Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Neuroscience. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 28 February 2024

The Processing of Hydrodynamic Stimuli With the Fish Lateral Line Systemfree

The Processing of Hydrodynamic Stimuli With the Fish Lateral Line Systemfree

  • Joachim MogdansJoachim MogdansInstitute of Zoology, University of Bonn


All fish have a mechanosensory lateral line system for the detection of hydrodynamic stimuli. It is thus not surprising that the lateral line system is involved in numerous behaviors, including obstacle avoidance, localization of predators and prey, social communication, and orientation in laminar and turbulent flows. The sensory units of the lateral line system are the neuromasts, which occur freestanding on the skin (superficial neuromasts) and within subdermal canals (canal neuromasts). The canals are in contact with the surrounding water through a series of canal pores. Neuromasts consist of a patch of sensory hair cells covered by a gelatinous cupula. Water flow causes cupula motion, which in turn leads to a change in the hair cells’ receptor potentials and a subsequent change in the firing rate of the innervating afferent nerve fibers. These fibers encode velocity, direction, and vorticity of water motions by means of spike trains. They project predominantly to lateral line neurons in the brainstem for further processing of the received hydrodynamic signals. From the brainstem, lateral line information is transferred to the cerebellum and to midbrain and forebrain nuclei, where lateral line information is integrated with information from other sensory modalities to create a three-dimensional image of the hydrodynamic world surrounding the animal.

For fish to determine spatial location and identity of a wave source as well as direction and velocity of water movements, the lateral line system must analyze the various types of hydrodynamic stimuli that fish are exposed to in their natural habitat. Natural hydrodynamic stimuli include oscillatory water motions generated by stationary vibratory sources, such as by small crustaceans; complex water motions produced by animate or inanimate moving objects, such as by swimming fish; bulk water flow in rivers and streams; and water flow containing vortices generated at the edges of objects in a water flow. To uncover the mechanisms that underlie the coding of hydrodynamic information by the lateral line system, neurophysiological experiments have been performed at the level of the primary afferent nerve fibers, but also in the central nervous system, predominantly in the brainstem and midbrain, using sinusoidally vibrating spheres, moving objects, vortex rings, bulk water flow, and Kármán vortex streets as wave sources. Unravelling these mechanisms is fundamental to understanding how the fish brain uses hydrodynamic information to adequately guide behavior.


  • Sensory Systems


Fish comprise by far the largest group of extant vertebrates, and estimates of species number are up to about 34,800 ( July 2022). They exhibit a wide range of shapes and sizes, life histories, and life styles and are found in nearly every aquatic habitat. While most fish have well-developed visual, acoustic, tactile, gustatory, and olfactory systems (von der Emde et al., 2004), some even have an electric sense for the detection of weak electric fields (Caputi, 2017; Crampton, 2019; Nelson, 2011; von der Emde, 2013). In addition, all fish as well as some aquatic amphibians have a mechanosensory lateral line system for the detection and processing of minute water movements (see, e.g., Bleckmann & Mogdans, 2014; Bleckmann & Zelick, 2009; Coombs & Montgomery, 1999; Mogdans & Bleckmann, 2012; Mogdans et al., 2003; Webb, 2014a, 2014b).

The lateral line was initially described as a series of pores along the side and on the head of fish, which lead to underlying subdermal canals. The first recognition of these pores dates back to Stenon in 1664, Lorenzini in 1678, and Rivinius in 1687 (Parker, 1904). Hofer (1908) discovered that the lateral line is sensitive to weak water currents which strike the body wall. Despite this early finding, the lateral line, for a long time thereafter, was believed to be an organ for the detection of low-frequency sounds. By surgical elimination of the lateral lines, von Frisch discovered that they are not needed for sound reception (von Frisch & Stetter, 1932). This was confirmed by Dijkgraaf, who was the first to point out that the lateral line responds to water motions relative to the surface of the fish (Dijkgraaf, 1963). This does not mean, however, that the lateral line has no ability to detect sound. Even though it is now well established that the lateral line serves as a detector for water motions, there is good evidence that it also detects low-frequency sounds, at least close to the sound source (Braun & Coombs, 2000; Braun et al., 2002; Higgs & Radford, 2013; Radford & Mensinger, 2014).

