughly 1% are certified by each of BAP, Friend of the Sea (FOS),³¹ and ASC respectively.³² The overall proportion of aquaculture certified according to such standards is quite low—in part because the majority of farmed fishes are raised in China, where certification is uncommon.³³ According to some informed advocates and academics, the best standards for farmed fish welfare are the RSPCA Assured standards.^{34, 35} Currently, the RSPCA Assured standards only cover Atlantic salmon and rainbow trout, both in limited farming circumstances.³⁶

Brief Overview of the Literature on Farmed Fish Welfare

There appears to be a relatively large and detailed body of literature relating to farmed fish welfare compared to some other areas in effective animal advocacy.³⁷ However, farmed fish welfare does not seem to have been investigated to the same extent that terrestrial farmed animal welfare has been. We do not aim to cover the entire body of literature in this report. Though there’s a growing interest in the field, we are far from having reached a scientific consensus. Assessing farmed fish welfare seems complex and, in some ways, more difficult than assessing the welfare of other farmed vertebrates. Other commonly farmed species are usually referred to at the taxonomic level of species (e.g., cows or pigs), while farmed fishes refers to an entire infraclass of species. The same infraclass taxonomic rank for birds (aves) includes a wide variety of diverse species such as penguins, hummingbirds, and ostriches. Thus, there is much greater diversity within farmed fishes with regard to characteristics relevant to welfare (e.g., ethology and anatomy) than in any other commonly farmed species such as chickens, turkeys, or rabbits. This greater diversity seems to stem from fishes’ long and distinct evolutionary history from other commonly farmed animal species’—as a result, fishes have evolved and adapted differently to various aquatic environments.

There appears to be much more relevant research on some fish species popularly farmed in Europe (such as Atlantic salmon and rainbow trout) relative to other populous species (such as carps and catfishes).³⁸ Due to the wide variety that exists within the infraclass of farmed fish species, there would be limited validity to generalizing this research beyond the specific species in question.³⁹ For example, some species display significantly different behavioral preferences (e.g., being broadly benthic or pelagic) and significantly different anatomies, resulting in differences in welfare requirements across species.⁴⁰ Even within a given fish species, welfare requirements may differ significantly as a result of the fishes’ individual characteristics and developmental stage.⁴¹

The available literature on farmed fish welfare, in general, has a number of significant limitations. For instance:

• Much of the literature uses metrics like health and growth, which may be poor indicators of fish welfare.⁴²
• There are formidable gaps in the literature.⁴³
• The literature on fish welfare sometimes lacks pre-analysis plans and may be biased towards non-null results.

Furthermore, very few welfare standards, pieces of scientific literature, or advocacy groups currently provide specific prescriptions for farming parameters to improve farmed fish welfare. Of those that do, it’s often unclear what their recommendations are based on—when it is clear, they appear to be based on quite limited evidence.⁴⁴

Indicators of Farmed Fish Welfare

The relevant literature, as a whole, uses dozens of different welfare indicators.⁴⁵ Also, much of the purported farmed fish welfare literature focuses on welfare indicators that, prima facie, seem more concerned with ensuring and maximizing the profitability of production rather than welfare.⁴⁶ We are aware of few attempts to systematically rank different farmed fish welfare indicators in terms of importance. These include the Salmon Welfare Index Model (SWIM 1.0) and various models by the European Food and Safety Authority (EFSA). We found these systematic rankings to be useful starting points, but avoided putting too much weight on them due to concerns about their methodology.⁴⁷

We now aim to discuss different welfare indicators in a way that can be meaningfully built upon in order to reduce uncertainty regarding their potential. Specifically, we will discuss behavioral indicators of welfare, cortisol level measurements, mortality rates, water quality, dissolved oxygen levels, and stocking density.

Behavioral Indicators of Welfare

To our knowledge, the most recent (and perhaps the only) comprehensive review of behavioral indicators of farmed fish welfare is Martins et al. (2012).⁴⁸ The review notes that behavior “represents a reaction to the environment as fish perceive it and is therefore a key element of fish welfare.” In addition, according to Huntingford et al. (2008), “behavioural welfare indicators have the advantage of being fast and easy to observe and therefore good candidates for use ‘on-farm.’”⁴⁹

A subset of behaviors that Martins et al. (2012) suggest as welfare indicators include:

• Changes in food-anticipatory behavior⁵⁰
• Feed intake⁵¹
• Individual swimming activity⁵²
• Group swimming activity⁵³
• Ventilation rate⁵⁴
• Acts of aggression⁵⁵

There are a number of limitations to using behavioral indicators of welfare, including but not limited to the following:

• There is variation in behavior within species.⁵⁶
• There is variation in behavior across species.
• Most behavior occurs below the surface of the water and therefore may not be visible to an observer at surface level.
• Behavioral welfare indicators can be difficult to quantify and can be variable over time.
• Chosen behaviors may not be representative of truly preferred behaviors if they’re chosen under unnatural, imposed circumstances (e.g., shallow cages).

Our limited impression is that some behaviors may be promising farmed fish welfare indicators, but at this stage we are quite uncertain about which behaviors are the most promising.

Cortisol Level Measurements

Scientists use measurements of cortisol,⁵⁷ a stress response hormone, to assess the presence of negative feelings in animals.⁵⁸ The conjecture is that negative feelings will generally manifest in a stress response in animals, as they do in humans. Some of the main strengths of this welfare indicator are that the stress response in fishes does indeed seem similar to that of other vertebrates,⁵⁹ and corticosteroids (the class of steroid hormones that includes cortisol) seem to be central to the mood control and emotions of a range of vertebrate species.

There have been a number of reviews of stress and cortisol in fishes. Ellis et al. (2012) presents a summary of some of the significant limitations to using cortisol measurements.⁶⁰ Some of the limitations include:

• A single cortisol measurement gives limited information about the duration of the stress; instead it only provides a momentary glimpse of HPA activity.⁶¹
• Stress isn’t necessarily detrimental to welfare.⁶²
• Single time-point measurements of cortisol have limited value because it’s important to account for diurnal and seasonal variations as well as environmental and genetic factors.
• There could be significant intraspecies variation in cortisol levels in response to a stressor.

Another notable disadvantage of using cortisol measurement as an indicator of welfare is that it usually requires taking a blood sample, which is an invasive and stressful process.⁶³ Given the abovementioned significant limitations to its use, proper assessment of cortisol levels probably requires either repeated measurements of cortisol in the same individual fish, or comparison to a control group. Cortisol measurements are currently one of the more difficult measures to make in the field and would be very resource-intensive for a third-party certification program. Our current view is that properly applied cortisol measurements seem to be a potentially useful indicator of welfare, but not a practical choice for use in welfare certification programs.

Mortality Rates

Our understanding is that the only review of mortality as an indicator of farmed fish welfare is Ellis et al. (2011).⁶⁴ Perhaps unsurprisingly, the review notes that mortality “has received insufficient attention as a fish welfare topic.” It seems relatively uncontroversial to us that mortality has a clear relationship with welfare—particularly insofar as many causes of death also cause suffering. However, low mortality rates don’t necessarily indicate high welfare. Ellis et al. (2011) outline some important considerations:

• Mortality is a relatively easily understandable welfare indicator.
• In some cases, regular removal of dead fishes can be used to benchmark mortality rates.⁶⁵
• Differences in population sizes typically quantified at different stages of the farming process can be used to estimate mortality rates.
• Mortality rates close to zero don’t necessarily indicate good welfare, as welfare may otherwise be poor but not result in mortality.

Mortality rates alone don’t account for the degree of suffering—for instance, death can either follow acute trauma or persistent chronic suffering. Mortality itself is the effect of poor welfare, rather than the cause of it. It is unclear to us what factors are the greatest contributors to pre-slaughter mortality in farmed fishes, and we doubt whether adequate information can be found in any one resource.⁶⁶ A number of welfare indicators have been linked to mortality rates, including water quality and stocking density. Some other factors that may contribute to high pre-slaughter mortality rates include disease and cannibalism.⁶⁷

A strikingly high mortality rate is present in farmed fishes compared to other commonly farmed vertebrates. By compiling information mainly from FAO reports, and in some cases combining this with our own analysis, we estimate that pre-slaughter mortality rates for some of the most commonly farmed fishes range from approximately 15%–80% over the entire production cycle.⁶⁸ Note that the mortality rates may be quite high in very young fishes and much lower as fishes approach slaughter.⁶⁹ If we account for the lifetime of fishes, the mortality rates are more comparable to those of other invertebrates than they initially seem.⁷⁰

Despite our limited understanding of which factors contribute most to the high mortality rates on fish farms, our limited impression is that pre-slaughter mortality rates seem in some ways to be a promising indicator of farmed fish welfare. To be clear, high mortality rates seem to indicate low welfare but low mortality rates don’t necessarily indicate high welfare. Also, it seems that mortality measurements can easily be made during practices that already occur on fish farms. In our view, these reasons alone make pre-slaughter mortality rates a promising welfare indicator.

Environmental Indicators of Fish Welfare

Water quality and dissolved oxygen

Water quality seems to be widely regarded as one of the most—if not the most—important factor for farmed fish welfare.⁷¹ For instance, a 2014 Farm Animal Welfare Council report considered water quality to be primary among the most important factors affecting fish welfare. In the few attempts that we are aware of to systematically rate the importance of different welfare indicators, water quality receives a high score. For example, according to a 2013 model of welfare assessment for Atlantic salmon, two of the four most important welfare indicators pertain to water quality.⁷² Similarly, the European Food and Safety Authority (EFSA) models cite water quality as being highly important.

Water quality refers to a variety of measurements rather than a single specific welfare indicator. These include dissolved oxygen (DO) levels, temperature, pH levels, ammonium levels, and more.⁷³ In the remainder of this section, we focus mainly on DO levels because (i) DO seems to be one of the most, if not the most, important indicator for water quality, (ii) the issues with accurate DO measurements may be generally indicative of the issues with other water quality indicators, and (iii) we deemed it beneficial to focus on other welfare indicators as opposed to covering additional water quality indicators in detail. We’d like to note that water temperature also seems to be a very important aspect of water quality, and that water temperatures largely influence DO requirements.⁷⁴

DO is quite simply the amount of oxygen dissolved in water.⁷⁵ It is widely regarded as one of the most important welfare indicators for farmed fishes.⁷⁶ For instance, one EFSA model ranks DO as the most important welfare indicator for carp and the single most important abiotic factor for salmon ongrowing.⁷⁷ Low DO levels seem to strongly correlate with pre-slaughter mortality rates⁷⁸ and in extreme cases they may even result in mass mortality.⁷⁹ Fishes appear to show preferences for locations with higher DO levels.⁸⁰ Depleted DO levels appear to result in reduced growth, relative to growth at non-depleted DO levels.⁸¹ Feed conversion ratios also appears to be lower at low DO levels.⁸²

Within the aquatic environment, DO concentrations will vary as a result of a number of factors such as the flow rate of water, temperature, and time of day.⁸³ Optimal DO levels depend on multiple factors including fish size, water temperature and salinity, and the altitude of the farming operation.⁸⁴ Optimal levels are also likely to vary by fish species, with both lower and upper limits that should be avoided.⁸⁵ For instance, pangasius and some carp require relatively low amounts of DO in their water.⁸⁶ Furthermore, optimal DO levels seem to vary with the life stage of fishes.⁸⁷

Our impression—given our limited engagement with the literature which itself has significant limitations—is that DO seems to be widely regarded as one of the most important indicators of farmed fish welfare. It appears to be fairly strongly associated with a number of important outcomes for farmed fishes, and it seems quite likely that DO levels may significantly correlate with farmed fish welfare. Measuring DO and temperature seem like essential steps to evaluating water quality. Care should be taken when prescribing and measuring DO levels to account for i) the specific features of the overall farming operation that impact the optimal DO levels (e.g., fish species, expected temperatures) and ii) the daily fluctuations in DO levels.

Stocking density

Defining and measuring stocking density may be slightly more complex for farmed fishes than for terrestrial species because fishes move in three dimensions to a greater extent. For fishes, stocking density refers to the fishes’ weight per unit volume (as opposed to area) or fishes’ weight per unit volume in unit time of water flow.⁸⁸ Unfortunately, economic considerations push producers to increase stocking density. High stocking densities have been associated with a variety of negative outcomes, including but not limited to:

• Increased aggression⁸⁹
• Greater rates of physical injury,⁹⁰ specifically fin damage⁹¹
• Decreased growth rates⁹²
• Increased pre-slaughter mortality rates⁹³
• Increased stress⁹⁴
• Increased susceptibility to disease⁹⁵
• Decreased water quality⁹⁶

It appears that the specific mechanisms for the abovementioned negative effects are not agreed upon.⁹⁷ There is disagreement and conflicting information about the most important way in which stocking density affects farmed fish welfare. Some claim that the greater amount of negative social interactions (e.g., tail nipping, or other possible acts of aggression) caused by higher stocking densities are the primary cause of compromised welfare. Meanwhile, others argue that the degradation of water quality is the primary adverse effect caused by increased stocking density.⁹⁸ While the negative effects of increased stocking density on farmed fish welfare are commonly attributed to both water quality degradation and increased frequency of negative interactions between fishes,⁹⁹ the relative impact of those two factors appears to have been subject to limited examination. There is some evidence suggesting that the impact of high stocking densities on welfare may significantly vary depending on whether water quality can be maintained.¹⁰⁰ There is further reason to think that all else equal, lowering the stocking density may have a limited impact for farmed fishes. For instance, one very large-scale study on chickens suggested that lowering stocking density in isolation did not have the expected benefits for their welfare.¹⁰¹ The precise relationship between stocking density and welfare is difficult to infer, with one often-cited study on Atlantic salmon suggesting that stocking density may have relatively little impact on welfare until an inflection point is reached.¹⁰²

Our limited impression is that stocking densities seem likely to be negatively associated with farmed fish welfare. With the current production methods, stocking density is most likely so high—given the strong economic incentives for producers to do so—that it often negatively influences farmed fish welfare to a significant degree. However, stocking density itself may have a limited direct impact on welfare and may have a greater impact on welfare when it affects water quality.

