An overview of fish bioacoustics and the impacts of anthropogenic sounds on fishes

Fishes use a variety of sensory systems to learn about their environments and to communicate. Of the various senses, hearing plays a particularly important role for fishes in providing information, often from great distances, from all around these animals. This information is in all three spatial dimensions, often overcoming the limitations of other senses such as vision, touch, taste and smell. Sound is used for communication between fishes, mating behaviour, the detection of prey and predators, orientation and migration and habitat selection. Thus, anything that interferes with the ability of a fish to detect and respond to biologically relevant sounds can decrease survival and fitness of individuals and populations. Since the onset of the Industrial Revolution, there has been a growing increase in the noise that humans put into the water. These anthropogenic sounds are from a wide range of sources that include shipping, sonars, construction activities (e.g., wind farms, harbours), trawling, dredging and exploration for oil and gas. Anthropogenic sounds may be sufficiently intense to result in death or mortal injury. However, anthropogenic sounds at lower levels may result in temporary hearing impairment, physiological changes including stress effects, changes in behaviour or the masking of biologically important sounds. The intent of this paper is to review the potential effects of anthropogenic sounds upon fishes, the potential consequences for populations and ecosystems and the need to develop sound exposure criteria and relevant regulations. However, assuming that many readers may not have a background in fish bioacoustics, the paper first provides information on underwater acoustics, with a focus on introducing the very important concept of particle motion, the primary acoustic stimulus for all fishes, including elasmobranchs. The paper then provides background material on fish hearing, sound production and acoustic behaviour. This is followed by an overview of what is known about effects of anthropogenic sounds on fishes and considers the current guidelines and criteria being used world‐wide to assess potential effects on fishes. Most importantly, the paper provides the most complete summary of the effects of anthropogenic noise on fishes to date. It is also made clear that there are currently so many information gaps that it is almost impossible to reach clear conclusions on the nature and levels of anthropogenic sounds that have potential to cause changes in animal behaviour, or even result in physical harm. Further research is required on the responses of a range of fish species to different sound sources, under different conditions. There is a need both to examine the immediate effects of sound exposure and the longer‐term effects, in terms of fitness and likely impacts upon populations.

Since the onset of the Industrial Revolution, there has been a growing increase in the noise that humans put into the water. These anthropogenic sounds are from a wide range of sources that include shipping, sonars, construction activities (e.g., wind farms, harbours), trawling, dredging and exploration for oil and gas. Anthropogenic sounds may be sufficiently intense to result in death or mortal injury. However, anthropogenic sounds at lower levels may result in temporary hearing impairment, physiological changes including stress effects, changes in behaviour or the masking of biologically important sounds.
The intent of this paper is to review the potential effects of anthropogenic sounds upon fishes, the potential consequences for populations and ecosystems and the need to develop sound exposure criteria and relevant regulations. However, assuming that many readers may not have a background in fish bioacoustics, the paper first provides information on underwater acoustics, with a focus on introducing the very important concept of particle motion, the primary acoustic stimulus for all fishes, including elasmobranchs. The paper then provides background material on fish hearing, sound production and acoustic behaviour. This is followed by an overview of what is known about effects of anthropogenic sounds on fishes and considers the current guidelines and criteria being used world-wide to assess potential effects on fishes.
Most importantly, the paper provides the most complete summary of the effects of anthropogenic noise on fishes to date. It is also made clear that there are currently so many information gaps that it is almost impossible to reach clear conclusions on the nature and levels of anthropogenic sounds that have potential to cause changes in animal behaviour, or even result in physical harm. Further research is required on the responses of a range of fish species to different sound sources, under different conditions. There is a need both to examine the immediate effects of sound exposure and the longer-term effects, in terms of fitness and likely impacts upon populations.

K E Y W O R D S
behaviour, criteria, effects, guidelines, hearing, sound *This paper is dedicated to two long-time friends and colleagues, Colin Chapman and Richard R. Fay. Colin and Dick have been major contributors to our understanding of fish bioacoustics. Moreover, their work and their thinking has inspired many of the studies described in this review.

| INTRODUCTION
The past several decades have seen an increasing level of interest in the potential effects of anthropogenic sounds on aquatic life. The sounds added by humans into aquatic environments (both marine and freshwater), include those from shipping, dredging, sonars, seismic airguns used for oil and gas exploration, underwater explosions and construction, including pile driving, as well as many other activities.
Anthropogenic sounds such as these have increased in recent times as a result of increases in shipping, harbour developments, the construction and operation of offshore wind farms, tidal and wave energy generation, dredging and cable and pipe laying, seismic surveys for oil and gas and offshore oil developments. Although initial concern regarding anthropogenic sound focussed on the effects upon marine mammals (NMFS, 2018;NRC, 1994;Southall et al., 2007), there is now growing concern over potential effects upon those organisms that make up a much larger part of the aquatic biomass, fishes and more recently, invertebrates and zooplankton . Concern has also been expressed recently over effects upon freshwater fishes (Bolgan et al., 2016;, since they have received far less attention in research on noise effects studies. The added sounds in the aquatic environment may have a wide range of effects on fishes. Exposure to very intense sounds may result in mortal injuries, but far more important issues are associated with sounds that are detectable by fishes and which may affect their behaviour, causing them to move away from their migration routes, leave favoured habitats in which they feed or breed, interfere with communication using sound, affect reproductive behaviour (where sound is used to attract mates and facilitate spawning), or prevent the detection of other biologically important sounds. As a consequence, the addition of anthropogenic sounds to the aquatic environment has the potential to do significant harm to fishes.
From an historical perspective, fish bioacoustic studies up until the early 21st century asked basic questions about hearing, sound communication and behaviour. While such studies continue, many studies since about 2005 have focussed on the potential adverse effects of sounds on fishes. A driving force in this change has been the need by regulators, industry, environmental groups and scientists to develop guidelines and criteria that can be used to assess whether particular sounds have deleterious effects on individual fishes or affect populations. There has also been a need to employ such information in regulations intended to protect fishes and ecosystems.

| OVERVIEW
The purpose of this paper is to improve understanding of the issues related to the potential effects of anthropogenic sounds on fishes and to point to the need to examine effects not only on individual animals, but also to those on fish populations and ecosystems. However, since many readers may not be familiar with fish bioacoustics, we also include some background material to assist readers in understanding and interpreting data on the effects of anthropogenic sound. Accordingly, the paper starts with a brief discussion of underwater sound in order to introduce major concepts that are critical for understanding potential effects of anthropogenic sound. This is followed by a discussion of fish bioacoustics for those not familiar with the topic. We then focus on data on the potential effects of anthropogenic sound on fishes. It will become clear that there are major gaps in our knowledge that need to be filled in order to facilitate the development of appropriate and effective sound exposure criteria and the guidelines to implement them. Indeed, it is critical to understand that current criteria are still provisional and that substantially more data are required before firm criteria can be set. The review therefore ends with suggestions as to the most critical current data needs. This paper is not intended to be a complete review of all the literature. Rather, our focus is on the major issues related to potential effects of anthropogenic sound on fishes and to help readers understand those aspects that are especially important. However, we do provide further citations so that those interested can delve deeper into the growing literature on the topic and we include a number of recent reviews that refer to the wider literature. Furthermore, the papers we do include are those we think are amongst the most informative and critical to understanding the main issues. At the same time, we do include a number of papers that we see as being problematic, so that as well as providing insight into work that is critical to understanding the effects of sound on fishes, we also provide information on work that may lead to misunderstanding.