Since the early studies by Dijkgraaf, research has generated enormous progress in all fields related to lateral line function and anatomy, including behavior, neurobiology, biomechanics, theoretical biophysics, and biomimetics. To provide an overview of our current understanding of the lateral line system, this article will briefly introduce various types of hydrodynamic stimuli and describe peripheral anatomy and central connections of the lateral line system. In the main part, the article will review neurophysiological studies that have been conducted to find out how the physical information that is provided by various types of hydrodynamic stimuli is received and represented by the lateral line system and how this information could be used to determine location and identity of a stimulus source.

Hydrodynamic Stimuli

Hydrodynamic stimuli are generated by biotic and abiotic wave sources at the water surface or in midwater. They provide a wealth of information to the lateral line system and are processed by the central nervous system to determine location, distance, nature, motion direction, and velocity of a wave source.

One type of biotic water movements consists of surface waves, which are typically generated by struggling insects fallen into the water or by aquatic animals contacting the water–air interface in order to breathe or feed (Bleckmann, 1988). Fish that feed on the water surface use their lateral line system to detect and localize the origin of surface waves generated by prey (Bleckmann et al., 1989). Biotic midwater stimuli include water movements generated by swimming and breathing invertebrates (Morad et al., 2010; van Duren & Videler, 2003; Yen et al., 2003), suction currents produced by predatory fish (Day et al., 2005), and vortices like those in the wake of swimming fish (Bleckmann et al., 1991; Hanke & Bleckmann, 2004). Other stimuli of biotic origin are changes in the flow fields produced around fish bodies during swimming and gliding. These can be analyzed by the lateral line system to inform the animal about the dimensions and surface structure of nearby objects (see, e.g., Hassan, 1985, 1986; von Campenhausen et al., 1981) and their spatial arrangement (Burt de Perera, 2004a, 2004b). Finally, oscillatory stimuli generated by body vibrations of conspecifics, on the one hand, may provide important communication signals during social behavior (see, e.g., Satou et al., 1991, 1994). On the other hand, these stimuli may be of disadvantage since they could be detected by predators. However, the propagation distance of oscillatory stimuli is limited to the near-field of the source, restricting the detection range of the lateral line system to within a few body lengths of the receiver, thus allowing these stimuli to be detected only at close range (Kalmijn, 1989).

Abiotic water motions are generated by leaves fallen onto the water surface; by water currents, wind, temperature gradients, and salinity gradients; and by gravity. They may be regarded as noise to the lateral line system because they can mask biotic water motions and thus impair the detection of local wave sources, such as prey items (see, e.g., Bassett et al., 2006; Engelmann et al., 2000; Kanter & Coombs, 2003). However, abiotic water motions like running water may also provide information that allows fish to determine the direction and velocity of the surrounding water and to find energetically favorable locations in a stream or creek (Liao, 2007; Liao et al., 2003a; Przybilla et al., 2010).

Design and Function of the Lateral Line Periphery

To understand lateral line function, a description of the sensory receptors is indispensable. The basic sensory units of the lateral line system are the neuromasts (Dijkgraaf, 1952). They occur freestanding on the skin (superficial neuromast, SN) or within fluid-filled canals (canal neuromast, CN). The canal fluid is connected to the water surrounding the fish via canal pores (Figure 1). Typically, there is one CN halfway between two adjacent pores (Coombs et al., 1988).

Figure 1. Schematic drawing of lateral line neuromasts. Upper: Superficial neuromasts. Lower: canal neuromasts. A neuromast consists of oppositely oriented hair cells (green, orange) underneath a gelatinous cupula (blue), which are innervated by afferent nerve fibers.