Tentative Conclusions

The literature on assessing farmed fish welfare is vast, and we don’t cover it in detail in this report. There exist a variety of welfare indicators, and relatively little work has been done to determine which are the most important. Our current belief is that given the various limitations to different welfare indicators, it seems that no single measurement should be used to assess farmed fish welfare. Instead, different promising indicators should be used to provide a more holistic estimate of farmed fish welfare. Pre-slaughter mortality rates are likely to be a promising welfare indicator due to their simplicity and ease of measurement, as well as the likelihood that they are representative of welfare. Water quality—of which DO levels and water temperature seem likely to be some of the best measures—is also a promising indicator. While high stocking density is associated with many negative impacts on farmed fish welfare, much of its impact on welfare may be indirect through its negative impact on water quality. When properly applied, cortisol measurements may be one of the better measurements of negative mental states in farmed fishes. That being said, single time-point measurements may have significant limitations, making this measure difficult to apply in production environments. There are also some other indicators that we’re unsure will be practically applicable given our current knowledge. For example, while behavioral indicators could quite plausibly be used to assess the subjective mental states of fishes, we’re currently unsure how to best utilize them. Certain welfare indicators do seem like they should be used, particularly DO levels and mortality rates in all settings, and properly measured cortisol levels in laboratory settings.

We are quite uncertain about how to best assess farmed fish welfare and do not have a complete understanding of many of the scientific factors involved, including the relevant ethology, environmental chemistry, and biology. We also don’t have a thorough understanding of all the commonly used welfare indicators. Other welfare indicators that we think may be worth further consideration include:

• Indicators regarding the presence of various pathogens¹⁰³
• Indicators regarding injuries and physical deformities
• Genetic indicators¹⁰⁴
• Indicators derived from preference tests¹⁰⁵

Overall, our view is that there is significant uncertainty around how to assess the welfare of farmed fishes, and the available literature seems to generally support our view. Further work evaluating different welfare indicators may lead us to significantly update our conclusions.

Reforms to Improve the Welfare of Farmed Fishes

This report aims to reduce uncertainty about the impact of different interventions on the net welfare of farmed fishes. In doing so, we sometimes considered how a specific change would impact the expected number of farmed fishes. We also considered some interventions’ expected impact on the sales prices of farmed fishes, since demand for fish seems to exhibit significant price elasticity.¹⁰⁶ The remainder of this section presents a brief analysis of some promising welfare interventions. Specifically, we will discuss slaughter practices, DO levels, environmental enrichment, farmed fish genetics, and the use of wild-caught fish in farmed fish feed.

Changing Slaughter Methods¹⁰⁷

Literature on changing slaughter methods offers the most detailed research and analysis on how to practically improve farmed fish welfare. As a result—and because some slaughter practices seem to cause significantly less suffering than others—we completed a lengthier analysis of “humane”¹⁰⁸ slaughter than we did for other interventions. In addition, because it is the first intervention we discuss, we provide further relevant background information to a greater extent.

A brief overview of common slaughter practices¹⁰⁹

The most common methods of slaughtering fishes are asphyxiation and live chilling¹¹⁰ (i.e., placement in an ice slurry). These common slaughter methods are considered distressing¹¹¹ and are not approved by existing farmed fish welfare guidelines.¹¹² In practice, death by asphyxia happens several ways, but it typically means that fishes are removed from the water and exposed to air, resulting in the gills collapsing due to reduced oxygen intake. Depending on temperature and humidity, it can take minutes to hours for fishes to die from asphyxia.¹¹³ The general idea of live chilling is to “simultaneously chill, sedate and kill the fish by suffocation.”¹¹⁴ It seems that the length of the stunning period and the live chilling process can take over a period of at least 12 minutes to several hours.¹¹⁵

A 2017 IBFC report provides what seems to be the most detailed information on slaughter for some of the most frequently farmed fish species in Europe (Atlantic salmon, rainbow trout, common carp, gilthead sea bream, and sea bass).¹¹⁶ Among other things, the report examines the extent to which OIE standards for stunning and killing methods are achieved globally for these species. The standards include verifying successful stunning, using a backup stunner if necessary, and ensuring death takes place before fishes regain consciousness. The OIE standards support the use of stunning¹¹⁷—electrical and/or percussive—for killing farmed fish. Other methods do not meet their standards.¹¹⁸ However, the OIE guidelines lack specific details (e.g., prescribed stunning parameters for individual fish species).^{119, 120}

Perhaps of particular interest to animal advocates is that in China, where most farmed fishes are produced, the slaughter of common carp (a species of carp) does not meet OIE guidelines. The fishes are largely sold alive or dead in markets, and alive in restaurants. They are not slaughtered in commercial processing plants and are not electrically stunned, which sometimes occurs in Europe.¹²¹ Instead, in restaurants, a physical strike to the head is often followed by evisceration or decapitation. After harvest at the aquaculture facility, carp are commonly killed via asphyxia.¹²² Our limited understanding is that the treatment of common carp in China is generally indicative of the treatment of other carp species in China.

“Humane” stunning methods

The two stunning methods which are widely referred to as “humane” are electrical stunning and percussive stunning. Sources state that by 2014, 17 farmed fish species had research-backed methods for humane stunning, all of which were electrical or percussive stunning.¹²³ Of these, 16 species featured some method of electrical stunning, eight referred to some form of percussive stunning, and seven featured both. Another set of respected standards for farmed fish welfare by informed animal advocates and academics are the RSPCA Assured standard.¹²⁴ For fishes, RSPCA currently only has standards for Atlantic salmon and rainbow trout, both in limited farming circumstances.¹²⁵ With regard to slaughter, mechanical percussive and electrical stunning are approved for both species by the RSPCA.¹²⁶

Percussive stunning involves delivering a physical blow to the head, above or adjacent to the brain, of sufficient strength to damage the brain and cause unconsciousness. For effective percussive stunning, sufficient strength must be delivered for the specific fish being struck, and the appropriate location must be struck on each particular fish. Electrical stunning involves applying electrical current of sufficient strength, duration, and frequency, such that loss of consciousness is induced.¹²⁷ There are two types of resultant stun: that which stuns and kills (electrocution) and that which stuns only (electronarcosis).¹²⁸ Given different sizes, anatomies, and circumstances, different species require different electrical stunning parameters.¹²⁹

It’s important to note that the literature on “humane” stunning/killing parameters seems to be based to a large extent on one or two studies that have rather small sample sizes. For example, Lambooij et al. (2008) determine that the electrical stunning parameters for Nile tilapia are at times partly based on a sample size of eight,¹³⁰ while Robb et al. (2010) identify group sizes of ten or fewer.¹³¹ Such small sample sizes leave significant room for concern about the extent to which results are representative.¹³²

Norway and Chile account for 80% of the world’s farmed Atlantic salmon production, producing 56% and 24% respectively.¹³³ In Norway, 95% of salmon are slaughtered following electrical stunning, percussion, or live chilling with CO₂.¹³⁴ In Chile, salmon are mainly slaughtered in slaughterhouses via percussion. IBFC notes that the main hazard of percussive stunning is mis-stunning caused by a variation in fish size,¹³⁵ but doesn’t attempt to quantify the proportion of mis-stunning. Roth et al. (2007) present what seems to be the main piece of evidence supporting specific percussive stunning parameters for Atlantic salmon, but they note that their sample size was insufficient to conclude what force would stun 100% of individuals.¹³⁶ Presently, the extent to which Atlantic salmon in commercial settings are successfully stunned after the first percussive stunning blow is not clear to us. However, our limited impression is that the manufacturers of stunning equipment do complete some further testing to set the stunning parameters for their equipment,¹³⁷ but we have not come across any reports outlining the manufacturers’ methods and results.

Still, the basic reasoning for preferring properly applied percussive or electrical stunning over other common slaughter methods is that via the other common methods, the final stressful stages of slaughter are relatively long in duration.¹³⁸ Properly applied, electrical stunning or percussive stunning should significantly reduce the duration of those final stressful experiences. Percussive stunning often kills fishes itself (and if not, it should be quickly followed by appropriate treatment), while electrical stunning seems more likely to require post-stun treatment.¹³⁹ In a “humane” slaughter process, any fish not killed during the immediate stunning process should be killed before consciousness is recovered.¹⁴⁰

Comparing promising stunning methods

There appear to be advantages and disadvantages to electrical and percussive stunning depending in part on the species of fish being stunned.

Table 2: Key Welfare Concerns and Comparative Advantages and Disadvantages of Percussive Stunning and Electrical Stunning of Farmed Fishes¹⁴¹

Assessing insensibility

A key part of “humane” slaughter is to reduce suffering during slaughter by rendering fishes insensible before death. Difficulties arise with assessing whether a supposedly stunned fish has been rendered insensible, as opposed to merely immobile. An immobile fish would not be able to demonstrate that they are conscious, but would still experience the suffering involved in slaughter.¹⁴⁸

It seems clear that a fish who is swimming, struggling, or maintaining equilibrium hasn’t been successfully stunned. However, the absence of those behaviors would not reliably indicate whether a fish was truly insensible rather than just immobile. In contrast, testing for the eye roll reflex seems like a more reliable method for assessing consciousness (or lack thereof).¹⁴⁹ In the absence of the eye roll reflex, it seems one can be reasonably certain that fishes are insensible. On the other hand, only methods involving electroencephalography (EEG) can determine with a very high degree of certainty whether a fish has been made unconscious via stunning,¹⁵⁰ but these methods only seem viable in experimental settings.¹⁵¹ While the absence of behavioral signs—such as swimming, maintaining equilibrium, and the eye roll reflex—perhaps shouldn’t be considered as highly conclusive evidence of unconsciousness, they are probably the best available indicators in the field.

Direct financial costs of changing slaughter practices

We are primarily interested in the effects of “humane” slaughter on farmed fish welfare. However, we are also interested in the effects on cost production, since this could determine the tractability of fish producers adopting new welfare policies. Changes in production costs could also influence the sales prices of farmed fish products, in turn shaping the subsequent demand for those products. In their 2017 report, the IBFC estimated the cost of replacing common slaughter practices with stunning for certain species in some European countries.¹⁵² Results show that the switch to stunning—particularly in larger facilities—does not lead to a substantial increase in production costs. In some cases, it may even reduce costs.¹⁵³ Given the price elasticity of demand for fish, if humane slaughter methods lead to increased cost per farmed fish, demand could decrease for that specific species as a result. However, much of the decrease in demand could come from consumption shifting to other species of fish.¹⁵⁴

Tentative conclusions for changing slaughter practices

We still have a number of questions for further consideration with regards to changing slaughter practices. Regarding assessing stunning success, it seems relatively uncontroversial that EEG is the only truly reliable measurement, and that assessing the eye roll reflex one of the best ways to assess stunning in commercial settings. Percussive stunning or electrical stunning currently seem to be the most promising options for farmed fish slaughter. When properly applied, they seem very likely to be preferable to the common methods of asphyxia and live chilling. However, we would like to see further research determine what proportion of fishes are mis-stunned in commercial settings for the various species and stunning methods.

It seems that the most respected standards by informed advocates and academics are the OIE guidelines and the RSPCA Assured standard. The cost of complying with OIE guidelines for the most numerous farmed fishes in European countries (and possibly other species) seems non-prohibitive to inclusion in welfare asks. If implemented, these guidelines would generally cause decreases in demand for the affected fish products as a result of increased production prices. However, this could be caused by demand shifting to other farmed fish products.

Managing Dissolved Oxygen Levels

As indicated above, Dissolved Oxygen (DO) levels appear to be one of the most important farmed fish welfare indicators. It is not clear what DO levels usually are in farmed fish operations; we were unable to identify this information for many of the main species in the literature search, and it seems that it probably doesn’t currently exist in any one resource.¹⁵⁵ The limited information available—combined with our general knowledge about usual farming practices (e.g., high stocking densities, limited care for water quality)—suggests that it’s likely that farmed fishes are generally subjected to undesirable DO levels.