| ADDITIONAL BACKGROUND INFORMATION
For those interested in broadening their understanding of general issues of fish bioacoustics (hearing, sound production, behaviour etc.), there are papers in a volume by Webb et al. (2008) as well as several more recent reviews (Ladich, 2014;Ladich & Fay, 2013;Putland et al., 2018). More detailed reviews of potential effects of anthropogenic sound on fishes (and other aquatic animals) can be found in papers by the authors of this review  and in the reports of several meetings on the Effects of Noise on Aquatic Life (www.an-2019.org; Hawkins et al., 2008;Popper & Hawkins, 2012 and the open access Proceedings of Meetings on Acoustics (www. go.umd.edu/UcA). Finally, a general overview of effects of anthropogenic sound on animals is provided by Slabbekoorn et al. (2018).

| Acoustic terminology
It is important to distinguish between sound and vibration. Sound is generated by the movement of an object, such as a loudspeaker, or a pile being driven, in a medium such as air or water (Urick, 1983). The term vibration refers to the actual motion of the sound source. As the sound propagates from the source it can be detected as the pressure fluctuations in the medium, above and below the local hydrostatic pressure (the sound pressure). However, sound is also accompanied by a back-and-forth motion of the medium, referred to as the particle motion. (For a clear visualisation of sound pressure and particle motion see: www.dosits.org/science/sound/what-issound/). The term noise is often used to describe unwanted sounds that are considered to be unpleasant, loud or disruptive to hearing, or that can hinder detection of a particular signal. In some cases, however, the terms ambient noise or background noise may also be used, as it is in this paper, to describe sound generated by natural sources, as well as by anthropogenic sources, especially where they may interfere with the detection of animal and other sounds. 5.2 | Sound pressure, particle motion and the substrate Sound pressure is a scalar quantity that acts in all directions. It can be described in terms of its magnitude, as well as its temporal and frequency characteristics. In contrast, particle motion is a back-and-forth motion and, as such, is a vector quantity. Accordingly, particle motion is described not only by specifying its magnitude and temporal and frequency characteristics, but also its direction of motion.
Sound pressure is expressed in SI units of pascals (Pa) or micropascals (μPa). Particle motion may be expressed in terms of the particle displacement (SI unit: metre m), or its time derivatives: particle velocity (meter per second m s −1 ) or particle acceleration (meter per second squared (m/s 2 ). Sound intensity is the product of the sound pressure and the particle velocity, for which the SI units are A fundamental point is that all fishes (including elasmobranchs) detect and use particle motion, particularly at frequencies below several hundred Hz .
Thus, the detection of particle motion is integral to hearing in all fishes (and invertebrates) and it is used to locate the direction of the source, even in those fishes that are also sensitive to sound pressure Nedelec et al., 2016). As a consequence, when investigating the effects of sounds upon fishes, it is important to describe the sounds in terms of particle motion , as well as sound pressure. This may be done by measuring the particle motion directly (Amorim et al., 2018;Roberts & Breithaupt, 2016) or by conducting experiments under freefield acoustic conditions, where the particle motion can be predicted from measurements of the sound pressure .
Until recently, most studies of sound and fishes have only included measurement of the sound pressure and very few have considered particle motion in a biologically relevant context. This was not just because investigators did not fully appreciate the importance of particle motion, but also because of the difficulty in obtaining instrumentation to measure the particle motion (e.g., Lumsdon et al., 2018;Martin et al., 2016).
While it is possible to estimate particle velocity from measurements of the sound pressure (or by measuring the pressure gradient), this can only be done in locales that are distant from reflecting boundaries (the water surface or bottom) or other acoustic discontinuities (MacGillivray et al., 2004), since such surfaces have significant influence on the sound field and thus, on the levels and directionality of particle motion. Under such conditions, sensors are needed that not only detect particle motion (whether particle displacement or its time derivatives: particle velocity or particle acceleration) per se but are also able to detect the vector components in three dimensions.
Passage of sound and vibration into the substrate, which can be caused by sources such as pile driving, dredging and seismic surveys, may result in waves propagating through the substrate, both as compression waves and interface waves . The interface waves are often referred to as ground roll (Hazelwood et al., 2018). These waves travel slower than the speed of sound and can have strong particle motion components. They may also generate evanescent sound pressure and particle motion waves that propagate through the water.

| Sound metrics
It is very important to always refer to a sound using the proper measures, or metrics, that best describes that sound.

| Continuous sound
Continuous sound (e.g., from shipping) is generally presented as the root mean square (dB rms ) sound pressure or particle motion level, measured over a specified time interval, for a specified frequency range. The roughness of continuous sounds may be especially important when considering effects, using a statistic often called kurtosis (Henderson & Hamernik, 2012). However, while of potential importance, and while mentioned more and more frequently, kurtosis has yet to be applied to fish (or marine mammal) bioacoustics.

| Impulsive sounds
Impulsive sounds (e.g., from pile driving) are best presented as the instantaneous peak level, the dB peak . That is, the level of the zero-topeak sound pressure or particle motion. Alternatively, the total energy within the pulse may be described by the sound exposure level (SEL; Popper & Hastings, 2009). The SEL is the integral, over time, of the squared sound pressure, normalised to a reference time of 1 s. The SI unit of sound exposure is the Pascal squared for 1 s (Pa 2 s −1 ). The SEL may be specified for a single impulse or strike (the SEL ss ). However, when impulsive sounds are repeated, for example when fishes are exposed to pile driving for a long period, it is appropriate to estimate the cumulative SEL (SEL cum ) associated with a series of pile strikes.
The SEL cum is the total noise energy to which the animal is exposed over a defined time period (Popper & Hastings, 2009).
Another important characteristic of impulsive sounds is the rise time, which is the time a signal takes to increase from 10% to 90% of its highest peak value. The rise time may affect the response of animals and may be especially important in terms of injury, where sharp rise times may be especially damaging.

| Frequency spectrum
The frequency spectrum is also important. The sound pulse is composed of a range of frequencies, expressed in terms of the level at each frequency measured over a given bandwidth. The bandwidths utilised are generally 1 Hz or 1 / 3 octave (an octave is a doubling of frequency). It is important to specify the frequency bandwidth as different animals respond to different frequency ranges.
6 | NATURAL SOUNDS IN THE AQUATIC ENVIRONMENT 6.1 | Ambient sound Aquatic environments are rarely silent. Ambient sound (often termed ambient noise) consists of sounds generated by physical sources such as wind, waves, precipitation and ground movement (geophony), together with biotic sounds (biophony) produced by a variety of marine organisms, including mammals, fishes and invertebrates. Examining the soundscape involves describing the characterisation of ambient sound in terms of its spatial, temporal and frequency attributes and the types of sources contributing to the sound field.

