The most repeated consideration, in most guiding documents, is to maintain the welfare of the study animals and minimise or mitigate pain, suffering and distress (AFS 2014; CCAC 2005). This may be achieved through less invasive techniques and refining procedures in ways that reduce the intensity and durations of exposure to pain, suffering and distress (Mohr 2013; NRC 2011).

The number of individuals captured, manipulated or killed for the purpose of research is a fundamental consideration of any scientific study involving live animals. The appropriate sample size is contingent on the nature and objectives of the proposed research. Most guiding documents therefore advocate the principle of the three R’s (Russell & Burch 1959). For experimental research, particularly field research in which animal subjects cannot be replaced by computer or other surrogate models, the number of study animals should be minimised as far as practicably possible, but sufficient to provide representative information (UFRC 2014).

A key principle of justifying animal experimentation is that the cost incurred to the individual or population (e.g. stress, deaths) is outweighed by the benefit of the research (e.g. advancement of knowledge) to society (Knight 2011). Thus, in proposed studies for collecting or manipulating fishes, a cost–benefit analysis should underlie decisions regarding sample size (Animal Behaviour Editorial 2012). For biodiversity studies, a sample should be sufficient to capture phenotypic or genotypic variation, although this may be difficult to determine without prior knowledge of a taxon’s natural genetic variation (Miyamoto et al. 2008). In experimental biology, a minimum sample size may be determined to satisfy the required statistical power (Knight 2011). Voucher specimen collections should sacrifice the minimum number of animals necessary for effective species confirmation, taking into account sexual dimorphism, ontogenetic change in identifiable features and similarity to related species (Animal Research Review Panel n.d.).

During field collection, vulnerable life-history stages or life-cycle events, such as aggregations of breeding fish, and sensitive habitats should be avoided (AFS 2014). Species regarded as imperilled or threatened should not be subjected to invasive sampling studies, unless deemed necessary for the specific purpose of gaining information for improved conservation (AFS 2014). By-catch of non-target species should be minimised, but if captured these should be released promptly and with minimal injury (AFS 2014). The number of sampling sites, depending on habitat variation, distribution range and the purpose of collection (e.g. confirming range shifts), should be minimised. Investigators should also select a sampling method that has the least impact on the fishes and the local ecosystem (CCAC 2005).

Recent debate around the conservation ethics of killing endangered species for scientific research highlights the ethical challenge of justifying the lethal sampling of fishes in field work (Heupel & Simpfendorfer 2010; Minteer et al. 2014). For lethal sampling, particularly of threatened species, the proportion of the population affected by sampling (with implications for the viability of the sampled population) and the purpose of collecting, as it applies to the conservation management of the species, should be considered.

Certain fisheries journals, such as Journal of Fish Biology and Indian Journal of Fisheries, have explicit policies on the sizes and justification of lethal collections reported on in submitted manuscripts and will not accept manuscripts that involve the unmotivated killing or damage of IUCN Red List threatened or endangered species.

High levels of natural mortalities and rapid recoveries from population reductions are characteristic events in the early life histories of many fish species (UFRC 2014). It is therefore generally regarded by most ethical standards that the sampling of fish larvae that have not developed beyond exclusive reliance on their own yolk nutrients does not require ethical clearance (Canadian Council on Animal Care 2007). However, according to the SABS (2008), live, sentient non-human vertebrates, including eggs, foetuses and embryos of fish, are to be cared for and used in ways that are judged to be scientifically, technically and humanely appropriate. Furthermore, investigations involving the early life stages of fishes typically require the sacrifice of large numbers of study organisms. Therefore, a precautionary approach should be followed, in which sample sizes and by-catch are minimised, particularly when threatened species constitute the target or possible by-catch species (CCAC 2005).

For fish larvae, killing by immersion in ice or a fixative such as ethanol or formalin is generally acceptable; however, this should be preceded by an overdose of an appropriate anaesthetic (Craig 2006). Any proposed non-use of anaesthesia on the grounds of compromising the aim of the study should be fully justified a priori (Metcalfe & Craig 2011). When fish larvae become independently feeding forms, ethical authorisation must be obtained (CCAC 2007).