Limited information exists concerning optimal DO levels for various species of farmed fishes, and specific prescriptions for DO levels are quite rare. What’s more, existing prescriptions seem based on limited evidence. Based on our literature search, most Researchers seem to agree that the minimum DO level for all farmed fish species should be between 3 and 9 mg/L.¹⁵⁶ Some recommendations include a specific minimum DO level. These precise minimums do not seem clearly supported by evidence, though they are within the range one would expect based on the overall body of evidence. For instance, the Fish Ethology Database (FishEthoBase) for nile tilapia recommends that DO levels should be greater than 4 mg/L.¹⁵⁷ This information appears to be based heavily on Gilbert (1996), a study that doesn’t seem to provide strong support for this recommendation.¹⁵⁸ Similarly, compliance with GAA BAP standards initially require 4 mg/L in the effluent water for all pond farms, though the recommendation seems to be based on reasons other than farmed fish welfare.¹⁵⁹ The FishEthoBase for Atlantic salmon recommend maintaining levels over 7 mg/L,¹⁶⁰ citing references that similarly don’t seem to offer strong support for that particular threshold.¹⁶¹ In this case, the FishEthoBase recommendations note that further research is needed. The RSPCA food assurance program for Atlantic salmon and rainbow trout also advocate for levels over 7 mg/L, though they don’t provide citations in support of their recommendations.¹⁶²

Increasing DO levels may be achievable through aeration, the process of mechanically adding air to water in order to increase oxygen levels. The costs involved with aeration do not appear to be prohibitive for inclusion in a welfare ask. However, aeration may make farmed fish products slightly more expensive which could in turn decrease demand.¹⁶³ Improved DO levels will also likely cause faster growth and an increased feed conversion ratio which would lead to greater harvest weight or less time spent on farms to meet conventional harvest weight. However, greater feed efficiency could decrease cost per fish by decreasing feed costs, possibly causing subsequent increases in demand for that type of farmed fish and decreases in demand for other farmed fish species.¹⁶⁴

Environmental Enrichment¹⁶⁵

The many types of environmental enrichment (EE) include “(i) physical enrichment, including modifications or additions to the tanks, that is, structural complexity; (ii) sensory enrichment, which concerns stimulation of the sensory organs and the brain; (iii) dietary enrichment, encompassing type and delivery of food (note the distinction from nutrient enrichment, which concerns addition of nutrients to the feed); (iv) social enrichment, adding contact and interactions with conspecifics; and (v) occupational enrichment, relating to reduction of physical and psychological monotony by introducing variation to the environment and possibilities for exercise and performance of preferred behaviours.”¹⁶⁶ The effects of various environmental enrichments on many different species of fish have been studied with a focus on improving fish welfare and decreasing production prices. Näslund & Johnsson (2014) offers the most recent comprehensive review that we are aware of for environmental enrichment.¹⁶⁷ The study reviews the effects of dozens of physical structures and substrates on fishes,¹⁶⁸ and reports evidence for a number of EE interventions. We haven’t vetted the quality of the studies included in the review so we place limited confidence on their results. Still, the review provided some interesting insights:

• With regards to Atlantic salmon, access to shelters has reportedly been shown to reduce basal plasma cortisol levels and improve metabolism in juveniles. Some studies suggest that incubation substrates improve survival rates of juvenile salmon, while one study suggests that providing a partial cover leads to lower stress levels and improved growth in Atlantic salmon.
• Provision of shelter structures has reportedly been shown to have beneficial effects in several species of catfish, who are nocturnal in their natural environment. However, floating covers appear to have overall negative effects on survival and growth of pond-reared channel catfish.
• One study suggested that structural EE may increase aggression in nile tilapia.¹⁶⁹

FishEthoBase offers information about specific types of habitats that evidence suggests may improve welfare for Atlantic salmon, nile tilapia, and gilt-head bream.¹⁷⁰ Their recommendations for different species include standards for covering tanks, suggestions for water velocity in the enclosures, and guidelines for the structures and substrates that should be used. We haven’t vetted the quality or relevance of the studies included in FishEthoBase’s literature reviews.

Some reviews note that more research needs to be done in order to understand environmental enrichment, as many of its effects are likely to be species-specific and dependent on the specific type of structure added.¹⁷¹ As such, it is essential to adapt EE to each specific species of fish. Although EE seems to offer many possible interventions—lighting, coloring of environment, floor substrate, etc.—our current understanding is that there is a dearth of information about which conditions to advocate for. At this stage it seems that further research is required to determine which environmental enrichments farmed fish advocates should promote.¹⁷²

Reducing Triploidy

The aquaculture industry sometimes intentionally creates triploid salmon, individuals with three complete sets of chromosomes instead of the typical two. It appears that this genetic change is used as a strategy to reduce potential problems of interbreeding between escaped farmed and wild salmon.^{173, 174} Triploidy has the potential to significantly affect welfare. For example, there is evidence that triploid Atlantic salmon have higher mortality rates and deformities than diploid salmon.¹⁷⁵ As such, some advocacy groups have expressed interest in asking fish farms to reduce their use of triploids.¹⁷⁶ Based on our initial research, this strategy could be promising as triploid salmon appear to have lower welfare and slower growth rates than diploid salmon.¹⁷⁷ Fraser et al. (2012) attempts to determine changes in welfare from switching from triploid to diploid salmon.¹⁷⁸ In general, their conclusions seem to slightly favor diploids over triploids. However, the effects of greater interbreeding between escaped farmed diploids and wild Atlantic salmon should also be considered.¹⁷⁹

Reducing the Use of Wild-Caught Fishes in Farmed Fish Feed

The tremendous number of small wild-caught fishes used as feed may be the most important consideration for reducing suffering in fish farming. A 2012 Fishcount estimate suggests that from 2005–2009, several hundred billion wild-caught fishes were used annually to make the fish oil and fishmeal for aquaculture.¹⁸⁰ It seems that in most cases, these wild-caught fishes are small pelagic fishes such as anchovies, mackerel, and herring.¹⁸¹ Some advocacy groups have expressed interest in requesting that fish farming practices use fewer wild-caught fishes by moving towards the highest percentage of plant-based feed possible.^{182, 183} Alternatively, they seek to switch farming practices from species that consume a large amount of wild-caught fish, like Atlantic salmon,¹⁸⁴ to species that can be farmed on a herbivorous diet at a lower cost.

There seems to be uncertainty abound the potential impact of such changes on wild animals. For instance, as Delon & Purves (2017) describe, the complex nature of ecosystems makes it very difficult to predict the extent to which some intervention would reduce, rather than exacerbate, wild animal suffering.¹⁸⁵ In a 2017 conversation with Matt Ball of One Step For Animals, he noted that “wild fish will die in some painful way, regardless of if they are caught by humans.”¹⁸⁶ Whether using wild-caught fishes in farmed fish feed leads to decreased net-welfare for wild animals may not immediately be clear. The general question of how different interventions affect wild fish welfare has been explored in somewhat more detail by Ray (2018), Tomasik (2015) A, and Tomasik (2015) B.¹⁸⁷ A key takeaway from these sources is that the net impact of interventions on wild fish welfare seems very uncertain. In these analyses, it appears that quality of life is a crucial consideration—if quality of life is generally positive for wild fishes, then reducing the use of wild-caught fish feed may positively impact them. If it is generally negative, then the net impact seems more uncertain.

Another point to consider is that some carnivorous fish species may have better survival rates and higher weights at slaughter per time spent on farms than some herbivorous species. For instance, consider Table 4 from the appendix indicating that compared to common carp, Atlantic salmon may have much better survival rates (~65% vs ~21%) and harvest weight per year of farming (~1,160–3,800 g/yr vs ~125–2,500 g/yr). These findings suggest that switching demand from farmed Atlantic salmon to farmed common carp may not be ideal—common carp could require significantly more farmed fish years than Atlantic salmon in order to satisfy equivalent demand levels. Similarly, shifting to plant-based feeds could potentially impact growth rates, in turn potentially increasing the number of farmed fish years required to meet the demand.¹⁸⁸ We don’t have a good sense of how greater reliance on plant-based feed would impact farmed fish prices. Our initial impression is that it seems fairly likely to cause increases in prices, at least in the short term.¹⁸⁹ We encourage considering all of these questions when attempting to determine the most effective reforms we should seek to reduce suffering.

Tentative Conclusions

Generally, the initial effects of changing common slaughter methods from asphyxia and live chilling to properly applied electrical or percussive stunning seem promising. The initial effects of improving DO levels seem promising as well. An initial survey of the evidence also seems to support the welfare benefits of opposing triploidy, at least for Atlantic salmon.

There are a few interventions that we currently feel are less promising because the evidence supporting them is mixed or lacking. These include (i) environmental enrichment and (ii) some methods for reducing the number of wild-caught fishes used in farmed fish feed. For environmental enrichment, further research may help us identify promising interventions. On the other hand, the effect of reducing the amount of wild-caught fishes in feed seems fraught with large degrees of uncertainty. Further, when advocating for switching a carnivorous species to plant-based feed or farming an herbivorous species instead of a carnivorous species, it’s worth considering the average survival rates, weight at slaughter, and lifespan of different farmed fish species.

We hope that further research will decrease our uncertainty about the following promising interventions:

• The use of anaesthesia to reduce stress at time of slaughter¹⁹⁰
• The use of sedatives during transport
• Minimizing of prolonged handling of fishes, especially out of water
• Water quality parameters other than dissolved oxygen
• Farmed fish genetics
• Stocking density¹⁹¹

We also hope to develop a stronger understanding of the other factors that affect the promise of farmed fish interventions, including the relevant economic considerations¹⁹² and the ethology of the most numerous farmed fish species. We encourage further discussion and analysis of promising farmed fish welfare interventions. Any interested parties are encouraged to contact us.¹⁹³

Questions for Further Consideration

There are many questions—some of which are listed below—that could cause us to significantly update our views.

• In the current farming systems, do some of the most numerous farmed fish species have significantly different welfare levels than others?
• How promising are the welfare indicators that we didn’t consider (e.g., indicators regarding the presence of various pathogens and parasites, indicators regarding injuries and physical deformities, genetic indicators, and indicators derived from preference tests)?
• How promising are the interventions that we didn’t consider (e.g., stocking density, use of sedatives, and water quality parameters other than dissolved oxygen)?
• What are the humane slaughter parameters for some of the most numerous farmed fishes such as crucian carp, silver carp, and grass carp?¹⁹⁴
• Are there better, yet still feasible, methods other than electrical and percussive stunning for reducing pain and distress before slaughter?
• What are the optimal dissolved oxygen parameters for some of the most numerous farmed fishes such as crucian carp, silver carp, and grass carp?

Check the resources

---

1. Much of the work of our Top Charities—such as advocating for individual diet change and contending for legislative changes—could contribute to changing societal norms and therefore may positively affect farmed fish welfare. On the other hand, some of these charities’ individual diet interventions may also cause consumers to replace their consumption of farmed land animals with fish, thereby increasing the demand for farmed fish products. 

2. For example, corporate outreach, online advertising, and undercover investigations could all be used to advocate for farmed fishes. 

3. For instance, see responses from The Humane League and Compassion in World Farming USA to the question: “There are many more farmed fishes than other species of farmed animals. Has your charity considered allocating more of their resources towards farmed fish advocacy?” 

4. For instance, blog posts by organizations such as PETA, Farm Sanctuary, and The Humane League promote fish sentience and highlight reasons to abstain from fish consumption. 

5. Animal Equality documented bluefin tuna slaughter, Mercy For Animals investigated farmed catfish slaughter, Essere Animali investigated sea bream, sea bass, and trout farms, and Animals Now has investigated fish farms. 

6. A non-exhaustive list follows:
   – Eurogroup for Animals lobbies E.U. and national governments. They also coordinate campaign work.
   – The Norwegian Animal Protection Alliance works to improve conditions for farmed fishes in Norway.
   – Compassion in World Farming is lobbying to increase fish welfare concerns within the European Commission.
   – The Humane Slaughter Association (HSA) appeared to play a central part in changing fish slaughter practices to techniques that involve less suffering.
   – Dutch animal protection societies have campaigned for the humane treatment of farmed fishes.
   – The RSPCA uses welfare labels that aim to convey to consumers whether farmed fishes are treated according to humane standards. 

7. For instance, see this post by Open Philanthropy Project examining why the U.S. corporate cage-free campaigns are succeeding. 

8. It isn’t totally clear that farmed fish advocacy would receive less public support than farmed land animal advocacy. For example, one Eurogroup for Animals/Compassion in World Farming survey (see Table 31/1 of their report) suggests greater support for better welfare protection of salmon than of pigs. Harrison & Hall (2010) suggest the opposite effect. 

9. The scale of a cause refers to its duration, incidence, and intensity. The most important causes deal with severe, large-scale problems, while less important causes might deal with local problems or with problems that have less severe impacts on those affected. For further information, please see our report on cause prioritization. 

10. While we haven’t deeply researched key facts about the capacity of fishes to suffer compared to other commonly farmed animals, it seems that general scientific opinion is in favor of fishes having the capacity to feel pain. For example, see this article thread for a list of responses to “Why Fish Do Not Feel Pain” (Key, 2016). 

11. Our rough impression is that farmed fishes generally have a quality of life that is perhaps comparable to or worse than that of most farmed birds. The sources of some major welfare concerns seem similar: they include high stocking densities that don’t align with natural preferences, clear indications of poor physical health, and high pre-slaughter mortality rates, among other things. 

12. Mood & Brooke (2012) use 2010 FAO data to estimate a range of 36.7 billion to 121.8 billion individual farmed fishes (midpoint approximately 80 billion). Following the same methodology, we use 2015 FAO data and estimate a range of 52 billion to 167 billion farmed fishes (midpoint 109.5 billion). See the appendix for more information. 

13. FAO (2016) notes that “[b]y 2014, a total of 580 species and/or species groups farmed around the world, including those once farmed in the past, had been registered with production data by FAO. These species items include 362 finfishes (including hybrids), 104 molluscs, 62 crustaceans, 6 frogs and reptiles, 9 aquatic invertebrates, and 37 aquatic plants.” Since drafting this report but prior to its publishing, the FAO released an updated report on the state of world fisheries and aquaculture.  

14. See Mood & Brooke (2012). 

15. Note that this question is different from the following important research question: How many farmed fishes are alive at any given moment based on the average of various species’ commercial lifespan?  

16. Mood & Brooke (2012), p. 2 report using FAO data from 2010 along with estimates of mean weights at harvest for different species to estimate the numbers of fish represented by global farmed fish production. These estimated mean weights were based on, and extrapolated from, fish size data for individual fish species obtained from a range of sources. 