| Fish sounds
Of the more than 33,000 species of fish, at least 800, from over 100 families, are known to produce sounds (Bass & Ladich, 2008). It is likely that with more studies, including freshwater fishes, many additional species will be shown to be sound producers. Many commercially important fish species produce sounds, including the Atlantic cod, Gadus morhua L. 1758 and haddock, Melanogrammus aeglefinus (L. 1758), both Gadidae (Hawkins et al., 1974;Hawkins & Chapman, 1966) and many croakers and drums (Sciaenidae; Ramcharitar et al., 2006). Sounds are produced in a wide range of contexts such as feeding, mating, or fighting (Hawkins & Myrberg Jr, 1983;Moulton, 1963).
The detection of sounds may be used by female fishes to locate vocal males and identify suitable mates (Casaretto et al., 2015). As a consequence, anything, including sounds from anthropogenic sources, that impedes the detection of these sounds can have an adverse effect on such fishes.

| ANTHROPOGENIC SOUND SOURCES
There are many sources of anthropogenic sound in the sea, lakes and rivers, with quite different acoustical characteristics Popper et al., 2014). Many commercial human activities introduce sound, either intentionally for a specific purpose, such as seismic surveys, or unintentionally as a by-product of activities such as shipping and offshore and even onshore construction work. Coastal areas and areas where a high degree of human activity takes place, may be quite noisy; including harbours and shipping lanes. However, some highintensity sources of underwater sound, such as pile drivers and seismic airguns, can be detected over distances of several thousand kilometres.
Thus, effects upon fishes may occur well away from the source itself.

| Seismic airguns
Airguns are impulsive sources used for seismic exploration for sub-sea gas and oil reserves as well as for geological research (Gisiner, 2016).
These devices use compressed air to produce a gas bubble which expands rapidly when released, creating a high intensity impulsive sound, primarily composed of energy below 200 Hz, but with the bulk of the sound from 20 to 50 Hz (Mattsson et al., 2012). The sounds are directed downward into the seabed, though there is also some spreading laterally and they are reflected from various geological formations and then detected by a long array of hydrophones towed by the seismic vessel (see Gisiner, 2016 for a detailed description of seismic surveys).

| Impact pile driving
Impact pile driving is widely used for the construction of bridges, harbours, wind farms and other offshore structures (Dahl et al., 2015;Popper & Hastings, 2009). Striking by the hammer results in vibration of the pile in water and in the substrate, thereby generating sounds that potentially affect nearby animals (Dahl et al., 2015;Hazelwood & Macey, 2016). The sounds produced by pile driving are impulsive, short (of the order of μs) and most of their energy lies below 500 Hz, though some energy may extend up to 1 kHz (Dahl et al., 2015). The sound levels (both sound pressure and particle motion) vary substantially, depending on numerous factors such as pile diameter, hammer size, substrate characteristics, etc. The sounds produced by pile drivers are often very intense with SEL ss often well-exceeding 180 to 200 dB re 1 μPa 2 s −1 and with very sharp rise times.

| Other industrial activities
Many other industrial activities contribute to underwater noise. Such activities generally produce sound that has the most energy at low frequencies (i.e., <1 kHz). Dredging, for example produces high levels of broadband noise (de Jong et al., 2016;Wenger et al., 2017) and is used to extract sand and gravel from the seabed and from lakes, maintain shipping lanes and to install pipelines and cables within the seabed. Activities onshore, including the passage of vehicles, may increase noise levels in the sea, lakes and rivers, especially if they generate substrate vibration.

| Operating wind turbines
Since c. 2000 there has been an enormous increase in the generation of electricity by wind farms located in coastal waters, especially in European seas. There is some concern that sounds from operating offshore wind turbines might affect fish behaviour, although the sounds generated are very different to those generated during wind-farm construction (Cheesman, 2016). Most sound from a wind turbine is concentrated in a narrow band, centred around 180 Hz and the sounds are generally below about 700 Hz (Madsen et al., 2006;Pangerc et al., 2016). However, there is also a particle motion component to the sounds generated by wind turbines, accompanying substrate transmission (Sigray & Andersson, 2012; P. Gopu and J. Miller, personal communication, 2018), although this has rarely been monitored and has often been ignored. There is currently limited information available on the acoustic characteristics of offshore turbines, including those utilising tidal and wave energy (Lossent et al., 2018;Schramm et al., 2017).