The sensory epithelium of a neuromast consists of hair cells like those in the auditory and vestibular system (McPherson, 2018). The ciliary bundles of the hair cells consist of stereovilli that are arranged with increasing length and an eccentrically located kinocilium. They are embedded in a gelatinous cupula that couples the hair cells to the surrounding fluid (Figure 1). Water motions actuate the cupula, which results in a shearing of the ciliary hair bundles. This conveys stress to proteinaceous tip links that connect the stereocilia within a hair bundle near their tips, resulting in the opening or closing of mechanically gated ion channels (Holt & Géléoc, 2017). Shearing towards the kinocilium causes opening while shearing away from the kinocilium causes closing of cation channels. This results in a depolarization or hyperpolarization, respectively, of the hair cell’s membrane potential and a subsequent increase or decrease in the firing rate of the innervating nerve fibers (Kroese & van Netten, 1989).

Each neuromast contains two populations of hair cells with antagonistically oriented hair cell bundles (Flock & Wersäll, 1962). Since hair cells are intrinsically directionally sensitive, a deflection of the cupula results in neuronal responses of opposite sign in the two hair cell populations; that is, hair cells of one population are depolarized, and hair cells of the other population are hyperpolarized (Figure 2). This leads to an increase or decrease, respectively, in transmitter release at the synapse to the innervating afferent nerve fibers and consequently to an increase or decrease in discharge rate in the innervating afferent nerve fibers. Since afferent fibers innervate only hair cells with identical hair bundle orientation (Faucherre et al., 2009; Flock & Wersäll, 1962; Görner, 1963), a deflection of the cupula will cause an increase in discharge rate in fibers innervating one hair cell population and a decrease in discharge rate in fibers innervating the other (Figure 2) (Münz, 1985). Neuromasts are also innervated by efferent nerve fibers, through which the brain most likely adjusts neuromast sensitivity (e.g., Flock & Russell, 1976; Roberts & Meredith, 1989). However, the efferent connections of the lateral line are poorly studied.

Figure 2. Schematic drawing of the effect of neuromast stimulation. Water flow (arrows) moves the cupula, resulting in a shearing of the ciliary bundles of the embedded hair cells. Depending on water flow direction, one hair cell population is depolarized and the other is hyperpolarized (compare orange and green colors on left and right sides of the figure). Subsequently, firing rate is increased or decreased, respectively, in the innervating afferent nerve fibers.

Functionally, lateral line neuromasts are velocity detectors with respect to the fluid surrounding the cupula. This is due to the fact that the cupula is driven by the viscous drag forces that are proportional to the velocity of the water (or canal fluid) flowing along the cupula. Neuromasts within canals, however, typically function as acceleration detectors since the velocity of the fluid in the canals is nearly proportional to the first full derivative of the velocity of the water outside the canal (Kalmijn, 1988, 1989). Since the acceleration of water is proportional to the pressure gradients causing the acceleration (Kalmijn, 1989; van Netten & McHenry, 2014), CNs can also be described as pressure gradient detectors (Coombs et al., 1996). In other words, a single CN responds in proportion to the pressure difference between neighboring canal pores.

The design of the peripheral lateral line can vary greatly in different fish species. For instance, number and placement of SNs and/or the number and structure of lateral line canals, the number and size of canal pores, cupula radius, cupula length, and the density of the canal fluid may differ between species (see, e.g., Bleckmann & Münz, 1990; Coombs et al., 1988; Janssen, 2004; Webb, 2014a, 2014b). It has been suggested that differences in lateral line design represent adaptations to the hydrodynamic conditions prevailing in the habitat of different species. However, to date, there has been no convincing evidence for this hypothesis (compare data by Beckmann et al., 2010; Montgomery et al., 1994; Vischer, 1989).