17. The 2017 FAO data is now available. 

18. See Mood & Brooke (2012) for more information. Mood & Brooke have been updating their estimates as new FAO datasets are released, and now estimate that between 51.1–167.4 billion (midpoint 109.5 billion) farmed fishes were slaughtered in 2017. 

19. Mood & Brooke estimated using 2015 FAO data that between 48.0–160.5 billion (midpoint 104.3 billion) farmed fishes were slaughtered that year. We are not sure why there is a discrepancy between our estimates. 

20. For example, ACE estimates that for the four most consumed farmed finfish in the U.S. the mortality rate prior to slaughter is 18%–60% for salmon, 5%–35% for tilapia, 10%–38% for pangasius, and 12%–65% for catfish (as 90% subjective confidence intervals). Please see Table 4 of the appendix for further information. 

21. “Production of freshwater fish in 2015 was dominated by carps (Cyprinidae, 20.4 million tonnes or 71.1 percent). The largest producer of all carps (cyprinids) in 2015 is China (78.7 percent), followed by India (15.7 percent).” (FAO: World Aquaculture 2015: A Brief Overview (2017), p. 8)  

22. For example, this FAO fact sheet notes that “China generates 67% of world aquaculture production of fish, crustaceans and molluscs (32,414,000 tonnes in 2005).” 

23. “In Europe, however, the farming of mainly salmonids is preferred, including salmon and trout along with turbot. Norway is the most important aquaculture nation in Europe, with about 1 million tonnes of farmed fish, followed by Spain with a good 250,000 tonnes; France takes third place with 220,000 tonnes.” (World Ocean Review: A Bright Future for Fish Farming (2013), p. 83)  

24. From Mood & Brooke (2012), p. 25 

25. “While all regions are expected to expand their aquaculture production, the largest expansion is expected in SEA [southeast asia] and IND [India]. SEA is expected to represent 15.9 percent of global aquaculture production in 2030, while IND would represent 9.2 percent. LAC [Latin America and Caribbean] and South Asia (excluding India) (SAR) are also projected to experience large aquaculture growth over the 2010–30 period. Middle East and North Africa (MNA) and Sub-Saharan Africa (AFR) also show substantial expected growth over this period, but they begin from much lower production levels in 2010 compared to other regions.” (World Bank: Fish to 2030 (2013), p. 41)  

26. “Compared to agriculture and livestock sectors, except for a few countries, aquaculture could still be considered a young food producing sector.” (FAO (2015), p. 2)  

27. See Table 24.1 from Amos & Sullivan (2018) where inland aquaculture is indicated as accounting for 87% of finfish farming by volume. Note that the table also provides information about the volume of fish farming in the 12 countries with the greatest production levels. 

28. “Across a variety of species, the steps and methods of aquaculture production are principally the same. To obtain young, two primary systems are employed: 1) eggs and milt, a sperm-containing secretion of the testes of fish, are hand-collected from broodstock and mixed to induce fertilization or 2) broodstock spawn in captivity and fertilized eggs or swimming fry (young, post-larval fish) are subsequently collected. When starting with fertilized eggs, they are incubated until hatched. When fry begin actively searching for food, they are collected and transferred to a nursery to grow and feed until a preset time, size, or mass is reached: perhaps 2–3 months for tilapia, 4–6 months for catfish and trout, 6–8 months for bass, and 8–16 months for salmon. Next, the fish are transferred from the nursery to a grow-out facility where they remain until reaching market size: 5–6 months for tilipia, 15–18 months for catfish, 15–20 months for trout, 18–24 months for bass, and 18–36 months for salmon. Times for all species can vary widely by many months depending on water temperature and quality, feed quality and availability, and stocking density.” (HSUS (2008), p. 2)  

29. “Ponds can be natural or artificial, typically with low water refreshment. Tanks are often fiberglass with a high water turnover rate. Raceways are long, linear structures designed so water flows into one end and out the other with high turnover. Cages or pens are usually made from mesh or net screens and are submerged in larger bodies of water, often lakes or seas for species requiring saltwater.” (HSUS (2008), p. 2)  

30. See page 37 of Potts et al. (2016). 

31. We did not find any FOS standards that were directly related to farmed fish welfare after quickly skimming marine aquaculture and inland aquaculture standards.  

32. “GLOBAL G.A.P. accounted for almost half of all certified aquaculture production, while BAP, ASC and FOS shared near-equal portions of the remainder.” (Potts et al. (2016), p. 37)  

33. See figure 2.11 from Potts et al. (2016).  

34. “The RSPCA standards are the best standards. They are evidence-based and updated annually. A lot of other standards don’t make such changes so frequently. The RSPCA to date has considered welfare in the same way THL would, whereas some other groups allow mutilations and other undesirable things. The RSPCA now has wrasse standards and that was certainly something that was missing. The RSPCA are also improving the transport times and reducing them more and more, although they still need to reduce them even more. The RSPCA also need to improve the handling and issues with starvation. There are always things the RSPCA standards can improve on, but they are as good as it gets right now.” —Conversation with Aaron Ross and Vicky Bond of The Humane League (2018) 

35. “RSPCA Assured has the scientific and veterinarian community backing it, it is more independent, publishes papers rather than receiving backing from the industry, no stakeholders from the industry despite being well versed in the industry’s conditions. Disadvantages include that it is eurocentered. However, they have clear guidelines and a robust auditing method.” —Conversation with Doctor Lynne Sneddon of University of Liverpool (2018) 

36. For example, the RSPCA Welfare Standards for Farmed Atlantic Salmon (2018) note that the “standards for farmed Atlantic salmon (Salmo salar) are used to provide the only RSPCA-approved scheme for the rearing, handling, transport and slaughter/killing of farmed Atlantic salmon. The standards cover the two distinct phases of farming (freshwater and marine farming)” (p. iii) and that “[a]t present the RSPCA welfare standards for farmed Atlantic salmon only apply to diploid fish. The RSPCA is monitoring the work being done on the specific needs of triploid salmon, such as diet, management and health care, in order to investigate how and whether their welfare needs can be properly satisfied, which will ultimately determine whether they will be allowed to be used in future under the standards” (p. 36). It seems that the standards don’t cover recirculated aquaculture systems.  

37. For instance, our literature search returned hundreds of relevant results while our literature search for the leafleting intervention report returned many fewer relevant results.  

38. Two individuals we spoke with mentioned that very little welfare research exists on catfish compared to salmon. See our Conversation with Professor Victoria Braithwaite of Penn State University (2018) and our Conversation with an Anonymous Animal Welfare Specialist from a Major Animal Advocacy Organization (2018). 

39. For instance, “[d]ifferent species display a wide variation in physiological responses to stressors associated with aquaculture. Elevations in plasma cortisol levels can differ by as much as two orders of magnitude among different species of fish following identical stressors […].” (Ashley (2007), p. 8)  

40. EFSA 2009 note that there is a “marked interspecies variation in brain anatomy, often reflecting sensory specialization, fundamental differences in embryonic development, or the degree of cell migration and proliferation (Butler, 2000)” (p. 11). Lines & Spence (2012) also note that “[s]ome other species of farmed fish including catfish, pangasius, carp and tilapia are very resistant to percussive stunning due to the shape and protection of the skull (Marx et al., 1999; Lambooij et al., 2007). Electrical stunning may be a more suitable stunning method for these species” (pp. 157–158). Some probably important areas where differences between these fish species would influence their welfare requirements include the extent to which the fishes in question are benthic or broadly pelagic, solitary or social, and migratory or sedentary. So, for instance, while the benthic, relatively solitary, and sedentary fishes may strongly prefer certain substrate, relatively low stocking density and relatively little movement, a pelagic fish may place less importance on substrate, strongly prefer higher stocking densities in order to exhibit schooling behavior, and/or prefer a greater amount of movement to attempt to satisfy migratory needs. 

41. Huntingford et al. (2006) note the varied behavioral profiles amongst individuals of the same species and Huntingford & Kadri (2008) state that individual fishes exhibit different behavioral strategies and varied magnitudes of neuroendocrine responses to stressors. Fraser et al. (2012), p. 193 also indicate that “[r]ecent reviews have not only highlighted the species-specific differences in fish welfare requirements, but also the differences due to within-species genomic variation (Chandroo et al., 2004). Previously, it had been suggested that triploids could be considered a separate species to diploids (Benfey, 2001) based on the physiological differences between triploids and diploids (reviews, Benfey, 1999; Piferrer et al., 2009), adding emphasis to the need to evaluate triploid fishes separately to their diploid counterparts.” 

42. For example: “Most fish welfare research has been directed towards determining if the fish are healthy and growing well, with the assumption that good health and growth is a meaningful indicator of good welfare (e.g., Turnbull et al. 2005).” (North et al. (2008), p. 244)  

43. For instance, the RSPCA Atlantic Salmon standards note that for Atlantic Salmon (the farmed fish species that seems to have been most studied) “difficulties arise in specifying details in relation to several issues (for example, acceptable maximum stocking densities) due to the lack of scientific research examining fish welfare under different commercial systems.” What’s more, despite the annual farming (in the billions) of Grass Carp, across ten key welfare criteria in the relatively comprehensive Fish Ethology Database there is no welfare area that achieves a score of “high” certainty. 

44. For a specific example, see the Managing Dissolved Oxygen Levels section of this report. 

45. For instance, Stien et al. (2013), pp. 38–39 lists the following seventeen welfare indicators: “water temperature, salinity, oxygen saturation, water current, stocking density, lighting, disturbance, daily mortality rate, appetite, sea lice infestation ratio, condition factor, emaciation state, vertebral deformation, maturation stage, smoltification state, fin condition, and skin condition.” 

46. “Most fish welfare research has been directed towards determining if the fish are healthy and growing well, with the assumption that good health and growth is a meaningful indicator of good welfare (e.g., Turnbull et al. 2005).” (North et al. (2008) p. 244)  

47. Stien et al. (2013) calculate the weighting factor for each welfare indicator according to the method proposed by De Mol et al. (2006). In a 2017 Open Philanthropy Project report, Cotra notes that “[w]e don’t consider FOWEL [the model used in De Mol et al. (2006)] to be a very reasonable procedure for mapping housing descriptions to welfare scores. In particular, the translation of individual sentences in the scientific literature into associations between attribute levels and welfare categories did not strike us as a useful modeling choice. When we requested access to a partial sample of the database of scientific statements and their weighting scores, we did not find it straightforward to translate most of them into a relation between an attribute level and a weighting category with a certain degree of intensity. We disagreed multiple times with the association intensity or welfare category chosen in the paper.” What’s more, EFSA (2008) also notes that “[t]he percentage of the individuals affected by the hazard which are likely to die as a result of exposure to the hazard were scored (see Table 9). This parameter is not used in the risk estimate calculation.” This seems like a suboptimal analysis choice due to the fairly clear relationship between mortality and welfare.  

48. See Martins et al. (2012). 

49. See Huntingford et al. (2008). 

50. “For instance, bottom-feeding flatfish have been shown to have improved welfare (measured as feed intake and feeding motivation) when fed sinking pellets compared with floating pellets (Kristiansen & Fernö 2007) […]. In brief, any observed changes in foraging-related behaviours, unless coupled with a change in management practices, is highly likely to be an indicator of changes in either the motivational state, health and/or welfare status of fish. This is supported by studies showing similar central signalling systems (corticotrophin-releasing factor, CRF/urocortin activation) involved in both stress-induced anorexia and negative emotions, at least in higher vertebrates (Heinrichs & Koob 2004).” (Martins et al. (2012), p. 21)
    “Food-anticipatory activity is usually considered as an indicator of good welfare in fish (Folkedal et al., 2012; Kristiansen et al., 2004; Martins et al., 2012; Spruijt et al., 2001) since it may indicate a behavioral priority for foraging instead of being prostrate.” (Colson et al. (2014), p. 156)  

51. “Indicators associated with the response to chronic stress, including physiological outcomes, disease status and behaviour, provide a potential source of information on the welfare status of a fish. These include loss of appetite, impaired growth and muscle wasting, immunosuppression, suppressed reproduction and reduced cognitive ability. Clearly, observing such changes strongly suggests that the well-being of the fish has been significantly compromised […].” (Braithwaite & Ebesson (2014), p. 247)  

52. “It has been suggested that the vertical swimming behaviour in Atlantic halibut (H. hippoglossus L.) described above (Kristiansen et al., 2004) and the circular shoaling behaviour seen in Atlantic salmon (S. salar L.) in sea cages (e.g. Oppedal et al., 2001; Juell et al., 2003) may represent abnormal behaviour or a kind of stereotypy comparable to the pacing seen in some zoo animals (Lymbery, 2002; Kristiansen et al., 2004). Stereotypies are fixed sequences of behaviour performed repetitively in the same way with no obvious function and have been used as indicators of reduced welfare (Mason, 1991a,b). While some stereotypies cause injuries and are clearly of welfare concern, the consequences of others are not so clear and there is no one to one correspondence between stereotypy and welfare (Barnett and Hemsworth, 1990; Dantzer, 1991; Mason, 1991a,b). The causes of stereotypies and what they tell us about welfare are a matter of debate (Dawkins, 1998).” (Ashley (2006), p. 18)  

53. “Assessments of group behaviour can be used as an operational on-farm welfare indicator and are what most fish farmers use daily to evaluate the hunger, stress level and health status of fish. Group swimming behaviour is defined as the spatial distribution and swimming activity of groups of fish held within an aquaculture production unit and covers shoal structure, polarisation, the horizontal and vertical distribution of the group and their swimming speed and direction.” (Martins et al. (2012), p. 26)  

54. “Ventilatory activity is the flow of water ventilated over the gills per unit time. In brief, during routine inspection, sustained hyperventilatory activity does not tell the observer much about the origin and/or intensity of a stressor (see Barreto & Volpato 2004) but, combined with other observations (such as location in the tanks or reactivity), it can be a sensitive indicator of fish welfare. Furthermore, correlations between increased ventilatory activity and other welfare indicators, such as blood lactate, glucose and haematocrit, are well established (e.g. White et al. 2008). Nevertheless, one should interpret an increase in ventilatory activity with caution as an increase may be linked with positive experiences and therefore not necessarily with poor welfare.” (Martins et al. (2012), p. 12)  

55. For example, “[t]wo common behavioural indicators of poor welfare in farmed fish are aggression and swimming activity.” (Lopez-Olmeda & Vasquez (2011), p. 147)  

56. “An individual’s responsiveness to stress is influenced by its coping style or behavioural syndrome, which should be considered when interpreting the variation in behavioural responses. The extent to which an individual is bold or shy should not be used as a welfare indicator, but one may infer a welfare problem when the behaviour defined under the bold/shy continuum changes; for example, when a bold individual starts exhibiting typical shy behaviours such as decreased exploratory behaviour, one may suspect that it is sick or chronically stressed. Still, one should interpret such changes with care, as coping styles are considerably plastic and may change with context (e.g. Ruiz-Gomez et al. 2008).” (Martins et al. (2012), p. 20)  

57. Lopez-Olmeda & Vasquez (2011) and Ellis et al. (2012) note that cortisol levels are a commonly used indicator for poor fish welfare. 