| Vessel noise
A significant proportion of anthropogenic noise in the ocean and other water bodies is created by motorised vessels, including large ships, fishing and pleasure boats (Pine et al., 2016;Rossi et al., 2016). Most vessels, and especially large ships, produce predominately low frequency sound (i.e., <1 kHz) from onboard machinery and hydrodynamic flow around the hull. Cavitation at propeller blade tips is also a significant source of noise across all frequencies (Ross, 1987(Ross, , 1993. Low frequency sounds from ships can travel hundreds of kilometres and can increase ambient noise levels over large areas of the ocean (Ellison et al., 2012;Southall, 2005).
Ambient noise levels in busy shipping lanes have recently increased (Hildebrand, 2009), across much of the frequency spectrum (Sertlek et al., 2016), but especially at lower frequencies (<500 Hz; Erbe et al., 2012;Bittencourt et al., 2014). Large numbers of smaller pleasure and recreational vessels, including things like jet skis (Erbe, 2013), may also result in substantial increases in noise levels in coastal waters, lakes and rivers. Ice-breaking ships can be a significant source of sound in polar regions. There is a long history of fish hearing studies (Moulton, 1963;Tavolga, 1971). It is likely that all fishes (including elasmobranchs) detect sound and use it to learn about their environment (e.g., Ladich & Fay, 2013).
Until recently, however, most studies have focussed on determination of hearing capabilities of fishes to sound pressure signals, despite it being clear that most fishes (and all elasmobranchs) primarily detect particle motion . (As an aside, lampreys (Petromyzontidae) also have an ear that has many characteristics in common with other vertebrates and both morphological (Popper & Hoxter, 1987) and recent physiological results  suggest that they only detect particle motion). There is a need to investigate the hearing abilities of lampreys and many other fishes, under conditions where the particle motion can be monitored or estimated and the ratios of these two potential stimuli can be varied. Such experiments have been reviewed in a number of recent papers, including Hawkins (2014) and Putland et al. (2018).
In addition to not focussing on particle motion, many studies have in glass 500 ml vials). In such environments, the sound fields presented to the fish are generally very complex and quite unlike the sound fields that a fish would encounter in a normal aquatic environment . As a result, such experiments often leave open questions regarding the actual nature of the sound field to which the animals were exposed and the stimuli to which they responded . Ideally, hearing experiments should be carried out in specially designed tanks (Duncan et al., 2016;Hawkins & MacLennan, 1976;Rogers et al., 2016) or in natural aquatic environments, where both the particle motion and the sound pressure levels can be monitored precisely.
Keeping these caveats in mind, it is possible to get some appreciation of hearing capabilities of fishes. For example, every species studied to date is able to hear. In addition, the majority of fishes detect sounds from <50 Hz, even as low as 10-30 Hz, or even lower (Sand & Karlsen, 2000) to perhaps 300-500 Hz. Fishes that can detect sound pressure hear to perhaps 1000 Hz. And, a much smaller number of species have specialisations that enable them to detect sounds to 3-4000 Hz (Ladich & Fay, 2013).
Because relatively few experiments on the hearing of fishes have been carried out under suitable acoustic conditions, valid data that provide actual hearing thresholds are available for only a few species (thresholds are generally defined as the lowest level of sound that can be detected 50% of the time). Figure 1 shows the measures of hearing, expressed as audiograms. determined in the open-sea, rather than in a laboratory tank, for: the flatfish common dab Limanda limanda (L. 1758) (Chapman & Sand, 1974); the Atlantic salmon, Salmo salar L. 1758 (Hawkins & Johnstone, 1978); the G. morhua; (Chapman & Hawkins, 1973); the Atlantic herring, Clupea harengus L. 1758 (Enger, 1967). The L. limanda and S. salar are only sensitive to particle motion and have a relatively narrow bandwidth of hearing (up to c. 300-500 Hz), whereas species like G. morhua, where the gas-filled swimbladder is close to the ear, are sensitive to sound pressure and show an increased hearing bandwidth (Fay & Popper, 1974;Sand & Hawkins, 1973).
It is important to understand that the swimbladder (and other gas-filled cavities) potentially plays a major role in fish hearing. This is because the gas within the swimbladder is compressible and changes volume in response to fluctuating sound pressures (sound) and this results in the swim bladder serving as an acoustic transformer, translating sound pressure into re-radiated particle motion (Sand & Hawkins, 1973). This produces higher levels of particle motion at the ears that stimulates the otolith organs . Thus, having a gas bubble or a swimbladder close to, or connected to, the ear enhances the hearing abilities of fishes since the ear is not only stimulated directly by the particle motion component of the sound, but also indirectly by the particle motion reradiated from the gas bubble to the ear in response to sound pressure. The actual contribution of the indirect stimulation varies by species and depends on the distance between the bubble and the ear. For example, in G. morhua, hearing at low frequencies (<110 Hz), is based on the detection of particle motion, but at higher frequencies it is based on sound pressure due to the closeness of the anterior end of the swimbladder to the ear.
Indeed, deflation of the swimbladder in G. morhua reduces sensitivity to sound pressure (Sand & Enger, 1973) and similar results have been shown for the goldfish Carassius auratus (L. 1758) (Fay & Popper, 1974).
In contrast, species like S. salar, despite having a swim bladder, are only sensitive to particle motion since the swimbladder is more distant from the ear (Hawkins & Johnstone, 1978;Knudsen et al., 1992). Other species, such as C. harengus (as all Clupeiformes) has a specialised connection between a gas bubble as the ear and shows sensitivity to a much wider range of frequencies and this can extend to >100 kHz in clupeids of the shad family Alosinae (Mann et al., 1998;Mann et al., 2001). Finally, species that do not have a swimbladder or other gas bubble, such as flatfishes, some scombrids and some gobies, only detect particle motion and hear over a narrower bandwidth than G. morhua.
In addition to having a gas bubble that improves hearing sensitivity and bandwidth, a number of fish species have additional adaptations that mechanically link the swimbladder to the ear, thereby carrying the motion of the swimbladder to the ear without attenuation of the signal as a result of distance of travel. Best known of these adaptations are the Weberian ossicles, a series of bones that connect the swimbladder to the inner ear in otophysan fishes. Popper & Fay, 2011). In other species, the swimbladder has extensions that come close to, or may actually contact, portions of the inner ear and most notably to the saccule, the otolith organ most fre-  The lower the thresholds (y-axis), the more sensitive the fish is to a sound. Thus, Clupea harengus has best hearing of all of these species over a wider range of frequencies. Note that the thresholds in Gadus morhua and C. harengus obtained under quiet conditions may be below natural ambient noise levels, especially at their most sensitive frequencies. In the presence of higher levels of noise, the thresholds would be raised, a phenomenon referred to as masking. Gadus morhua and C. harengus are sensitive to both sound pressure and particle motion, whereas Limanda limanda and Salmo salar are only sensitive to particle motion. The reference level for the particle velocity is based on the level that exists in a free sound field for the given sound pressure level. n.b., For the particle velocity levels in this figure to match the sound pressure levels in a free sound field it is necessary to calculate an appropriate particle velocity reference level. If the standard reference levels are used, then the curves will not match one another and so they are not included here to keep the figure relatively simple. Fig. ©  For the more sensitive fishes, hearing is not limited by the lowest level they can hear in a quiet environment, but by their ability to detect and discriminate biologically important sounds against the ambient noise background (Figures 1 and 2). In such conditions, the level of noise limits the lowest sound level that an animal can detect. This interference with detection of a biologically relevant sound by another sound, or noise, is generally known as masking and it is commonly found in all vertebrates, including fishes (Fay & Megela Simmons, 1999). As an example of masking, G.morhua only show best hearing sensitivity under the quietest sea conditions (Figure 2; Chapman & Hawkins, 1973). Any increase in the level of ambient sea noise results in a raising of the auditory threshold and a decline in the ability of the fish to detect, locate and recognise particular sounds. Critically, the masking of biologically relevant sounds occurs not only as a result of increases in natural ambient sea noise (caused by wind and rain) but also by any additional sounds added to the environment by humans. However, fishes that do not hear well may be less likely to have their hearing sensitivity affected by masking noise, since the lowest sound level they can detect may be above the level of the background noise (Hawkins & Johnstone, 1978).
Although the detection of sounds may be affected by the presence of masking sounds, it is also clear that fishes can use frequency filters to improve sound detection. They can also discriminate between different sound frequencies and intensities. They are also able to determine the direction from which sounds come (sound source localisation), a critical ability since this enables fishes to move towards potential food sources or away from predators (Fay, 2005;Fay & Megela Simmons, 1999;Sand & Bleckmann, 2008).

| The ear
Fishes detect sound with paired inner ears (Figure 3), located in the cranial cavity lateral to the brain at the level of the medulla (Figures 3   and 4), that closely resembles ears found in other vertebrates. Since a fish's body is the same density as water, there is no need for any external structures (external or middle ears) to carry sound to the sensory regions of the ear. The ear consists of three semi-circular canals and associated sensory regions (ampullae) that are primarily involved in detection of angular acceleration and three otolith organs (saccule, lagena, utricle) that are involved in hearing and positional senses . There is very substantial variation in the morphology of the ears of fishes and particularly in the regions associated with hearing Retzius, 1881;Schulz-Mirbach et al., 2018;, leading to the suggestion that there is very substantial diversity in hearing mechanisms (and potentially capabilities) in different species .
The auditory parts of the ear, the otolith organs, each have a sensory epithelium that lies in close contact with a dense calcium carbonate structure, the otolith (Figures 3 and 4). The sensory epithelium (often referred to as a macula) has many sensory hair cells that are very similar to those found in the mammalian ear ( Figure 5). When a fish is exposed to particle motion, the body, along with the sensory cells, move with the water, while the far denser otoliths move at a different amplitude and phase. This results in bending of the cilia on the apical surface of the sensory cells, releasing a neurotransmitter and sending a signal to the brain through an afferent neuron.
A critical role of the ear in fishes is involvement with determination of sound source direction . The sensory hair cells are morphologically polarised and the response of an individual cell changes with bending in different directions. Thus, each cell is directionally sensitive. Furthermore, the cells are organised into orientation groups in which all of the kinocilia are in the same direction ( Figure 5). These hair cell orientation patterns, which vary in different species (Popper & Coombs, 1982), show graded responses to particle motion from various directions, thereby enabling a fish to determine direction by comparing information from different receptor groups (Fay, 2005;Sand & Bleckmann, 2008).