Visual sampling, such as underwater visual census (by snorkelling or SCUBA) and underwater video surveys, is common in marine, estuarine and freshwater environments (Ellender et al. 2012; Willis, Millar & Babcock 2000) and provides a valuable tool for collecting qualitative and quantitative fishery-independent data on fish density, diversity, community structure and behaviour (Brock 1982). Underwater videography is increasingly used for assessing and monitoring fish communities, as digital video provides standardised methodology and permanent records (Langlois et al. 2010). The non-destructive nature of these methods makes them suitable for sensitive habitats and protected areas (Mallet & Pelletier 2014), and in most cases, ethical justification for their use is not necessary. In some instances, however, divers conducting visual or video observations or the deployment of remote equipment may cause habitat damage. Investigators should thus attempt to minimise any potential impact on the habitat.

Numerous sampling techniques involve the actual capture (such as trapping, netting or hooking) of target fishes and are thus inherently invasive. Investigators therefore need to understand the uses, limitations and threats of these methods, in order that the most appropriate method be employed, to provide the most satisfactory result, while minimising pain, suffering and distress (EU 2010).

Netting: There are a variety of netting methods for the capture of fishes, some passive and some active. Gill nets are passive nets, hung vertically in the water column. The gills or fins of fish swimming into the net become entangled in the mesh, preventing their escape. Mesh size is an important consideration, as larger mesh sizes usually result in the capture of larger fish, while providing some form of exclusion for smaller species or individuals. The use of gill nets for sampling is generally accepted, although gill nets can result in cuts on the body from the filament of the net or the drowning of obligate air-breathing fishes, amphibians, reptiles and mammals (Ellender et al. 2016).

Fyke nets are also passive nets, comprising a vertical wall of net that guides fishes into an enclosed cod-end or bag at either end. They are generally considered non-destructive, as fish are free to swim in the cod-end once captured, allowing them to be released relatively unharmed. However, fyke net placement is an important consideration, as nets set in areas of low oxygen concentration can result in elevated fish mortality. Furthermore, fyke nets are not species selective and mortality of by-catch, particularly drowning of obligate air-breathing animals, can occur (Larocque, Cooke & Blouin-Demers 2012). Mortality can also result from predatory fishes, frogs or crabs that are caught in the cod-end or from external sources, such as otters attacking fishes through the net (Ellender et al. 2016). Given the associated predation risk to captured fishes, fyke nets may be unsuitable in certain environments, particularly if they pose a high capture threat to endangered species. Simple gear modifications, such as creating air spaces using floats to allow captured animals to breathe (Larocque et al. 2012) or selective mesh placed over the entrance to exclude mammals or reptiles, can however reduce mortality and by-catch. Reducing soak time or checking nets at shorter intervals enables by-catch and target species to be processed and released more rapidly. Owing to the potential injuries, by-catch and mortalities associated with gill and fyke netting, motivation for their use is warranted and identification of potential by-catch species and an assessment of their conservation status should form an integral part of the planning phase of any project using these gears.

Active net gears include, inter alia, seine nets and various forms of trawl nets. Seine nets are vertical nets, with a floating top line and weighted bottom line, actively hauled to encircle a set area of water and pull the fish within that area to shore. They are a preferred gear for sampling fish communities in shallow littoral environments such as beaches, estuaries and artificial impoundments (Lapointe, Corkum & Mandrak 2006). A cod-end at the centre of the net acts as a corral point for fish as they are brought to shore. Seine netting is generally regarded as non-destructive, although mortality of small-bodied species and post-release mortality of discarded by-catch species have been recorded (Kennelly & Gray 2000). Fish mortality can be minimised by handling fish within the submerged cod-end before bringing the net onshore. In some cases, seine netting can damage submerged macrophyte habitat in haul regions, although the impact is usually minor (Bayley & Herendeen 2000).