58. “The most objective means available for assessing the presence of negative feelings in animals is generally accepted (although perhaps not explicitly) to be the measurement of stress response indicators.” (Ellis et al. (2012), pp. 174–175)  

59. “The stress response in fish is very similar to that of other vertebrates and can be described in three stages (Barton, 2002; Iwama, 2007). The primary response involves the activation of two neuroendocrine axes. The hypothalamus-sympathetic-chromaffin cell axis produces catecholamines (adrenaline and noradrenaline) from the chromaffin cells, the equivalent of adrenal medulla in tetrapods. The second axis is the hypothalamic-pituitary-interrenal tissue (HPI) axis, with the production of corticosteroids (mainly cortisol, in teleosts) from the interrenal tissue, the equivalent of adrenal cortex in tetrapods.” (Galhardo & Oliveira (2009), p. 4)  

60. See Ellis et al. (2012) for more information. 

61. For more information, see Otovic & Hutchinson (2015). 

62. “For example, cortisol secretion is increased by exercise and by presumably pleasurable activities such as mating, nursing and exploration.” (John Wiley & Sons (2013), p. 120)  

63. Braithwait (2017) notes that “[f]or many animals this requires that a blood sample be taken, but as circulating cortisol is secreted across fish gills and in their urine, the water around the fish can be sampled to assay cortisol production. This technique has several advantages to taking blood samples; the fish can be repeatedly sampled with very little handling required and the fish do not need to be anaesthetized.” 

64. See Ellis et al. (2011). 

65. “In some rearing systems, such as salmon sea-cages, the removal of dead fish is a regular daily process, economically justified by the need to reduce possible sources of disease. This is achieved by hand nets, SCUBA divers, hauling of collection ‘socks,’ and commercial suction devices (Sangster 1991; Soares et al. 2011). Mortality records from such collections have recently been used to retrospectively benchmark mortality rates through the production cycle (Soares et al. 2011).” (Ellis et al. (2011), p. 196)  

66. “Aunsmo et al. (2008) provide the example of bacterial ulcers: although bacterial infection ultimately caused death, infection could not have occurred without an initial mechanical trauma. However, the mechanical trauma would not, on its own, cause death that only ensued after infection.” (Ellis et al. (2011), p. 190)  

67. Some reviews on farmed fish cannibalism include Pereira et al. (2017), Hecht & Pienaar (1993), and Baras & Jobling (2002).  

68. See Table 4 in the appendix and Ellis et al. (2011), p. 19 for more information. 

69. See Table 4 in the appendix for more information. 

70. According to Ellis et al. (2011), p. 193, lamb mortality rates are generally around 10%–15% and broiler hen mortality rates are less than 2% during rearing to 18 weeks. However, the commercial lifespan of broilers is only 6–8 weeks so the mortality per unit of time for broilers may be quite comparable to that of farmed fish. For instance, the National Chicken Council reports an estimated 4.4% mortality rate over a 47-day lifespan for broilers. Note that the incentives for the National Chicken Council are such that they are more likely to underestimate than overestimate the mortality rate. Still, that would be the equivalent to a daily mortality rate of ~0.09% for broilers. For comparison, calculating the average daily mortality rate for farmed fish using the information in Table 4 from the appendix gives an estimated average daily mortality rate of ~0.13%.  

71. Cooke (2016), Compassion in World Farming (2009), North et al. (2008), and MacIntyre et al. (2008) all suggest that water quality is an essential indicator of farmed fish welfare. However, it is important to note that the literature may generally falsely equate health with welfare. 

72. Stien et al. (2013) report that the four variables with the highest weighting factors are: mortality (% per day) with a weighting factor of 21, DO levels with a weighting factor of 17, temperature with a weighting factor of 16, and emaciation state with a weighting factor of 16.  

73. See MacIntyre et al. (2008) for further information about the various indicators of water quality.  

74. “Temperature is a vitally important physical property of the water in aquaculture systems. The temperature of the water regulates the amount of dissolved oxygen that a body of water can hold, the rate of decomposition and photosynthesis, which will affect the oxygen demand in pond systems and the ionisation of ammonia […] (Colt & Tomasso 2001). Additionally, increasing temperature increases the growth and infectiousness of many fish pathogens (Roberts 1975) and increases the toxicity of many dissolved contaminants (Wedemeyer 1996). All of these factors have the capacity to compromise the health of farmed fish. As fish are exothermic, increasing the water temperature increases the metabolic rate and hence oxygen consumption.” (MacIntyre et al. (2008))  

75. As defined by the Biology Online Dictionary, dissolved oxygen is “[t]he amount of free oxygen dissolved in water, expressed in mg/L, parts per million (ppm), or in percent of saturation, i.e. where saturation pertains to the maximum amount of oxygen that can be dissolved theoretically in water at a particular altitude and temperature.”  

76. For example, both Stien et al. (2013) and Waldrop et al. (2018) present DO levels as an essential indicator of farmed fish welfare.  

77. For more information about the EFSA’s risk scores for Atlantic salmon, download the supplementary material table from the report’s supporting information section, and see the “water oxygen content” row in the “abiotic factors – ongrowing” sheet. Note that the EFSA defines salmon ongrowing as “the sea-water stage of the life cycle, beginning at the time when fish are properly adjusted to the sea-water environment and ending at slaughter (usually 13–20 months after sea-water transfer).”  

78. “Bhatnagar et al. (2004) also suggests that DO levels of 1–3 ppm has sublethal effect on growth and feed utilization, while DO levels of 0.3–0.8 ppm is lethal to fishes. Ekubo and Abowei (2011) also cautioned that fish are likely to die if exposed to less than 0.3 mg L-1 of DO for a long period of time.” (Makori et al. (2017), p. 27)  

79. “In extreme cases, acute hypoxia results in mass mortality (Thronson & Quigg 2008, Stauffer et al. 2012); in less extreme cases, suboptimal DO concentrations result in decreased growth, appetite, immune function, swimming performance and fish welfare (Oppe – dal et al. 2011a, Remen et al. 2012, 2014, Burt et al. 2013, Kvamme et al. 2013).” (Oldham et al. (2017) p. 145–146)  

80. See Franklin (2013), p. 114 for more information. 

81. “Oxygen depletion in water leads to poor feeding of fish, starvation, reduced growth and more fish mortality, either directly or indirectly (Bhatnagar and Garg, 2000).” (Bhatnagar (2013), p. 1983)  

82. See for instance Figure 1, p. 40 of Boyd & Hanson (2010). The following sources also outline the effects of DO on feed conversion:
    – “Channel catfish growth is significantly lower when fish are reared under conditions of constant low compared to constant high DO concentration (% saturation) because feed intake decreases as DO percent saturation decreases and results in slower growth (Andrews et al. 1973; Buentello et al. 2000).” (Green et al. (2012), p. 188)
    – “In this regard, Bergheim et al. (2006) and Duan et al. (2011) reported that fish growth and feed efficiency were affected by DO availability, and fish always showed good feed efficiency when fed at enough DO in water. Abdel-Tawwab et al. (2014) reported that growth of Nile tilapia was significantly retarded at low DO level. The low feed intake and low growth observed in fish at low DO conditions were because fish appetite and digestibility was reduced (Tran-Duy et al. 2012; Gan et al. 2013). Thus, it could be concluded that high growth under normal DO conditions resulted mainly from better feed consumption and nutrient digestibility. Similar results have been obtained from channel catfish, Ictalurus punctatus (Buentello et al. 2000); Hippoglossus hippoglossus (Thorarensen et al. 2010); Japanese flounder, Paralichthys olivaceus (Duan et al. (2011); and grass carp, Ctenopharyngodon idella (Gan et al. 2013); all these fish experienced reduced feed intake and growth under hypoxic conditions.” (Abdel-Tawwab & Monier (2015), pp. 1,268–1,269)
    – “Reduced growth and feed consumption resulting from chronic hypoxia have been observed in a wide variety of fish species, including Nile Tilapia (Teichert‐Coddington and Green 1993), Striped Bass Morone saxatilis (Brandt et al. 2009), and Rainbow Trout (Bernier and Craig 2005; Glencross 2009). (Torrans et al. (2015), p. 489)
    – “Feeding is often strongly affected because search, digestion and food assimilation are significant components of many fishes energy budget and thus are limited by oxygen availability (Doudoroff & Shumway 1970; Remen et al. 2012).” (Franklin (2013), p. 114)  

83. According to Oldham et al. (2017) p. 145, DO varies with changing light, temperature, currents, wind, and rainfall. Compassion in World Farming (2009), p. 5 also notes that the flow rate of water influences DO and the dilution and dispersal of waste. Section 1.1 of FAO (1984) lists time of day as a factor influencing DO. 

84. See Alabaster & Lloyd (1982), p. 129, Boyd & Zimmerman (2010), p. 240, and MacIntyre et al. (2008), p. 169 for more information.  

85. See Colt (2006), p. 152 for more information. 

86. “Mood & Brooke (2012), pp. 25–26 note that highly intensive pangasius catfish farming has greatly expanded in Vietnam in the last decade due in part to their “ability to breathe atmospheric oxygen, which makes them able to tolerate low levels of dissolved oxygen and highly polluted water.” Also see EFSA (2009), p. 3 for more information about different species’ ability to live in relatively low DO levels. 

87. “Sensitivity of fish to low concentrations of dissolved oxygen (DO) differs between species, between the various life stages (eggs, larvae, and adults), and between the different life processes (feeding, growth, and reproduction, which in turn may depend on swimming ability, and specialized behaviour which may also be influenced by DO). Any DO standards set for fisheries must take all these into account, bearing in mind the type of fishery, the times and places the fish occur, and the likely impact on the fishery of impairment of each part of the life cycle.” (Alabaster & Lloyd (1982), p. 129)  

88. From Ashley (2007), p. 13 

89. “High densities are associated both with increased competition, aggression, and physical injury (due to increased contact between fish and between fish and the housing/net), and with degradation in water quality.” (Ashley (2007), p. 14)  

90. “High densities are associated both with increased competition, aggression, and physical injury (due to increased contact between fish and between fish and the housing/net), and with degradation in water quality.” (Ashley (2007), p. 14)  

91. “One trend that has been observed repeatedly is damage to fins as stocking density increases. This has been observed in both rainbow trout and Atlantic salmon (MacLean et al. 2000, Turnbull et al. 2005, North et al. 2006, see Chapter 9).” (Turnbull et al. (2008), p. 115)  

92. According to Bosworth et al. (2015), p. 441, individual growth rates of channel catfish decrease with increased stocking density. Similarly, Sahin et al. (2014), p. 134 note that rainbow trout growth rates decrease with higher stocking density and Garcia et al. (2013), p. 55 observe the same effect for tilapia. Salas-Leiton et al. (2008), p. 87 observe a more complex relationship between growth rates and stocking density, nothing that “[m]ost published studies describing an inverse relationship between growth parameters and the rearing stocking density were carried out using juveniles within a weight range of 1–15 g. This could suggest stocking density as a more likely factor limiting fish growth during early phases of the overall ongrowing period. It has been shown that the putative negative effects caused by stocking density on growth parameters seem to be mitigated as fish approaches to adult size (Duarte et al., 2004; Sánchez et al., 2007). There is a maximum stocking density beyond which point growth is depleted in some flatfish species (Björnsson, 1994; Howell, 1998; Schram et al., 2006).” 

93. See Adams et al. (2007), p. 337 for more information about the impact of stocking density on mortality rate. Also see Coulibaly et al. (2007) p. 65 for information about the link between cannibalism and high stocking densities. 

94. For example, Lupatsch et al. (2010), p. 245 and Sahin et al. (2014), p. 132 state high stocking density as an important stress factor for farmed fishes. 

95. For example Sirakov & Ivanchena (2008), p. 150 and Roberts (2012), p. 205 state high stocking density as an important source of diseases for farmed fishes.  

96. For example, Hosfeld et al. (2009), p. 236 state high stocking density as a cause for deterioration of water quality “due to a reduction in DO, a build up of fish metabolites and carbon dioxide followed by a reduction of pH level.” 