| EFFECTS OF ANTHROPOGENIC SOUND
There are very few experimental examples of sound being sufficiently loud to result in death or mortal injury to fishes. However, far more importantly from the perspective of potential effects, is that anthropogenic sound, even at levels far lower than those that might result in mortality, may result in temporary hearing impairment, physiological changes, changes in behaviour and the masking of biologically important sounds ( Masking in the Gadus morhua and Salmo salar by ambient noise. The thresholds were determined using a pure tone signal at a frequency of 160 Hz. The ambient noise (natural sea noise, augmented by white noise from a loudspeaker) is expressed as the spectrum level at that same frequency (dB re 1 μPa/Hz). Closed symbols, thresholds to natural levels of ambient noise; open symbols, thresholds to anthropogenic noise. n.b., The thresholds in S. salar were only influenced by high noise levels, above the natural ambient levels of noise (data from Hawkins, 1993). However, as pointed out by Hawkins et al. (2015), there are large gaps in our knowledge of effects of sound on fishes that need to be filled if we are to fully understand the implications of exposure to anthropogenic sounds.

Ear
Gas-gilled swim bladder Cranial cavity  would have ratios of pressure and particle motion unlike those that an animal would ever encounter in the wild (Parvulescu, 1964;Rogers et al., 2016). Thus, even if a fish shows a particular behaviour pattern in response to a tank-based sound, the same sound produced in the wild may have very different acoustic characteristics and thus may or may not elicit the same behaviour as in the tank. It is important to monitor the particle motion as well as the sound pressure and where possible to ensure that the acoustic conditions under which experiments are conducted are similar to those the fish would experience in the wild. Where the particle motion is properly monitored as well as the sound pressure, some physiological experiments on captive animals may provide some useful information on the levels that produce particular effects. However, it is necessary to be circumspect in extrapolating the findings to wild animals.
Finally, in considering behaviour, it is also important to recognise that the responses of fishes may vary with their age and condition, as well as under different environmental conditions. Moreover, responses may vary with different sound sources, or with the same , The approximate dividing lines between orientation groups) Another issue is that many fishes migrate to feeding areas or spawning grounds and may subsequently return to other locations. Acoustic cues guide the orientation, habitat selection and settlement of many fishes, but these processes may be impaired if degradation alters reef soundscapes.
Sounds are also important for many fish species for spawning. In observed the behaviour of schools of S. sprattus and S. scombrus in mid water at a quiet coastal location, using an echosounder. Sprattus sprattus is sensitive to sound pressure, while the S. scombrus is likely to be sensitive only to particle motion. The fish were exposed to short sequences of repeated impulsive sounds, simulating the strikes from a pile driver, at different sound levels. Results showed that the incidence of behavioural responses increased with increasing sound level. There have also been a number of studies of the response of captive demersal species to pile driving sounds. For example, Neo et al. (2014) found that that intermittent sounds may yield longer-lasting behavioural effects than continuous sounds (Neo et al. 2015). Moreover, ramp-up procedures, where sounds are slowly increased in level so as to warn fishes of impending sounds, do not necessarily lead to mitigation (Neo et al. 2016). At the same time, these studies were done in enclosures that did not resemble natural acoustic environments and many were done with D. rerio, a species that is small, thrives in small tanks and which hears far better than most (if not all) species likely to be exposed to pile driving operations. Kastelein et al. (2015Kastelein et al. ( , 2017 determined acoustic dose-response relationships for behavioural responses to the play back of pile driving sounds by D. labrax in a netting enclosure within a very shallow rectangular pool, where the sound field was nothing like that in the wild.
It was concluded that if wild D. labrax were exposed to pile driving sounds at the levels used in the study, there were unlikely to be any adverse effects on their ecology, because their initial responses were short-lived. However, the experiments were carried out on fish that had spent their whole lives in captivity.
In a more detailed series of experiments on laboratory-bred juve-  (Løkkeborg et al., 2012), although they generally returned to the original site within hours or days after the end of the seismic operation (Engås et al., 1996;Engås & Løkkeborg, 2002). Other studies have shown that fish may respond to approaching vessels by diving towards the seafloor or by moving horizontally out of the vessel's path, with reactions often initiated well before the vessel reaches the fish (Ona et al., 2007).

| Effects upon hearing sensitivity
Exposure to sounds may result in hearing loss as a result of damage to the sensory cells of the inner ear or the innervating neurons. While temporary hearing loss (TTS) occurs in fishes, there is no evidence for permanent hearing loss (PTS). Indeed, PTS may not occur in fishes since they can repair or replace sensory hair cells of the inner ear that have been lost or damaged (Smith et al., 2006;Smith & Monroe, 2016). TTS is a short duration decrease in hearing sensitivity resulting from exposure to intense sounds or sounds of long duration. After termination of the sound, normal hearing ability returns over a period that may range from minutes to days, depending on many factors, including the intensity and duration of exposure (Amoser et al., 2004;Smith et al., 2006;Smith & Monroe, 2016). However, during a period of TTS, animals may be placed at some risk to survival in terms of poorer communication, inability to detect predators or prey and difficulty in assessing their environment.
TTS has been demonstrated in a number of fish species from a diverse array of sounds (Smith & Monroe, 2016) but in all cases, TTS was only found after multiple exposures to intense sounds (e.g., < 190 dB re 1 μPa rms) or as a result of long-term exposure (e.g., tens of minutes or hours) to somewhat less intense sounds. Even when a signal source caused TTS in some individuals or species, it did not occur in other specimens or other species (Popper et al., 2005;Popper et al., 2007). In most cases, normal thresholds returned within a few hours to several days. There is also evidence that, given the same type and duration of sound exposure, a much more intense sound will be required to produce TTS in fishes that do not hear well compared with fishes that do hear well Smith et al., 2004). Since TTS can arise from prolonged exposure to sound (though this is not always so), it is not likely to be of great significance for fishes that are only briefly exposed to a source Popper et al., 2007).
Of far greater concern is that TTS may occur when there is longterm noise exposure such as in harbours and other areas where there is a long-term increase in sound level. While limited, TTS is correlated with damage to sensory hair cells of the ear and it has been shown that recovery from TTS occurs in parallel with repair or replacement of sensory cells (Smith et al., 2006;Smith et al., 2011). Other studies have shown that exposure to intense sound may result in hair cell damage, but they did not examine whether this was accompanied by a loss of hearing (Casper et al., 2013b;Enger, 1981;Hastings et al., 1996;McCauley et al., 2003). At the same time, studies of other species or other types of intense sounds have not resulted either in TTS or hair cell damage (e.g., Halvorsen et al., 2013;Popper et al., 2005;Popper et al., 2007).
Clearly, there is still a question as to whether TTS occurs in fishes exposed to anthropogenic sounds and, if so, which sounds will result in TTS. Moreover, there appears to be broad species variation as to whether TTS will occur and there is even evidence that different genetic stocks of the same species may or may not show TTS Popper et al., 2007). Moreover, none of the studies on TTS to date have determined whether the loss of hearing (or lack of loss of hearing) is correlated with exposure to sound pressure or particle motion. Finally, none of the studies have been done on wild animals where there is the potential to escape from areas of intense sounds, or to test whether a small change in hearing threshold has any real impact on fitness .