Trawling refers to nets towed from a boat, which usually sample hyper-benthic environments. The use of dragged or towed gears, which may damage the habitat, benthic environment or animals sampled (Feyrer et al. 2013), should be avoided or minimised (UFRC 2014). Alternative methods should also be sought where possible; for example, Feyrer et al. (2013) investigated the use of attaching a camera to the open cod-end of a trawl net, allowing fish to be recorded but to pass through the net unharmed, greatly reducing mortality.

Electrofishing: Electrofishing is one of the most effective and commonly employed sampling methods in shallow freshwater environments (Snyder 2003). While generally considered non-destructive, it can have negative impacts (Snyder 2003). In extreme cases, severe spinal injuries, internal haemorrhages, bleeding at the gills, physiological stress, asphyxiation and harmful effects on embryos have been recorded (Snyder 2003). It is therefore recommended that researchers familiarise themselves with, and apply, recommendations for international best practises regarding electrofishing (e.g. Beaumont et al. 2002; Goodchild 1991). Current strengths and settings and the electrofishing gear itself should be suited to the target species and conditions in the study area, such as conductivity, water depth and survey area (Bohlin et al. 1989). When sampling threatened or endangered fishes, electrofishing should be used with caution and alternative methods employed where possible (Ellender et al. 2012). If multiple electrofishing passes are intended, all fish captured should immediately be removed from the net and placed in aerated buckets, filled with ambient water, to avoid prolonged exposure. Bohlin et al. (1989) and Snyder (2003) provide comprehensive reviews on standardising electrofishing protocols and harmful effects of electrofishing.

Hook-and-line fishing: Hook-and-line fishing is commonly used in a range of research applications, such as quantifying abundance, assessing community structure and for tagging studies (Bennett & Attwood 1993; Dunlop, Mann & Van der Elst 2013; Willis et al. 2000). Catch-and-release is often employed, whereby fishes are released to the capture environment after undergoing procedures such as measuring or tagging, with minimal impact to the fishes and environment (Cooke et al. 2013a). However, excessive fight times or post-capture air exposure and high water temperatures have been shown to induce physiological stress responses in hook-and-line caught fishes, with implications for their successful release (Cooke & Suski 2004). In addition, fish captured by hook-and-line fishing may be subjected to stress as a result of barotrauma injuries if rapidly brought to the surface (Keniry et al. 1996), or damage to the gills or viscera through hook ingestion (Domeier, Dewar & Nasby-Lucas 2003). The potential impacts of hook-and-line fishing should thus be thoroughly considered before using this method to capture endangered species or in excessive water depths where barotrauma is likely.

Relevant mitigation measures, such as choice of hook design and size, optimal fishing times and minimised handling times need to be considered (see Cooke & Suski 2004). The use of circle hooks has been shown to decrease the incidence of gut- or deep-hooking (Cooke & Suski 2004) and post-release mortality (Prince et al. 2007) and the use of barbless hooks can reduce injury and facilitate hook removal (Schaeffer & Hoffman 2002). Brownscombe et al. (2016) provide guidelines for best practice in catch-and-release recreational fisheries.

Ichthyocides (poisons) and anaesthetics have been used extensively to determine fish species composition, particularly in freshwater and shallow reef environments (Ackerman & Bellwood 2000; Willis 2001). Rotenone, for example, is a commonly used ichthyocide, highly effective at killing most fish species (AFS 2014; Willis 2001). Although such techniques are effective for assessing cryptic species and provide greater species and family counts than non-destructive methods (Willis 2001), they can cause excessive mortality for sensitive species, are non-selective, vary widely in effectiveness, provide inefficient sampling of highly mobile species and cannot be used in sensitive areas or where sampling is to be repeated (Brock 1982). Furthermore, recent studies have highlighted that many non-target organisms are susceptible to ichthyocide concentrations used for fishes (e.g. Dalu et al. 2016). Careful consideration, especially of collateral impacts on non-target biota, is therefore necessary, prior to using ichthyocides. Clearwater, Hickey and Martin (2008) discuss several ichthyocides for pest control; however, a comprehensive review of ichthyocides and their use in fish research is lacking.