Central Lateral Line Pathways

The sensory information that is received by the neuromasts is processed at all levels in the brain. The ascending lateral line pathway starts with afferent nerve fibers which are innervating the neuromasts. They course to the brain in the anterior, the posterior, and the medial lateral line nerves, which innervate different parts of the lateral line periphery, namely head, trunk, and back, respectively (e.g., Northcutt, 1989; Puzdrowski, 1989; Song & Northcutt, 1991). The fibers terminate predominantly in the brainstem medial octavolateralis nucleus (MON), the first site of central integration of lateral line information (e.g., McCormick & Hernandez, 1996; New et al., 1996). Output neurons in the MON project predominantly to the contralateral midbrain torus semicircularis (TS) (e.g., McCormick & Hernandez, 1996). From there, lateral line information reaches the midbrain optic tectum and sensory areas in the fish forebrain (Striedter, 1991; Wullimann, 1996). There are also descending projections from the telencephalon to the nucleus praeglomerulosus (Wullimann, 1996), from the nucleus praeeminentialis pars ventralis to the MON, and from the posterior eminentia granularis to the MON (McCormick & Hernandez, 1996; Striedter, 1991). However, little is known about the function of these efferent connections. For a more detailed account of the central nervous organization of the lateral line system, see (Wullimann & Grothe, 2014).

Sensory Processing

The sensory information that is received by the neuromasts must be analyzed by the central lateral line system in order to determine the direction and velocity of water movements and to compute spatial location and identity of a wave source. Neurophysiological experiments have been performed at all levels in the ascending pathway, including afferent nerve fibers and neurons in the brainstem MON and the midbrain TS, to find out how lateral line neurons are analyzing hydrodynamic information. Since a single nerve fiber may innervate a single CN or a small number of SN (Münz, 1989), its firing reflects the neuronal activity of the hair cells in one or, at best, a few neuromasts. In contrast, brainstem and midbrain lateral line neurons are integrating the inputs from numerous neuromasts, which may be distributed widely across the animal, and thus their firing behavior reflects higher-order processing. The types of hydrodynamic stimuli that were applied in these studies include oscillatory water motions generated by a stationary dipole source (i.e., a sinusoidally vibrating sphere) or more complex water motions like those produced by an object moving alongside the fish, by vortex rings delivered through a pipette, by bulk water flow, and by turbulent flow.

Processing of Single-Frequency Dipole Stimuli

Numerous studies have investigated the responses of lateral line neurons to single-frequency stimuli generated by a stationary dipole source (i.e., a small sinusoidally vibrating sphere). Dipoles sources generate well-defined and highly reproducible stimuli (Harris & van Bergeijk, 1962; Kalmijn, 1988, 1989). Lateral line afferents respond to a dipole stimulus with a change in their ongoing firing rate (e.g., Mogdans & Bleckmann, 1999). Fibers are sensitive to peak-to-peak water displacements as low as 0.01 μ‎m (at 100 Hz) at the surface of the skin and to frequencies between less than 1 Hz up to at least 200 Hz, and they exhibit phase-coupled action potentials (Sand et al., 1975). Phase coupling is weak at low stimulus amplitudes and increases with increasing stimulus amplitude until it reaches saturation (Figure 3). Firing rate also increases with increasing stimulus amplitude but only at stimulus levels that are about 10 dB greater than those causing significant phase coupling (Mogdans & Bleckmann, 1999). Thus, the amplitude of a dipole stimulus is represented both by the degree of phase coupling and by firing rate (e.g., Coombs et al., 1998; Mogdans & Bleckmann, 1999).

Figure 3. Response of a goldfish afferent nerve fiber to the stimulation of the lateral line system with sinusoidal water motions. A: Original recording. Spike train (red), raster plot (each marker represents one action potential) across five stimulus presentations (black) and stimulus trace (blue) (50 Hz, peak-to-peak vibration amplitude 4 μ‎m) are shown. B: Level-response function. Discharge rate (red line connecting circles re: left-hand axis) and synchronization coefficient R as a measure of the strength of phase coupling (blue line connecting triangles re: right-hand axis) are plotted as function of stimulus level (in dB). An attenuation of −20 dB corresponds to a peak-to-peak sphere displacement of 425 μ‎m.