| Stress
Animals showing no overt sign of responding to an environmental stimulus may, nonetheless, experience physiological changes that are often referred to as stress responses. These are often similar to stress effects to sound exposure found in terrestrial animals (Gourévitch et al., 2014;Kight & Swaddle, 2011;Weilgart, 2017;Wysocki et al., 2006). Stress may include hormonal, autonomic, immune and behavioural responses that may initially allow fishes (as other animals) to adapt to adverse conditions. However, some stressors may change the state of physiological processes and affect homeostasis, thus having an adverse effect upon the animals' health and well-being. Very little is known about stress effects in fishes and the significance of such effects in response to anthropogenic sounds is even less clear (Tennessen et al., 2016).
There is an increasing body of literature on potential stress effects of exposure to both continuous and impulse anthropogenic sounds (Buscaino et al., 2010;Celi et al., 2016;Nedelec et al., 2015; Sierra-Flores et al., 2015). However, as for behavioural studies, there is a wide range of species used, a diverse set of exposure paradigms, very different results depending on species and paradigm, and, most importantly, all of these studies have been done in the laboratory.
Consequently, one must be cautious in extrapolating to how a fish might respond to a stressor in the wild where the fish's movement is not restrained and it could, potentially, move away from a stressor. It is also important to distinguish between normal or tolerable variations in response to environmental stress from those changes that will have consequences for survival and reproduction. At present, critical examination of these long-term changes in fishes as a result of sound exposure is lacking.
In considering potential physiological effects, a critical issue is that potential effects of sounds on the physiology of fishes, as measured by various stress parameters, are quite variable and are not particularly instructive with regard to how exposure might affect fishes.
In particular, all of the studies to date, including both long and shortterm exposures, were made on captive animals in enclosed areas where the fishes could not avoid the sounds. Thus, the acoustics were different than those an animal would encounter in the wild and the fish could not move away from the disturbing sound. Thus, it is possible that it is not the sound itself that resulted in the stress response, but the inability of the animals to move away from the sound. and mid-frequency sonars resulted in no mortality Popper et al., 2007), nor did exposure to seismic airguns (Popper et al., 2005;. There are, however, some data showing that fishes receiving high intensity and particularly impulsive, sounds will experience damage to body tissues. This damage appears to result from rapid oscillation of the walls of the swimbladder when stimulated by an impulsive source. In such cases, it appears that the swimbladder expands and contracts rapidly, thereby damaging the proximate organs including liver, kidney, gonads and the swimbladder itself (Halvorsen et al., 2012b;Halvorsen et al., 2012c).

| Death and injury
For example, of five species exposed to high intensity simulated pile driving signals (Casper et al., 2013a;Halvorsen et al., 2012b;Halvorsen et al., 2012c), only the hogchoker Trinectes maculatus (Bloch & Schneider 1801), a flatfish without a swim bladder, showed no tissue damage (Halvorsen et al., 2012b). At the same time, exposure to very high intensity continuous signals that did not contain any impulsive components showed no tissue damage in five different species (Halvorsen et al., 2012d;Halvorsen et al., 2013;Kane et al., 2010;Popper et al., 2007).
A recent set of studies, using a pile driving sound as a stimulus, enabled investigators to quantify the physical effects of sound exposure on various tissues (Casper et al., 2012;Casper et al., 2013aCasper et al., , 2013bCasper et al., 2017;Halvorsen et al., 2012aHalvorsen et al., , 2012bHalvorsen et al., , 2012cPopper et al., 2013). While these results directly relate to pile driving, they are also likely to give guidance for potential effects of other impulsive sounds on fishes and so they have been incorporated into POPPER AND HAWKINS 703 FISH the most recent guidelines for fishes on interim sound exposure criteria (Table 2; Popper et al., 2014;Andersson et al., 2017).
In brief, results from these studies showed a general correlation between the extent of tissue damage and the cumulative level of sound energy to which fish were exposed. For example, there was no tissue damage in one of the main study species, Chinook salmon Oncorhynchus tshawytscha (Walbaum 1792), following exposure to sounds below an SEL cum of 210 dB re 1 μPa 2 s −1 . At an SEL cum that was a few dB higher (but with sounds given over the same time period), internal injuries started to appear and when the level reached 219 dB re 1 μPa 2 s −1 there were massive internal injuries that would likely result in death. Studies with other species showed that while there is some variation in SEL cum required for onset of physiological effects, this is always at SEL cum levels >203 dB re 1 μPa 2 s −1 (Casper et al., 2013a;Halvorsen et al., 2012b).
At the same time, results show that the effects do not support the idea of an equal energy hypothesis, which is an idea based on the premise that the same effect will show up as long as the total energy to which a fish is exposed remains the same (Woodbury & Stadler, 2008). Instead, experimental results clearly show that the degree of effect depends upon a combination of the energy within single strikes (SEL ss ) and the number of strikes, but the effect is not predictable from just knowing the cumulative energy (Casper et al., 2016;Halvorsen et al., 2012c).
Studies subsequently found that O. tshawytscha and hybrid white bass Morone chrysops (Rafinesque 1820) x striped bass Morone saxatilis (Walbaum 1792), recovered from all apparent physical effects within 10 days of exposure (Casper et al., 2012(Casper et al., , 2013a. However, it was made clear that recovery took place in the laboratory and that animals in the wild with similar injuries would have lower fitness and be more susceptible to predation and disease until they fully recovered. This is a concrete example of the need to be cautious in interpreting the results of laboratory experiments. An additional question was whether hearing was affected by exposure to up to 960 sequential simulated pile strikes. Limited data showed that damage to ear tissues did not show up until the SEL cum was 216 dB re 1 μPa 2 s −1 (Casper et al., 2013b). However, both species studied have swim bladders that terminate some distance from the ear and so movement of the swimbladder walls would not directly affect the inner ear. It is possible that fishes with gas-filled organs near or directly associated with the ear would show damage at lower sound exposure levels due to the impulsive movement of the organ walls, much as they damage other nearby tissues.

| EFFECTS ON FISH POPULATIONS AND THE WIDER ECOSYSTEM
The studies described previously have largely dealt with effects upon individual animals. However, for fishes, unlike marine mammals, perhaps the greater concern lies with effects upon populations rather than individuals (Hawkins & Popper, 2016a;Pirotta et al., 2018). The   (Halvorsen et al., 2012b(Halvorsen et al., , 2012c. Temporary threshold shift (TTS) based on Popper et al. (2005). The same peak levels are used both for mortality and recoverable injury since the same sound exposure level (SEL ss ) was used throughout the pile driving studies. All criteria are presented as sound pressure even for fishes without swim bladders since no data for particle motion exist. Relative risk (high, moderate, low) is given for animals at three distances from the source defined in relative terms:  reproduce, or rear young and to quantify how this impairment may affect the abundance of the species concerned.
For species where there is limited knowledge of ecological interactions, an alternative risk assessment tool is required. Fisheries biologists have recently considered new risk-based approaches in assessing the effects of fishing upon species for which there are only limited data on key population parameters. The productivity susceptibility assessment (PSA) approach (Patrick et al., 2010) has been applied to fish stocks to determine the effect of human activities upon fishes.
Such an approach attempts to evaluate the vulnerability of fish stocks to fishing; based on their biological productivity and potential for resisting adverse effects. This approach has been increasingly used to identify species at risk within multispecies fisheries (Hobday et al., 2011;Smith et al., 2007) and may have wider applicability in assessing risks from noise exposure.