Sedatives or anaesthetics are commonly prescribed and used during prolonged or invasive procedures, to immobilise fish and reduce pain and physiological stress related to capture, handling and processing, through depression of the nervous system (CCAC 2005; UFRC 2014). Minor procedures usually require mild sedation, whereas intrusive (major) procedures, such as laparotomy, usually require deep anaesthesia (Thorsteinsson 2002). Where anaesthetics or sedatives are used, only approved drugs should be used and administered at a concentration and duration appropriate to the size and species of fish and in accordance with established guidelines and relevant legislation (AFS 2014; UFRC 2014).

The SANS 10386 (SABS 2008) recommends four anaesthetics for use on fishes in South Africa, including tricaine methanesulfonate (MS-222), benzocaine (and benzocaine hydrochloride), metomidate and ketamine hydrochloride. However, none of these is registered in Australia, New Zealand, the United States, Canada or Europe (Javahery & Moradlu 2012). Other anaesthetics common in fish research include 2-phenoxyethanol, clove oil, AQUI-S^TM (active ingredient isoeugenol) and quinaldine sulphate (Javahery & Moradlu 2012; Wagner & Cooke 2005). However, many of these are regarded as hazardous (Javahery & Moradlu 2012), particularly if fish are released and subsequently consumed by humans (Thorsteinsson 2002). Of these anaesthetics, only MS-222 is approved by the United States Food and Drug Administration (FDA) for use on animals in the United States (AFS 2014), but has a 21-day withdrawal period before fish can be released into the wild, negating its use in field studies. Eugenol (a derivative of clove oil) and carbon dioxide are the only immediate-release drugs approved by the FDA (AFS 2014) – yet neither is recommended for use in South Africa (SABS 2008), and carbon dioxide allows considerable trauma at shallow sedation and is not suitable for heavy sedation required in invasive procedures (Javahery & Moradlu 2012). Therefore, identifying an appropriate drug and concentration, particularly for fish intended for immediate release, is often a challenging task and should be based on thorough research of contemporary literature and legislation prior to any implementation (see Thorsteinsson 2002 for comprehensive review of anaesthetics).

Anaesthetics can induce prolonged negative metabolic effects and stress responses in fishes (CCAC 2005), as well as hypoxia from reduced respiration and vascular activity (Thorsteinsson 2002). In certain cases, this may lead to post-release mortality or increased predation risk (Cooke et al. 2005). Numerous studies on fishes, involving handling and even invasive procedures such as laparotomy, have deemed the use of anaesthesia inappropriate (Wagner & Cooke 2005). For large elasmobranchs, anaesthesia would require specialised lifting equipment, considerable risk to the animals and excessive quantities of chemicals; therefore, such studies usually rely on tonic immobility (e.g. Holland et al. 1999). Therefore, the use of anaesthetics should be evaluated prior to any study (Thorsteinsson 2002), particularly when fish are planned for release into the wild and in some cases may not be appropriate at all (Cooke et al. 2013b, 2016).

Tissue collection: A variety of methods have been used to successfully collect tissue or blood samples from anaesthetised fish without significant effects on their survival. Non-lethal tissue sampling includes fin-clipping, excision of scales and gill, muscle or skin biopsies (McCormick 1993; UFRC 2014). Fin-clipping is common for genetic sampling and its deleterious effects have been extensively studied. Many studies reported no negative effects and complete regeneration of the affected fins in weeks (Woodall, Koldewey & Shaw 2011), while others reported some decrease in survival or growth of fin-clipped fish (O’Grady 1984). The appropriate site to sample is species specific. For example, partial clipping of the dorsal fin was successfully used for genetic sampling of seahorses with no significant effect on mortality (Woodall et al. 2011).