97. For instance, “[r]educed growth and feed conversion ratio during chronic stress conditions have been attributed to a change in metabolism. This effect is based on the assumption that coping with stress increases the fish’s overall energy demand, which is then unavailable for growth (Wendelaar Bonga, 1997). On the other hand decreased feed consumption (Vijayan and Leatherland, 1988) social interaction (Papoutsoglou et al., 1998) and altered water quality (Pickering & Pottinger, 1987) may result in increased metabolic demands and additional expenditure of energy occurring at the expense of growth.” (Lupatsch et al. (2010), p. 245)  

98. “There is dispute as to the cause of the observed effects of increasing density, with water quality deterioration and/or an increase in aggressive behaviour being variously proposed. Both causes can theoretically generate the observed effects of increasing density, and the relative contribution of the two causes may depend upon the specific conditions. However, documentation of the relationship between density and the effects of aggressive behaviour at relevant commercial densities is lacking. Consequently only inferential evidence exists that aggressive behaviour generates the observed effects of increasing density, whereas there is direct experimental evidence that water quality degradation is responsible.” (Ellis et al. (2002), p. 493)  

99. “Reports of increased stocking density negatively affecting fish growth and other production traits are common and are typically attributed to direct effects of density such as social interactions and competition for food, indirect effects of increased density on degradation of water quality, which results in reduced production, or a combination of direct and indirect effects (Holm et al. 1990; Barcellos et al. 1999; Le Ruyet et al. 2008; Offem & Ikpi 2013).” (Bosworth et al. (2015), p. 441)  

100. “Several studies performed suggest no negative effects resulting from high rearing densities when maintaining water quality parameters at an adequate level in all density groups (North et al., 2006; Ellis et al., 2002; Bagley et al., 1994; Soderberg et al., 1993; Kebus et al., 1992; Blackburn and Clarke, 1990; Kjartansson et al., 1988; Soderberg and Meade, 1987; Soderberg and Krise, 1986).” (Hosfeld et al. (2009), p. 236)  

101. “Here we report on broiler welfare in relation to the European Union proposals through a large-scale study (2.7 million birds) with the unprecedented cooperation of ten major broiler producers in an experimental manipulation of stocking density under a range of commercial conditions. Producer companies stocked birds to five different final densities, but otherwise followed company practice, which we recorded in addition to temperature, humidity, litter, and air quality. We assessed welfare through mortality, physiology, behaviour and health, with an emphasis on leg health and walking ability. Our results show that differences among producers in the environment that they provide for chickens have more impact on welfare than has stocking density itself.” (Dawkins et al. (2004), p. 342) During the review of this report, one reviewer noted that this could be misleading as the difference between the stocking range of 42 and 34 kg/m² could be much more significant than the difference between the range of 34 and 25 kg/m² if the increase in suffering is non-linear as the stocking density increases. 

102. “Further analysis of the relationship between stocking density and the welfare score suggested that there was no trend up to an inflection point around 22 kg/m³, after which increasing stocking density was associated with lower welfare scores. This was a weak but statistically significant association.” (Turnbull et al. (2008), p. 117)  

103. For example, see Table 5 of Segner et al. (2011) for a list of various pathogens associated with rainbow trout farming. Some of the pathogens listed include various viruses, bacteria, and parasites.  

104. “A very different approach to gaining a broad oversight of the welfare of farmed fish is potentially provided by using modern molecular tools to explore patterns of gene expression in fish exposed to different challenges (25). Information obtained from studying the complete set of RNAs encoded by the genome of (in this case) a fish held under a specific set of conditions (genome-wide transcriptomics), complemented by the huge amount of information on the functions of known genes in other organisms, can give an integrated picture of the overall response of fish to aquaculture-related challenges. For example, in seabass, discrete clusters of genes can be identified based on their specific temporal patterns of expression following confinement stress. These patterns of expression are related to rapid metabolic activation, tissue repair and remodelling, reestablishment of cellular homeostasis and immune regulation (26).” (Huntingford & Kadri (2014), p. 238)  

105. Martins et al. (2012), p. 32 note that “[p]reference tests that allow animals to display their needs (Dawkins 2004) are used to assess fish welfare.” Also see Näslund & Johnsson (2014), p. 17 for information about indicators derived from preference tests.  

106. “A review of studies of fish price elasticities of demand in western markets concluded that demand responds closely to price (Gallet, 2009). In developing countries, price is likely to be an even greater determinant of whether poor consumers buy fish or not.” (Beveridge et al. (2013), p. 1,075)  

107. Marcus A. Davis from Rethink Priorities contributed substantially to this section. ACE would like to express significant gratitude for his help. 

108. Our use of “humane” does not indicate that ACE condones “humane” farming practices. We use “humane” in order concisely indicate that some slaughter practices seem to involve significantly less suffering than others. 

109. Please note that some of the following descriptions of slaughter practices may be distressing to some readers. 

110. See live chilling on Wikipedia for more information.  

111. According to HSA 2018, p. 1, “[t]he most common methods of slaughtering finfish (e.g. asphyxia in air or hypothermia in ice slurry) are likely to cause considerable distress.” Lines & Spence (2014), p. 258 also note that “[t]he majority of farmed fish are killed without prior stunning simply by asphyxiating them in air or in ice slurry. Ice slurry is operationally convenient because it cools and preserves the fish, while killing them with no additional effort or cost. The method is often referred to as stunning by cold shock but, although cheap and convenient, it cannot be considered to be humane […].” The IBFC (2017) describes one disadvantage of live chilling to be “[s]tress in fish due to steep drop in temperature.” 

112. The RSPCA’s Atlantic Salmon Standards permit applying a percussive blow, electronarcosis followed by bleeding, or electroduction as the stunning/killing methods for marine sourced trout. According to European Commission, (2018), p. 4, “[t]he OIE advises the use of electrical or mechanical (e.g. percussive stunning) methods for killing farmed fish. Other methods, including live chilling with CO₂, CO₂ stunning, chilling in ice water followed by electrical stunning, and asphyxia in ice, do not meet OIE standards.” 

113. “The time to death is temperature and moisture dependent and can take up to several minutes (Robb & Kestin, 2002). Under certain conditions of low temperature and high humidity, delay before death in carp may even be many hours (Oberle personal communication).” (EFSA (2009) p. 9)  

114. From EFSA B (2009), p. 20 

115. “After harvesting, fish are directly transferred to water/ice slurry tubs. Variable water/ice ratios are used to give liquid ice. This easy and quick procedure is used in Mediterranean countries for Mediterranean small sized species and in the U.K. for rainbow trout. Fish body temperature decreases rapidly as does their metabolic rate and movements (live chilling). Fish oxygen requirements also decrease markedly and the time to death can be prolonged. The fish die of anoxia. The length of the stunning period also can be long: trout 28–198 min, salmon 60 min, turbot 20 min, sea bream 20–40 min, sea bass 20 min, eel 12 min (Kestin et al. 1991; Huidobro et al. 2001; Wall 2001; Lambooij et al. 2002; Poli et al. 2002).” (Poli et al. (2005), p. 38)  

116. See the 2017 IBFC report here. 

117. In this context, “stunning” refers to a procedure that “should cause loss of consciousness and sensibility without avoidable stress, discomfort, or pain.” (IBFC (2017), p. 82). 

118. “The OIE advises the use of electrical or mechanical (e.g. percussive stunning) methods for killing farmed fish. Other methods, including live chilling with CO₂, CO₂ stunning, chilling in ice water followed by electrical stunning, and asphyxia in ice, do not meet OIE standards.“ (European Commission (2018), p. 4)  

119. The OIE guidelines state that “[t]here are many species of fish in farming systems and these have different biological characteristics. It is not practicable to develop specific recommendations for each of these species. These OIE recommendations therefore address the welfare of farmed fish at a general level.”  

120. “The OIE guidelines are very common sense, but what they lack is detail. They give general rules, but don’t provide parameters for achieving those goals. If you follow their rules you’ll achieve good welfare, but you need to know a lot about each individual species to meet to those rules.” —Conversation with Phil Brooke of Compassion in World Farming (2018) 

121. See Figures 9–11 of IBFC (2017), p. 69–70. 

122. “After harvest at the pond site, the carp are killed by asphyxia. In restaurants, fresh fish may be killed by a manual blow to the head followed by evisceration or decapitation. Asphyxia is still one of the commonly used methods in commercial practice. Carp in China is not processed in commercial processing plants. Slaughter of carp does not adhere to OIE standards.” (IBFC (2017), p. 79)  

123. See Table 1 of HSA (2018), p. 12. 

124. “The RSPCA standards are the best standards. They are evidence-based and updated annually. A lot of other standards don’t make such changes so frequently. The RSPCA to date has considered welfare in the same way THL would, whereas some other groups allow mutilations and other undesirable things. The RSPCA now has wrasse standards and that was certainly something that was missing. The RSPCA are also improving the transport times and reducing them more and more, although they still need to reduce them even more. The RSPCA also need to improve the handling and issues with starvation. There are always things the RSPCA standards can improve on, but they are as good as it gets right now.” —Conversation with Aaron Ross and Vicky Bond of The Humane League (2018) 

125. “The RSPCA welfare standards for farmed Atlantic salmon (Salmo salar) are used to provide the only RSPCA-approved scheme for the rearing, handling, transport and slaughter/killing of farmed Atlantic salmon. The standards cover the two distinct phases of farming (freshwater and marine farming) […] At present the RSPCA welfare standards for farmed Atlantic salmon only apply to diploid fish. The RSPCA is monitoring the work being done on the specific needs of triploid salmon, such as diet, management and health care, in order to investigate how and whether their welfare needs can be properly satisfied, which will ultimately determine whether they will be allowed to be used in future under the standards.” (RSPCA (2018), pp. 3–36) It seems that the standards don’t cover recirculated aquaculture systems.  

126. The RSPA standards permit an effectively applied percussive blow, electronarcosis followed by bleeding, or electrocution as their stunning/killing methods for marine sourced trout. 

127. From OIE (2015), p. 3 

128. From HSA, p.1 

129. “Electric stunning can be achieved both in water and out of water. Determination of the correct voltage, current or electric field that needs to be applied to stun the fish is dependant on the species, the orientation of the fish and the conductivity of the water. Further details for various species can be found in a range of publications including Lambooij et al. (2006, 2007, 2008), Lines and Kestin (2004, 2005), Robb et al. (2002), Robb & Roth (2003) and Roth et al. (2004).” (Lines & Spence (2012))  

130. See Lambooij et al. (2008). 

131. See Robb et al. (2010). 

132. For example, Lambooij et al. (2008) note that “[n]ile tilapia can be satisfactorily and humanely electro-stunned side-to-side using 1.0 Arms/dm² (50 Hz AC) top to bottom in freshwater. In a head-to-tail application of the electrical field the current can be reduced to 0.4 Arms/dm² (at 50 Hz AC).” However, that latter result appears to be based on only 8 fishes with all 8 being rendered unconscious at those parameter (see section 3.3, p. 92). However, the 95% confidence interval for the proportion of Nile Tilapia rendered unconscious at those parameters, according to the Wilson score interval, based on that sample size and success, would range from 0.68 to 1. Therefore, there could be a 2.5% chance that more than 30% of fishes wouldn’t be appropriately stunned in a head-to-tail application of the electrical field at 0.4 Arms/dm² (at 50 Hz AC). 

133. From IBFC (2017), p. 142 

134. For more information, see Figure 6 from IBFC (2017), p. 79. 

135. From IBFC (2017), p. 51 

136. Figure 3 from Roth et al. (2007), p. 195 suggests that all nine fishes stunned with a hammer force of greater than 78 Newtons, but less than 85 Newtons, were rendered unconscious by the blow. However, the authors note that at 78 Newtons, one mature male was not stunned.  

137. “Big manufactures like Optimar work with institutes to check that their machines work properly. However, a small producer in Eastern Europe might not have that kind of check. One of the things very much needed in the EU is a system for checking that equipment works properly and protocols amongst those using the equipment. Other manufacturers check in the field that animals have no signs of behavioural evidence of consciousness and may check later in the lab further via EEGs.” —Conversation with Phil Brooke of Compassion in World Farming (2018) 

138. According to EFSA (2009) p. 9, the delay before death for carp can take many hours when under conditions of low temperature and high humidity. Poli et al. (2005), p. 38 suggest the following lengths of stunning period for different species: “trout 28–198 min, salmon 60 min, turbot 20 min, sea bream 20–40 min, sea bass 20 min, eel 12 min.” Also see EFSA (2004), pp. 174–175 for further estimates of time to death for common killing methods.  

139. “Unlike percussive stunning, electrical stunning without suitable post-stun treatment is seldom permanent. The duration of unconsciousness depends on the electrical parameters of the stun, the duration of electrical application, and on post-stun treatment. Some species, if stunned for a suitable duration and then placed in still water or ice, will die from hypoxia before they are able to recover consciousness. These include salmon, trout, sea bass, and gilt head bream. However, if these fish are returned to moving or agitated water so that a flow of oxygen to the gills is maintained, they may recover. To minimise the risk of recovery during transport, it is therefore better to transport these fish to the processing factory in ice rather than ice slurry. Species which are tolerant of low oxygen levels require a follow-up killing method, such as immediate bleeding, to avoid recovery. Such species include turbot, halibut, eel, tilapia, carp and catfish. It is necessary to sever all of the gill arches on at least one side of the head, immediately after stunning.” Lines & Spence (2014), p. 260 

140. For example, the OIE (2015), p. 2 code states that “[s]tunning should not take place if killing is likely to be delayed such that the fish will recover or partially recover consciousness.” Similarly, Farm Animal Welfare Committee (2014), p. 4 notes that “[p]rior to killing an animal, either it must be rendered unconscious and insensible to pain instantaneously or unconsciousness must be induced without pain or distress” and that “[a]nimals must not recover consciousness until death ensues.” 