| SOUND EXPOSURE CRITERIA AND GUIDELINES
Sound exposure criteria essentially define those levels of sound from different sources that are likely to affect aquatic animals adversely, in order to regulate the generation of noise in aquatic environments. Significant efforts have been made over the past few years to develop criteria for aquatic vertebrates, including marine mammals, as well as guidelines for the use of these criteria (NMFS, 2018;Southall et al., 2007).
Substantially less effort has been placed on developing criteria and guidelines for fishes. However, interim sound exposure criteria for the onset of physiological effects on fishes for use on the United States west coast were proposed by the Fisheries Hydroacoustics Working Group (FHWG, 2008) but also see Popper et al. (2006) and Woodbury and Stadler (2008). More recently, a new set of interim criteria was proposed  based on a much stronger set of research and these raised the effective onset of effects levels, at least for physical effects, substantially and these interim criteria are now being used world-wide (Andersson et al., 2017).
Most work to date has focussed upon effects on marine mammals and marine fishes; much less is known about these effects in fresh water. However,

| Current interim guidelines
The term onset and the phrase onset of effect have been widely used in preparing guidelines on sound exposure criteria. However, it is clear that onset is viewed very differently by different investigators, regulators and others and that there is no clear definition of the term, particularly with regard to the potential effects of sound on fishes. In this review, onset refers to the lowest sound level that results in a statistically significant effect, in terms of physical damage to an animal or a significant change in behaviour. It should be noted that earlier papers that considered fishes used onset for any level of response, including a response by a single animal in a school (Woodbury & Stadler, 2008).
Thus, if there is scale loss in one fish within a group of many animals, that would be considered onset.

| Onset of physical effects
The interim sound exposure criteria, which are still in use, at least on the U.S. west coast (Caltrans, 2015;www.go.umd.edu/UcP), were based on a recommendation of dual criteria of peak sound pressure (SPL peak ) and cumulative SEL (SEL cum ) (Carlson et al., 2007;Popper et al., 2006;Popper & Hastings, 2009).
The rationale for dual criteria was that it was sometimes hard to determine one or the other measure when trying to set a signal level for onset of an effect and having alternative approaches provides a more conservative guideline for the protection of the animals. The SEL cum was suggested since animals are often exposed to many more than a single pile driving strike in succession and any effect would probably come from an accumulation of energy from the multiple strikes. However, as noted above, it is now clear that the SEL cum is probably an inappropriate measure of potential effects.
In 2008, the Fisheries Hydroacoustic Working Group adopted the interim dual-criteria model for onset of physiological effects from sound exposure (FHWG, 2008). However, these criteria were immediately criticised since they were based on very limited scientific research on effects of pile driving on fishes (Carlson et al., 2007;Popper & Hastings, 2009). The criteria were: Peak (SPL): 206 decibels (dB) re 1 μPa; SEL cum : 187 dB re 1 μPa 2 s −1 for fishes above 2 g; SEL cum : 183 dB re 1 μPa 2 s −1 for fishes below 2 g.

| Onset of behavioural effects
The U.S. National Marine Fisheries Service (NMFS), as well as other agencies, currently uses 150 dB re 1 μPa (rms) as the sound pressure level that may result in onset of behavioural effects (Caltrans, 2015).
This is based on a recent NMFS guidance document (www.go.umd. edu/Ucs) that says that sound pressure above the 150 dB rms level are expected to cause temporary changes in behaviour and these might include startle responses (though startle is not defined and has broad meaning to fish biologists), feeding disruption, area avoidance, etc.
However, there are a number of problems with the 150 dB rms criterion. First, its origin and scientific basis is not known (Hastings, 2008).
Second, the value is based on the assumption that fishes respond to sound pressure even though, as pointed out earlier, most fishes primarily detect particle motion (see also . Thus, any behavioural criteria should be based on the acoustic signals that the fish can actually detect and respond to. Finally, and perhaps most importantly, a single criterion value for behaviour does not take into consideration the very substantial species differences in hearing sensitivity, behaviour, etc., nor does it take into consideration POPPER AND HAWKINS 705 FISH response changes with animal age, season, or even motivational state (Neo et al., 2014).

| Recent criteria and guidelines
More recently, a set of interim criteria and guidelines for fishes was developed based on recent scientific advances (Table 2; Popper et al., 2014). Of major importance, the authors concluded that it was not possible to define sound exposure criteria for every possible sound source, type of response to the sound, or do an analysis for every fish species (or even all of those potentially listed in various locales).
Instead, they developed an approach that focussed on fish groups based on morphology of auditory apparatus (Table 3), on major sound types (e.g., pile driving, shipping) and major potential effects ( Table 1).
The overall intent was to provide the first science-based, but clearly interim, criteria for effects of anthropogenic sound on fishes and to provide a way to deal with the potentially insurmountable combinations of species and sources. The authors very carefully, however, pointed out that the proposed criteria were not complete due to lack of data (Table 2 provides examples of the several effects tables found in the guidelines) and that they expected that as more studies were done, the suggested criteria would evolve.
Finally, the authors of the guidelines made it clear that many of the acoustic impact assessments carried out on fishes in the past and upon which the interim guidelines were based, must be amended since they only considered sound pressure and did not take into consideration the potential effects from high levels of particle motion, something that must be done in future iterations of the guidelines Hawkins & Popper, 2016b;Nedelec et al., 2016;. There is growing international awareness that fishes do possess particle-motion receptors and that this must be taken into account in setting future criteria, once appropriate data are available.

| European guidelines for fishes
The monitoring of underwater noise is included in the European Union's Marine Strategy Framework Directive (MSFD; EU, 2008), which is concerned with ensuring good environmental status (GES) of European waters (Andersson et al., 2017;Dekeling et al., 2016;Tasker et al., 2010;Tasker et al., 2012;van der Graaf et al., 2012). The directive requires that the introduction of energy, including underwater noise, must be at levels that do not adversely affect the marine environment. No specific criteria for fishes are provided but indicators for achieving GES are specified.
The Swedish Environmental Protection Agency recently published a review that discusses regulation of pile driving sounds (Andersson et al., 2017). While the review focusses on sound pressure, the authors also strongly concurred with the idea that future guidelines for fishes must also be in terms of particle motion and must also consider signals  Fishes lacking swim bladders that are sensitive only to sound particle motion and show sensitivity to only a narrow band of frequencies (e.g., flatfishes, Pleuronectiformes; sharks skates and rays, Chondrichthyes).
2 Fishes with a swimbladder where that organ does not appear to play a role in hearing. These fish are sensitive only to particle motion and show sensitivity to only a narrow band of frequencies. This group includes salmonids (Salmonidae) and some tunas and mackerels (Scombridae), but many other species are likely to fit into this category as well. 3 Fishes with swim bladders that are close, but not intimately connected, to the ear. These fishes are sensitive to both particle motion and sound pressure, and show a more extended frequency range than groups 1 or 2, extending up to about 500 Hz. This group includes cod fishes (Gadidae), eels (Anguillidae), some drums and croakers (Sciaenidae), and perhaps other fishes.