Blood collection: Fish blood is collected for various reasons, such as haematology and clinical chemistry tests (Satheeshkumar et al. 2012), parasitological investigations (Shahi et al. 2013) and investigations on adaptations to subzero temperatures (Miya et al. 2014). Such studies have demonstrated that blood samples can be successfully obtained, even from fish less than 100 g in mass, without compromising their survival. Plastic syringes with a small volume of anticoagulant such as sodium or ammonium heparin or sodium citrate can be used to avoid the rapid clotting that can occur with the use of glass syringes (Stoskopf 1993). The size of the syringe and diameter of needle should be appropriate to the size of fish. The preferred site for blood sampling from anaesthetised fish is the caudal vein running beneath the vertebrae, using a lateral or ventral approach. Study objectives will determine the required blood volume and concentration of anticoagulant. Other methods used to collect blood from fishes include cardiac puncture and caudal bleeding (UFRC 2014). These methods, however, can compromise the survival of the fish. Needles are most effective when new and sharp and should be replaced or sterilised after each sample to avoid contamination and unnecessary injuries to the fish.

Marking (using dyes or fin-clipping) and tagging of fishes with external or internal tags has been used extensively to estimate population sizes, discriminate stocks, identify migration rates and movement patterns and assess fish growth and survival (McFarlane, Wydoski & Prince 1990). Numerous marking and tagging methods are used, although most are intrusive to some degree (Thorsteinsson 2002). The capture, handling and marking/tagging process can affect the behaviour, physiology, swimming capacity, predator avoidance, feeding, growth rate and ultimate survival of tagged organisms (Naef-Daenzer et al. 2005; Thorsteinsson 2002). Studies should therefore aim to minimise these impacts, for the welfare of the study animal, to ensure that the process provides scientifically valuable data, and to maximise the value from every individual that has been subjected to the stress of capture, handling and processing (Cooke et al. 2013b).

Marking: Marking techniques for fishes, such as external dyes, fin-clipping, hot branding (e.g. lasers) and freeze branding (e.g. liquid nitrogen), do not breach the skin or musculature and are thus generally considered to have minimal impact on fishes (Murray & Fuller 2000). However, such marks usually exhibit short-term retention, and brightly coloured external marks or dyes (and any external part of a tag) should be avoided, as these can affect social structure or interactions of the fish or reduce its predator avoidance or hunting capabilities (Guy, Blankenship & Nielsen 1996; Murray & Fuller 2000). Chemical tags are another form of marking and may be natural (i.e. accumulation from the local environment) or induced, through immersion, injection or ingestion of chemicals, which become incorporated into tissue or calcified structures (Nielsen 1992). Chemical tagging is long lasting and allows large numbers of individuals to be tagged and does not necessarily require anaesthetic or even handling (Guy et al. 1996), but often requires sacrifice of the animal at a later stage.

General tagging considerations: To minimise the effects of tags on fish and maximise data collection and tag retention, tag size and weight should be appropriate for the size, weight and species of fish (Cooke et al. 2011; Stasko & Pincock 1977). The tag placement and attachment method must also be appropriate for the study species (Naef-Daenzer et al. 2005; SABS 2008) and have minimal influence on the posture, buoyancy, locomotion (Stasko & Pincock 1977), behaviour, growth and survival of the fish (Bradford et al. 2009). Where necessary, the suitability of a chosen tag type, size and attachment method should be assessed prior to tagging studies on a new species (CCAC 2005; Murray & Fuller 2000).

External tag attachment: External attachment is used for simple plastic or wire tags and electronic transmitters and requires that an external or trailing part of the tag be anchored somehow within the musculature or skeletal structure of the fish. A common type of external tagging, particularly in southern Africa, is dart tagging (Maggs & Cowley 2016). Dart tags are inserted using a sharp, sterile tag applicator, usually at the base of the dorsal fin and anchored behind the pterygiophores (Attwood 1998). Electronic tags can also be attached externally in a similar manner or with wires or nylon cords inserted through the dorsal musculature of the fish and secured with a plastic or metal plate (Liedtke & Rub 2012; Thorstad et al. 2013). Sharks and large pelagic fishes are often tagged in situ in the water, by means of a dart tag or electronic tag attached to the end of a sharpened pole. For pelagic sharks, tags (particularly satellite tags) are also commonly attached by means of plastic or metal bolts through the dorsal fin (Thorsteinsson 2002).