141. Unless otherwise indicated, information in this table is derived from the Summary Table from OIE (2015) and Table 18 from the IBFC (2017) report.  

142. A mis-stun implies that consciousness is not lost immediately.  

143. For example: “In our study the salmon did not lose consciousness using the commercial stunner set at the used bars of <8.1 bar in practice. This event combined with the fact that several of the investigated salmon were conscious, but prohibited from expressing behavioural responses, is of concern.” (Lambooij et al. (2010), p. 111)  

144. “Some other species of farmed fish including catfish, pangasius, carp and tilapia are very resistant to percussive stunning due to the shape and protection of the skull (Marx et al. 1999; Lambooij et al. 2007). Electrical stunning may be a more suitable stunning method for these species.” (Lines & Spence (2012), pp. 157–158)  

145. From EFSA (2004), p. 163 

146. From Lines & Spence (2011), p. 159 

147. “Even so, a potential welfare hazard associated with the electrical stunning was noted during the on-site observation as it appears that the duration of the insensibility following electrical stunning may not always be long enough. For example, in Atlantic salmon, brain death (as defined by the onset of brain dysfunction using Electroencephalography) from bleeding following cutting of the ventral aorta takes up to 7 min at 6°C (Robb et al., 2000). Thus, with the relatively transient effects of the dry electrical stunning observed here, where only 4 to 7 min was required for Arctic char to regain equilibrium, there appears to be a considerable risk that fish may recover consciousness before death occurs from exsanguination.” (Gräns et al. (2015), p. 298)  

148. From EFSA (2009), p. 14 

149. Yue (2008), p. 3 note that a “method to assess consciousness in fish is monitoring the eye roll reflex, movement of the eyes when fish are rolled from side to side. When conscious, fish will attempt to remain upright when rolled to the side and their eyes will roll relative to the head. However, when unconscious, the eyes will remain fixed relative to the head, showing no movement at all. Hence, it is accepted that insensibility is achieved in the absence of the eye roll reflex.” Lines & Spence (2011), p. 159 also state that a “practical field approach is to pick the fish up and assess it for the presence of eye roll, then place it in water and observe it for any rhythmic motion of the lower jaw or opercula, any signs of coordinated motion, swimming or struggling to recover posture. The presence of the first two of these indicators should be considered as signs that the fish is recovering sensibility; the presence of the latter indicates that a fish that has already recovered consciousness (Kestin et al. 2002). Occasional spasmodic convulsions or gasps can be observed in effectively stunned fish and should not be a cause for concern. Since electricity stimulates the muscles directly, all observations need to be made once the electricity has been removed.”  

150. “In a state of unconsciousness the EEG is always abnormal (Lopes da Silva, 1983). Appearance of theta and delta waves and spikes on the electroencephalogram (EEG) towards an isoelectric line (minimal brain activity) indicates cessation of brain activity (Lopes da Silva, 1983).” (Lambooij et al. (2010), p. 107)  

151. Lines & Spence (2011) note that the practical applications of EEG are “seldom possible.” Kestin et al. (2002), p. 303 also state that EEG “is invasive, technically demanding and time consuming, and is thus more suited to laboratory-based studies than field or ship-based studies. Furthermore, it is unsuitable when the method involves a physical blow to the head which might destroy any electrode array (Robb and others 2000). If the effectiveness of slaughter methods is to be assessed in the field, more practicable methods of assessment are needed.” 

152. See IBFC (2017), p. 139 for estimates for Atlantic salmon, common carp, European sea bass, gilthead sea bream, and rainbow trout. 

153. See Section 11.2 of IBFC (2017), p. 139 for the economics of improving fish welfare. For the cost of adhering to improved stunning processes for Atlantic salmon, see Table 46 (p. 147). The cost of complying with OIE slaughter guidelines as a percentage of sales price was estimated between -0.22% and 1.47% for Atlantic salmon, between 2.85% and 28% for common carp, between -2.14% and 7.68% for rainbow trout, and between 0.56% and 1.9% for gilthead sea bream and sea bass. 

154. For example, Asche et al. (2005), p. 29 note that “while most seafood products have several substitutes, the substitutes tend to be other seafood products.” Quass & Requate (2012), pp. 4–5 also report that “[t]he degree of substitutability varies depending on how similar the species are, e.g. white fish or flat fish (Barten and Bettendorf 1989), and whether fish products are fresh, frozen, or otherwise processed (Chiang et al. 2001, Fousekis and Revell 2005).” 

155. For example, in channel catfish “[t]he traditional pond system typically produces 4,500–5,500 kg/ha of catfish with a maximum of 7,000 kg/ha (Brune, 1991; USDA, 2006). However, today, many farms in Alabama produce more than 10,000 kg/ha, and the amount of aeration provided is not adequate to consistently maintain minimum dissolved oxygen (DO) concentrations above 3 mg/L (Boyd and Hanson, 2010).” (Brown (2011), p. 72)  

156. According to Lawson (2013), most aquaculturists agree that the DO range should not drop below 5–6 mg/L, though the optimal range may vary between fish species. Colt (2006) reports the same range for most species, but notes that the suitable minimum for channel catfish, guppy, and eel is only 3.0–3.5 mg/L. MacIntyre et al. (2008) states that some recommendations for rainbow trout culture range from 5–9 mg/L. Both MacIntyre et al. (2008) and Bhatnagar & Devi (2013) report that there is significant disagreement in the industry regarding optimal DO levels. 

157. To the project leader’s knowledge, the FishEthoBase recommendations are the only private standards that provide recommendations for specific DO levels with supporting citations. 

158. The main support for the FishEthoBase recommendation seems to stem from the excerpt of Gilbert (1996), p. 31 which states that “[d]issolved oxygen was around 3.5–4 mg/L and frequently higher. Low oxygen levels in highly-stocked pre growing cages encouraged outbreaks of furunucolosis, but the signs quickly disappeared with an increase in the level of dissolved oxygen, or a reduction in density.” 

159. GAA BAP (2017), p. 10 indicates that their initial minimum recommendation for DO in the effluent water is 4 mg/L, but that compliant farms should aim to reach a minimum of 5 mg/L in the next five years. They note that this recommendation is based on their concern that “[e]ffluents with low dissolved-oxygen concentrations or high pH can negatively affect aquatic organisms in receiving water bodies.”  

160. See the FishEthoBase recommendations for Atlantic salmon for more information.  

161. Armstrong et al. (2003) cite Crisp (1996) when recommending 7.0 mg/L. However, it seems that the relevant data from Crisp (1996) does not in fact relate to Atlantic salmon. Instead, the table where this data is presented in Crisp (1996) (Table 4) draws from studies about rainbow trout and sockeye salmon.  

162. See FW 1.6 from RSPCA’s Atlantic Salmon Standards and FW 1.6 from RSPCA’s Rainbow Trout Standards. In a private communication with a biologist who was contracted by an animal advocacy organization to do research related to farmed fish welfare, we were told that RSPCA’s recommendations are based on the results of several dozen studies. We attempted to contact RSPCA to discuss this, among other questions, but the RSPCA staff member declined due to a scheduling conflict. 

163. “Increased net yields (2,539 lb/acre in 2009 and 6,005 lb/acre in 2010) in high-DO treatments does come at a cost for increased aeration: in 2009, the cost of aeration (electrical cost based on a local utility rate of $0.13/kW‐h) was $0.019/lb of fish produced (high-DO treatment) and $0.004/lb (low-DO treatment). The cost of aeration in 2010 was $0.023 and $0.007/lb in the high-DO and low-DO treatments, respectively. However, we believe that the increased net yields with similar FCRs make intensive aeration a viable management practice on commercial farms.” (Torrans et al. (2015), pp. 485–486)  

164. Figure 22 of the IBFC (2017) indicates that for Atlantic salmon, feed costs make up a significant portion of total production costs. Aeration, on the other hand, is relatively cheap—according to Torrans et al. (2015), pp. 485–486, it costs only a few cents per kilogram of fish. It seems plausible that increased costs from aeration might be counteracted by decreased costs from the resulting improved feeding efficiencies.  

165. Marcus A. Davis from Rethink Priorities contributed substantially to this section. ACE would like to express significant gratitude for his help. 

166. From Näslund & Johnsson (2014), p. 4 

167. See Näslund & Johnsson (2014).  

168. In this context a substrate refers to the lining at the bottom of the enclosure. For example, a sand bottom would be a different substrate than a concrete bottom.  

169. “Barreto et al. (2011) concluded that structural EE was negatively affecting Nile tilapia Oreochromis niloticus (Cichlidae) as it increased aggression, but in this study only a single, low level of complexity was investigated. Possibly, more structures could have reduced aggression.” (Näslund & Johnsson (2014), p. 9) 

170. See the FishEtho Base standards for Atlantic salmon, nile tilapia, and gilt-head bream. 

171. From Näslund & Johnsson (2014) and Williams et al. (2009) 

172. Some further resources on environmental enrichment include:
     – Loch Duart Salmon (2016)
     – Batzina & Karakatsouli (2012)
     – Mendiguchía et al. (2006)
     – Carroll et al. (2003) 

173. “Two recent reviews have highlighted the potential for adverse effects of aquaculture on wild salmon populations. Naylor et al. (2005) reported on the ecological, genetic, and socio-economic effects of escaped farmed salmon in the Atlantic and Pacific Oceans, and concluded that risks to wild populations, ecosystems, and society were high where salmon were farmed in their native range, when large numbers of salmon were farmed near small natural populations, and when exotic pathogens were introduced with farmed fish. Ferguson et al. (in press) focused on the genetic effects of farmed Atlantic salmon on wild salmon, and concluded that escaped farmed salmon have both indirect and direct genetic effects on wild populations, resulting in a loss of fitness (reduced recruitment) in wild populations, which are cumulative over generations in the event of continued escapes. Both reviews provided recommendations for a more sustainable coexistence of fish farming and wild salmon populations.” (Hindar et al. (2006))  

174. “The use of sterile Atlantic salmon (Salmo salar) triploids has been considered a possible strategy to reduce potential problems of interbreeding between escaped farmed and wild salmon (Heggberget et al., 1993; McGinnity et al., 1997; Cotter et al., 2000; Lacroix and Stokesbury, 2004).” (Taylor et al. (2011))  

175. Both Lerfall et al. (2017) and Stevenson (2007) report that triploids have higher mortality and poorer health that diploids. Lerfall et al. (2017) specifically note that triploids struggle compared to diploids when exposed to higher temperatures (15°C and 18°C), and that they are more sensitive to hypoxia than diploids.  

176. “CIWF and WSPA believe that biotechnology techniques involving chromosome manipulation (e.g. sex reversal and triploidy) should be prohibited.” (Stevenson (2007), p. 11)  

177. “Studies have indicated that triploid Atlantic salmon often show lower survival in freshwater (McGeachy et al., 1995), poorer growth in freshwater (McGeachy et al., 1995; O’Flynn et al., 1997) and saltwater (Friars et al., 2001), although improved growth performance in saltwater has been reported (Oppedal et al., 2003) and increased deformity prevalence (O’Flynn et al., 1997; Sadler et al., 2001) when compared with diploids.” (Taylor et al. (2011))  

178. See Fraser et al. (2012). 

179. “Two recent reviews have highlighted the potential for adverse effects of aquaculture on wild salmon populations. Naylor et al. (2005) reported on the ecological, genetic, and socio-economic effects of escaped farmed salmon in the Atlantic and Pacific Oceans, and concluded that risks to wild populations, ecosystems, and society were high where salmon were farmed in their native range, when large numbers of salmon were farmed near small natural populations, and when exotic pathogens were introduced with farmed fish. Ferguson et al. (in press) focused on the genetic effects of farmed Atlantic salmon on wild salmon, and concluded that escaped farmed salmon have both indirect and direct genetic effects on wild populations, resulting in a loss of fitness (reduced recruitment) in wild populations, which are cumulative over generations in the event of continued escapes. Both reviews provided recommendations for a more sustainable coexistence of fish farming and wild salmon populations.” (Hindar et al. (2006))  

180. See this Fishcount table for more details about the calculations, and Fishcount (2002) for more information about the significant welfare problems associated with wild-caught fishes. 