4
Fishes that have special structures mechanically linking the swim bladder to the ear. These fishes are primarily sensitive to sound pressure, although they also detect particle motion. They have a wider frequency range, extending to several kHz and generally show higher sensitivity to sound pressure than fishes in groups 1, 2, or 3. The group includes some of the squirrelfishes (Holocentridae), drums and croakers (Sciaenidae), herrings (Clupeidae) and the large group of otophysan fishes.

5
Eggs and larvae.
(species) thresholds, while different proportions of fishes respond at lower weighted levels. However, there are very few field data derived from wild fishes to support these chosen levels. Also, the concept of dB ht has not been accepted in any independent peer-reviewed publications. Indeed, extreme caution must be exercised in applying the dB ht (species) measure. Defining response criteria applicable to all species is a far too simplistic an approach to evaluating behaviour. Moreover, the data on hearing thresholds used for the dB ht approach should ideally be based on accurate behavioural threshold determinations rather than measures of inner ear responses, as the latter are susceptible to flaws .

| MAJOR RESEARCH GAPS AND RESEARCH RECOMMENDATIONS
In order to develop better guidelines and criteria, it will be critical to fill many gaps in understanding of the potential effects of anthropogenic sounds on fishes. The goal must be to increase knowledge in those areas that are most likely to enable sound exposure criteria to be revised, as per the expectation of the 2014 guidelines . There are many research gaps suggested earlier in this paper and in other publications (Normandeau, 2012a,b;Hawkins et al., 2015). Here we will focus on those questions and data needed to move forward most rapidly.

| Selection of species
It is clear, based on the diversity of fishes and their life styles that it is critical to obtain data from multiple species and a range of sizes and ages of fish within each species. There is likely to be substantial variation in potential effects depending on differences in anatomy, physiology and behavioural responses to various stimuli. Recent guidelines  suggested dividing fishes into several morphological groups that relate to the presence or absence and configuration of the swimbladder (see Tables 2 and 3). Having a representative set of species or fish types will be critical for future work on all aspects of effects of anthropogenic sound on fishes. adversely. Some anthropogenic sounds may frighten the fish away from preferred locales or from migration routes. While many behavioural effects are likely to be minimal and have little or no effect on fish fitness and survival, some behavioural responses may have substantial short and long-term effects upon them.

| Behavioural responses
The currently available data on behavioural responses, as shown earlier, are highly variable and have many problems that do not even start to provide any general principles on how fishes respond to anthropogenic sound. Moreover, there are numerous additional behavioural issues that need to be examined, from the sound levels that are likely to influence hearing (e.g., hearing studies, studies of hearing in the presence of maskers) to responses to sound pressure v. responses to particle motion

| Effects of particle motion
It is now clear that fishes are primarily detectors of particle motion and relatively fewer species of fish use sound pressure. Thus, criteria and guidelines must be developed in terms of particle motion as well as sound pressure. Yet, very little is known about hearing sensitivity to particle motion and it is imperative that such data be obtained.
Concurrently, it is imperative to measure the signal from anthropogenic sources in terms not only of sound pressure, as now done, but also in terms of particle motion.

| Development of dose-response data
Studies on physical effects of pile driving signals in fishes are needed that could lead to understanding dose-response relationships of different sound variables such as signal intensity, number of strikes, inter-strike interval, etc. Indeed, a recent study (Casper et al., 2017) suggests that the dose-response relationship is more complex than previously thought. Studies of dose-response relationships will provide insight not only for understanding the onset of physical effects or behavioural effects, but also for determining those levels above the onset level at which potentially harmful effects start to occur. Such information will enable regulators and others to be able to make better decisions on criteria, particularly if they are willing to accept the idea that just because there is a small effect, this may not affect the fitness of the animal. POPPER AND HAWKINS 707

| Hearing
Though here is a body of literature on the hearing of perhaps 100 fish species (Ladich & Fay, 2013), the greater portion of these data were obtained using sound pressure measures and do not reflect the fact that most fishes primarily detect particle motion. Moreover, most of the studies (particularly recently) used physiological measures (most often auditory evoked potentials; AEP) that do not reflect the sound processing capabilities of the whole auditory system and thus, do not reflect the actual hearing capabilities of an animal (Sisneros & Rogers, 2016). In order to understand fish hearing and the sounds that potential will affect behaviour, future studies must include particle motion and be done using behavioural methods that reflect how fishes actually respond to sound. Moreover, future studies need to be done in acoustic environments where sounds can be fully calibrated, such as in open bodies of water without physical constraints to reflect sounds, or in specially designed (and very expensive) tanks.

| Population studies
In contrast with marine mammals, where populations are small and there is concern for single animals, the greater interest for fishes is with populations of animals. Loss of an individual due to exposure to anthropogenic sound does not have the same implications for a species as does the effect on a population. Effects are the broad range of potentially measurable changes that may be observed in individuals, groups of animals, or even habitats as a result of sound exposure.
Impacts are effects that, with some certainty, rise to the level of deleterious ecological significance (Boehlert & Gill, 2010). Thus, the effect does not indicate the significance, whereas the impact deals with the severity, intensity, or duration of the effect upon animal populations and ecological communities. Such impacts can then be compared with those resulting from other stressors, including chemical pollution, fishing, pathogens, climate change etc. The ecosystem-wide consequences of exposure to sound also need to be evaluated. Effects may influence the dynamics of predation and other types of biotic interactions at the community level. Making assessments across species and communities and within the wider ecosystem, may be of considerable value.

| CONCLUSIONS
There is increasing concern about the effects of anthropogenic sounds upon aquatic animals, including fishes. It is evident, however, that there are major gaps in our understanding of the effects of these sounds and especially their effect upon animal populations and aquatic ecosystems. Much of the literature is limited in quality and many of the experiments have been carried out on captive fishes under laboratory conditions, rather than on free-living fishes in the wild. There is also a lack of information on the responses to particle motion, rather than sound pressure. It is evident that there are so many information gaps that it is almost impossible to come to clear conclusions on the nature and levels of anthropogenic sound that have potential to cause changes in animal behaviour, or even physical harm. There is need to carry out further research on the behavioural responses of a range of fishes to different sound sources, under different conditions. As well as investigating responses to sounds of short duration, information is also required on responses to continuous or repeated exposure. What are the immediate effects of sound exposure and what are the longer-term effects in terms of fitness and likely effect on populations?
At the same time, since there is an immediate need for updated criteria and guidelines on potential effects of anthropogenic sound on fishes, we recommend, as do our colleagues in Sweden (Andersson et al., 2017), that the criteria proposed by Popper et al. (2014) should be used. (We recognise that the suggestion of using the 2014 guidelines is potentially self-serving since we are lead authors on that document. However, as this document is growing in acceptance, we feel it important that we share our own thoughts and that of colleagues world-wide.) However, as new data become available, these criteria need to be updated and filled in. We also suggest that there is significant need to define what onset of effect means in terms of fishes. Is this, as often now used, the start of any effect even on a single animal, or is it some level that, while easily assessed, reflects some statistical value and which focusses on the population rather than on individuals.