External tagging is simple and does not necessarily require animals to be anaesthetised, or even brought out of the water (Thorsteinsson 2002). External tags, however, can have negative impacts, as they may increase drag and impede normal swimming ability and reduce growth or survival (Murray & Fuller 2000). The attachment mechanisms breach the skin and musculature and could lead to infection, tissue damage or necrosis, reduced growth or reduced reproductive capacity (Thorsteinsson 2002). External tags, particularly larger volume electronic tags, should thus be shaped and attached in such a way that minimises drag, entanglement in aquatic vegetation, irritation and constriction (CCAC 2005; Murray & Fuller 2000). Owing to the disadvantages of external tag attachment, a number of internal tag implantation techniques have been used.

Gastric insertion: Gastric insertion of devices involves voluntary ingestion of the device embedded in bait or the forcing of the device down the pharynx, past the cardiac sphincter and into the stomach by means of a glass or plastic rod (Liedtke & Rub 2012; Thorsteinsson 2002). The process is less invasive than surgical implantation (Winger & Walsh 2001), and the internal device overcomes many of the problems associated with external tags, such as entanglement, drag or wound development (Stasko & Pincock 1977). However, gastric insertion can only be used when fish are not feeding and there are several drawbacks to the technique. There is a risk of tag loss through regurgitation, the presence of the device may affect the ability of the fish to feed and injury may be caused to the oesophagus or stomach (Thorstad et al. 2013; Winter 1996). Furthermore, gastric insertion cannot be used on fishes with a food-crushing pharyngeal apparatus, and the process does not allow for long-term tag retention (Stasko & Pincock 1977).

Surgical implantation of tags or devices: Laparotomy (the surgical implantation of a device through an incision in the body wall) is becoming recognised as a superior technique to external attachment and gastric insertion, particularly for long-term tag retention, reduced long-term physiological stress and data that may be more representative of normal behaviour (Cooke et al. 2011; Thorstad et al. 2013). The technique is considered to have little or no methodological bias in the long-term, and the internal placement of the tags removes the potential entanglement and hydrodynamic interference that may result from external tags (Jepsen et al. 2002; Thorstad et al. 2013). However, the recovery time after surgical implantation may be greater than that for external attachment or gastric insertion (Thorsteinsson 2002). Furthermore, this process requires suitably trained or qualified personnel and is governed in certain states by formal legislation. In South Africa, surgical procedures are regulated under the Veterinary and Para-Veterinary Professions Act (SA Government, Act No. 19 of 1982). For all tagging procedures, aseptic technique should be maintained, although this is difficult in most field settings. Antibiotic powder (topical) or liquid (topical or by injection) can be applied post-tagging to prevent secondary infection. Before releasing tagged fish back into their natural environment, they should be monitored until having recovered from the effects of anaesthesia and surgery.

Where the sacrifice of study fish cannot be avoided (e.g. voucher specimen collection or physiological studies), humane euthanasia techniques should be used (SABS 2008). Animals may be euthanised by chemical or physical methods, and the selected methods should be predictable, minimise pain and stress, produce rapid loss of consciousness and be compatible with the scientific aims (American Veterinary Medical Association 2013; SABS 2008). Fishes to be preserved should be euthanised prior to immersion in formalin or other preservatives (CCAC 2005). Where feasible, euthanasia should consist of a two-step process with initial anaesthesia until loss of equilibrium, followed by a physical or chemical method to cause brain death (CCAC 2005). Rapid cooling, followed by prolonged exposure in ice-cooled water, is recommended for small-bodied fishes (AVMA 2013). Overdosing with anaesthetics, such as MS-222, is an accepted method (Neiffer & Stamper 2009), whereas the use of carbon dioxide, suffocation (removal from water) and decapitation alone are generally not. The AVMA (2013) provides comprehensive guidelines on euthanasia of animals.