181. From Table 6 of FAO (2009). 

182. “The second ask is to get the industry to move towards the highest percentage plant-based feed that they can, which prevents the death of other fish being fed to the salmon.” —Conversation with an Anonymous Animal Welfare Specialist from a Major Animal Advocacy Organization (2018) 

183. “CIWF also thinks farmers should be choosing species which naturally live in ponds and believes in rearing vegetarian and omnivorous species rather than carnivorous ones.” —Conversation with Phil Brooke of Compassion in World Farming (2018) 

184. “The protein requirement, which is primarily used for tissue growth, varies little among farmed species. Even so, the origins of proteins in animal feeds vary widely (see Fig. S3), with marine piscivorous species such as European seabass and gilthead seabream most dependent on fishmeal for protein (>40% inclusion). Fishmeal also represents a significant (20%–55%, depending on life history stage), but declining, contribution to total protein intake in rainbow trout and Atlantic salmon. In contrast, cereal and oilseed proteins play a much larger role in the diets of omnivorous carp and tilapia species and terrestrial animals than in the diets of piscivorous fish. The proportion of fishmeal and alternative protein ingredients in aquafeeds depends on the nutritional requirements of the species, relative commodity prices, and the regulatory environment of production systems (e.g., Europe prohibits the use of rendered animal products in feeds).” (Naylor et al. (2009))  

185. See Delon & Purves (2017). 

186. For more information, see our Conversation with Matt Ball of One Step For Animals (2017). 

187. See Ray (2018), Tomasik (2015) A, and Tomasik (2015) B. 

188. “Numerous studies have been carried out to examine the effect of replacing FMs and FOs with alternative ingredients in fish feeds. The most frequent finding is that partial replacement of FMs and FOs is possible without compromising growth, but complete replacement is usually not successful due to problems related to dietary palatability, the presence of anti-nutritional factors in some alternative ingredients and difficulties with maintaining dietary amino acid balance, adequate amounts of essential or semi-essential micronutrients and/or adequate provision of essential n-3 fatty acids (Francis et al., 2001; Opstvedt et al., 2003; Mundheim et al., 2004; Torstensen et al., 2005, 2008; Aksnes et al., 2006; Espe et al., 2006, 2007; Gatlin et al., 2007; Hevrøy et al., 2008; Miller et al., 2008; Turchini et al., 2009, 2011; Crampton et al., 2010; Pratoomyot et al., 2010).” (Bendisksen et al. (2011), p. 133)  

189. Béné et al. (2015) note that “[f]ormulated feeds are a significant factor in production costs, and this is a strong incentive to develop technology that will make feeds more affordable and sustainable.” Therefore, if alternative feeds lowered production costs, it’s likely that the industry would have already adopted them.  

190. One potential method of reducing fish stress before slaughter is the use of anaesthesia (Zahl et al. (2012), p. 251). There is some question as to whether fishes exposed to anaesthetics directly in their water have aversive reactions, and questions about the variable responses between species (Readman et al. (2013)). Such anaesthetic agents are approved for use in Australia, Chile, New Zealand, Korea, Costa Rica, and Honduras, but not in the E.U. (Lines & Spence (2014), p. 260). 

191. “Recent research shows that above 22 kg/m³, increasing density is associated with lower welfare for caged Atlantic salmon. However, in order to provide a safety margin, CIWF and WSPA believe that the maximum stocking density for Atlantic salmon in sea cages should ideally be 10 kg/m^3, with farmers who achieve a high welfare status and in particular low levels of injuries, disease, parasitic attack and mortality being permitted to stock up to a maximum of 15 kg/m³.” (Stevenson (2007), p. 7)  

192. In general, when interventions cause changes in production costs, they can in turn affect demand and in some instances lead to consumers substituting for other farmed fish products. The overall effects of such interventions on farmed fishes is complicated and the possible outcomes seem very hard to quantify. 

193. Individual email addresses can be found on ACE’s Meet Our Team page.  

194. Much is unknown about which exact stunning and slaughter procedures would be best for different species, including several commonly farmed species of carp. Research is lacking for most of the top 10 farmed finfish by volume including: grass carp (most produced by weight), silver carp, bighead carp, and catla. HSA (2018) supports scientific investigation into the humane slaughter of such high-volume species though they caution that some of these species have a low cost per kilogram which creates a low margin for increasing production costs via improved slaughter practices. HSA also supports scientific research into species that are high cost per kilogram (like tunas), and species which are in the same order as other species that have been studied for humane slaughter.  

Key Resources

European Food Safety Authority (EFSA). (2009). General approach to fish welfare and to the concept of sentience in fish. EFSA Journal, 7(2), 954.

Martins, C. I., Galhardo, L., Noble, C., Damsgård, B., Spedicato, M. T., Zupa, W., … Planellas, S. R. (2012). Behavioural indicators of welfare in farmed fish. Fish Physiology and Biochemistry, 38(1), 17–41.

Ellis, T., Yildiz, H. Y., López-Olmeda, J., Spedicato, M. T., Tort, L., Øverli, Ø., & Martins, C. I. (2012). Cortisol and finfish welfare. Fish physiology and biochemistry, 38(1), 163–188.

Branson, E. J. (Ed.). (2008). Fish welfare. John Wiley & Sons.

IBFC. 2017. Welfare of farmed fish: common practices during transport and at slaughter. Final Report. IBF Consortium: IBF Consumer Policy Centre, VetEffecT and Wageningen University & Research Centre (WUR). Study prepared for the European Commission Directorate-General for Health and Food Safety SANTE/2016/G2/009, September 2017. European Union Strategy for the Protection and Welfare of Animals 2012–2015. 186 pp. doi: 10.2875/172078

Näslund, J., & Johnsson, J. I. (2016). Environmental enrichment for fish in captive environments: effects of physical structures and substrates. Fish and Fisheries, 17(1), 1–30.

Fair Fish’s Fish Ethology Database

 

Conversation Summaries:

Conversation with Professor Victoria Braithwaite of Penn State University

Conversation with Doctor Lynne Sneddon of University of Liverpool

Conversation with Aaron Ross and Vicky Bond of The Humane League

Conversation with Phil Brooke of Compassion in World Farming

Conversation with an Anonymous Animal Welfare Specialist From a Major Animal Advocacy Organization

Check the appendix

Appendix

Further Information About Farmed Fish Production¹

There were significant variations in production tonnage between 2010 and 2015 for the most produced species. Table 3 provides a brief overview of these variations.

Table 3: Changes in FAO-reported production tonnage from 2010–2015

Please note that the estimates reported in Table 1 and the overall global estimates don’t account for several important factors. As a result, they likely significantly underestimate the total numbers of farmed fishes. For instance, farmed fishes that have relatively high pre-slaughter mortality rates aren’t accounted for in Table 1.² The reported numbers also don’t account for the fact that different farmed fishes can have significantly different lifespans before their harvest. This means that the reported slaughter totals may be substantially different than the total number of farmed fishes alive at any point in time for a given species. See Table 4 for rough estimates of survival rate, lifetime before slaughter, weight at slaughter, and slaughter weight per year of life for some of the main farmed fish species.

Table 4: Rough estimates of survival rate, slaughter weight, lifetime before slaughter, and slaughter weight per year of life for some of the most numerous fish species³

The numbers reported in Table 1 and the global estimate of the number of farmed fishes may also not take into account fishes that are used for restocking.^{14, 15} The numbers also doesn’t account for the number of fishes farmed as baitfish.¹⁶ Based on those factors, the actual number of farmed fishes alive annually could be several hundred billion.¹⁷

Notes on our Literature Search

Before completing this report, ACE already knew of multiple studies pertaining to farmed fish welfare. This knowledge came from the general research we had completed over the past few years, our involvement and communication with the effective animal advocacy community, and our interaction with popular media reports. In early/mid 2018, we used Google Scholar to further our knowledge and became aware of some other evidence and research. This moderate-depth literature search included searching for literature containing terms such as “fish welfare indicator systematic review,” “recommended dissolved oxygen carp,” and “recommended stocking density aquaculture.”¹⁸ The project leader also spoke with or exchanged emails with multiple informed sources in the field. These conversations led us to identify further relevant literature and resources. Contact our research team for a full list of the resources identified in the literature search, as well as notes from their reading.

We did not commit to a method for analyzing the literature before the project leader reviewed the literature search results. Below, some of our general methods are summarized, along with a brief rationale:

• If the project leader judged the literature as relevant,¹⁹ they read as much of it as they could in the available time and noted key quotes, tables, graphs, and occasional brief key impressions.
• The analysis generally prioritized benchmark programs, information that could potentially inform animal advocates, and/or research that academics indicated as being the most relevant. As a result, dissolved oxygen, stocking density, and slaughter considerations were generally prioritized.

We would like to note the contributions of Marcus A. Davis²⁰ and Matt Edwards²¹ to the drafting of this page.²²

---

1. Matt Edwards contributed substantially to this section. ACE would like to express significant gratitude for his help. 

2. For example, ACE estimates that for the four most consumed farmed finfish in the U.S. the mortality rates prior to slaughter are: 18%–60% for salmon, 5%–35% for tilapia, 10%–38% for pangasius, and 12%–65% for catfish (as 90% subjective confidence intervals). 

3. Unless otherwise indicated, the data for this table was found among these FAO pages in early 2018.  

4. Unless otherwise indicated, these are the estimated weights from Mood & Brooke (2012). 

5. Catla is a type of commonly farmed Carp.  

6. This value was estimated in ACE’s Guesstimate Model of U.S. per Capita Consumption of Farmed Finfish.  

7. These values were estimated in ACE’s Guesstimate Model of U.S. per Capita Consumption of Farmed Finfish. 

8. These values were estimated in ACE’s Guesstimate Model of U.S. per Capita Consumption of Farmed Finfish. 

9. These values were estimated in ACE’s Guesstimate Model of U.S. per Capita Consumption of Farmed Finfish. 

10. These values were estimated in ACE’s Guesstimate Model of U.S. per Capita Consumption of Farmed Finfish. 

11. These values were estimated in ACE’s Guesstimate Model of U.S. per Capita Consumption of Farmed Finfish. 

12. Roho Labeo is a type of commonly farmed Carp.  

13. These values are based on the reported time spans from the FAO fact sheet and the reported on-growing times from Lymbery (2001), p. 19.  

14. Restocking, or fish stocking, is the practice of raising fishes in a hatchery and releasing them into the wild to supplement existing populations or to create a population where none exists.  

15. For instance, Table 2 of the 2013 USDA Census of Aquaculture reports the number of farmed fishes sold and appears to exclude those “produced and distributed for conservation, recreation, enhancement, or restoration purposes.” See Table 22 and note that the number of Walleye distributed for restocking is greater than the total number of other fishes reported as sold in Table 2. Based on the data from Table 3, we estimate that at least 3.63 billion fishes were distributed for these purposes in the United States in 2013 alone. This number should be considered a lower bound, since (i) the numbers for some species are not reported, and (ii) estimates exclude pre-release mortalities.  

16. Table 4 of the 2013 USDA Census of Aquaculture suggests that at least 1.16 billion baitfishes were sold in 2013 in the United States alone, excluding goldfish and fishes categorized as “other baitfish.” Note that this estimate excludes fishes that die before they’re sold. This number could be substantial as high mortality rates are common in the aquaculture industry—Mischke (2012), p. 223 suggests that baitfishes may have a mortality rate of ~25%. It is unclear what age the baitfishes are usually sold at; Ohs et al. (2013) suggest that baitfishes are alive for several months before being sold, while Gunderson (2018) suggests that golden shiners (one of the most popular farmed baitfish) are sold at 1 or 1.5 years of age.  

17. Some weak evidence for this can be found in Leung et al. (2007). They report that in 2004, “711.6 billion freshwater fish fry and 2.41 billion marine/brackish fish fry were produced in China.” It is unclear what their estimate is based on, as they provide neither citations nor rational behind their calculations.  

18. This footnote provides further details about the literature search. Not every aspect of the search is recorded here but it should give readers a general idea of the process. The following made up the bulk of the literature search in early 2018 over Google Scholar (search terms in quotations, restrictions listed afterwards, then the number of results whose title was read before the project leader judged their relevance):
    – “fish welfare indicator systematic review”
    – “measuring welfare of fish,” year of publication restricted to 2017 or 2018, first 200 results inspected
    – “measuring welfare of fish,” year of publication restricted to 2016 only, first 200 results inspected
    – “measuring welfare of fish,” year of publication restricted to 2015, first 200 results inspected
    – “measuring welfare of fish,” publication date unrestricted, first 200 results inspected
    – “finfish welfare measure,” publication date unrestricted, first 200 results inspected
    – “finfish welfare,” publication date unrestricted, first 200 results inspected
    – “farmed fish slaughter,” publication date unrestricted to 2005 or after, first 200 results inspected
    – “environmental enrichment farmed fish,” publication date unrestricted, first 80 results inspected
    – “recommended water quality farmed catfish,” publication date restricted to 2006 or after, first 140 results inspected
    – “recommended dissolved oxygen catfish,” publication date restricted to 2006 or after, first 180 results inspected
    – “recommended dissolved oxygen carp,” publication date restricted to 2006 or after, first 160 results inspected
    – “recommended dissolved oxygen aquaculture,” publication date restricted to 2006 or after, first 160 results inspected
    – “recommended stocking density aquaculture,” publication date restricted to 2006 or after, first 160 results inspected
    – “recommended stocking density carp,” publication date restricted to 2006 or after, first 160 results inspected
    – “growth efficiency fish feed substitute salmon,” publication date restricted to 2006 or after, first 100 results inspected
    – “substitution elasticity fish europe,” publication date restricted to 2006 or after, first 80 results inspected
    In addition, for some of what seemed to be the most influential papers in the field, the project leader inspected all papers listed by Google Scholar’s “cited by” function. These influential papers (of which all were inspected unless otherwise indicated) include:
    – Martins et al. (2012)
    – Ellis et al. (2012)
    – Huntingford et al. (2006), first 200 results inspected
    – Stien et al. (2013) 

19. The project leader assessed eligibility by reading relevant sections of the literature in question. This eligibility assessment wasn’t checked by other reviewers of this piece. The inclusion/exclusion criteria were not clearly prespecified and were instead left to the judgment of the project leader. The project leader made a judgement call about when to stop for any given search term based on when they felt it was sufficiently unlikely that anything useful would be found. 

20. Davis is the Senior Researcher at Rethink Priorities.  

21. Edwards volunteered on this project and completed the section in the Appendix titled Further Information About Farmed Fish Production. Edwards is not formally affiliated with any organization. 

22. These contributions substantially contributed to the quality of the piece, which is greatly appreciated. In addition, this report was externally reviewed by Vicky Bond, Walter Sanchez-Saurez, Lewis Bollard, David Moss, and Douglas Waley. ACE would also like to express sincere gratitude for their contributions. Any mistakes remaining are our own.