When stimulated with different frequencies but equal displacement amplitude, two types of afferent fibers can be distinguished based on the slopes of the level-response functions, the frequency of best sensitivity, and the angle of phase coupling (e.g., Engelmann et al., 2000; Kroese & Schellart, 1992; Montgomery & Coombs, 1992; Montgomery et al., 1994; Münz, 1985). In the study by Kroese and Schellart (1992), response magnitude in one type of fiber increased by about 20 dB with a tenfold increase in frequency, best sensitivity was between about 25 Hz and 60 Hz, and action potentials exhibited a phase lead of about 120° with respect to sphere displacement for low frequencies (Figure 4). These values are typical for a velocity detector (Kalmijn, 1989); that is, fibers of this type presumably innervate SNs. In the other type of fibers, response magnitude increased by about 35 dB with a tenfold increase in frequency, best sensitivity was between 70 Hz and 120 Hz, and action potentials exhibited a phase lead with respect to sphere displacement of 180° (Kroese & Schellart, 1992). These values are characteristic of an acceleration detector (Kalmijn, 1989); that is, fibers of this type most likely innervate CNs.

Figure 4. Frequency response functions of trout afferent nerve fibers obtained by stimulating the lateral line system with sinusoidal water motions. A: Data from four fibers innervating superficial neuromasts. B: Data from four fibers innervating canal neuromasts. Different colors refer to different fibers. Identical colors in A and B do not refer to the same fiber. Gain is given in arbitrary decibels. For clarity, some of the curves are shifted vertically. Phase is given in degrees with respect to displacement. Phase values measured below −180° were shifted up by 360°.

Redrawn and modified from Kroese and Schellart (1992).

In most studies, the temporal response patterns of afferent nerve fibers have been described as sustained with little or no adaptation to sustained sinusoidal stimulation, thus representing stimulus duration (Coombs et al., 1998; Mogdans & Bleckmann, 1999). One study, however, showed that afferent nerve fibers of the lateral line, just like afferent nerve fibers in the auditory system of vertebrates (fishes: e.g., Coombs & Fay, 1987; Fay, 1985; Kuno, 1983; frogs: e.g., Megela & Capranica, 1981, 1982; mammals: e.g., Kiang et al., 1965; Westermann & Smith, 1984), can exhibit adaptive responses to sine-wave stimuli (Mogdans et al., 2017). Adaptation increases with increasing stimulus level and frequency and with increasing rise time of the stimulus envelope (Mogdans et al., 2017). This suggests that adaptation is a fundamental property implemented in the periphery of all hair cell–based sensory systems.

In contrast to afferent nerve fibers, the responses to sine-wave stimuli of brainstem neurons in the MON were characterized by greater degrees of adaptation and greater heterogeneity in terms of response patterns and phase coupling (e.g., Coombs et al., 1998; Künzel et al., 2011; Mogdans & Kröther, 2001). For instance, the temporal patterns of the responses of brainstem neurons range from primary-like sustained or adapting discharges to patterns that are considerably altered relative to the afferent input, like long-latency responses that take considerable time to reach a maximum, responses with periods of intermittent excitation and inhibition (Coombs et al., 1998), and responses that are characterized by a suppression of ongoing activity (Kröther et al., 2002). The meaning of the various types of central responses patters still needs to be disclosed.

Most notably, however, brainstem units are less sensitive to sine-wave stimuli than primary afferents (e.g., Caird, 1978; Coombs et al., 1998; Montgomery et al., 1996; Paul & Roberts, 1977; Wubbels et al., 1993). For instance, in goldfish, about 30% of the units in the MON do not respond to a stationary vibrating sphere, even when displacement amplitudes of up to 800 μ‎m are used (Mogdans & Goenechea, 2000). Such displacement amplitudes are substantially greater than those causing rate saturation in lateral line afferents. However, many of these seemingly insensitive units readily respond to the water motions generated by a non-vibrating sphere moving alongside the fish (Mogdans & Goenechea, 2000) (Figure 5). Similar data were obtained from units in the midbrain TS, where the vast majority (83%) of those units that responded reliably to a moving sphere did not respond to a stationary sphere vibrating with 160 μ‎m displacement amplitude (Plachta et al., 2003). These findings suggest that central lateral line neurons are not particularly well adapted for the analysis of sinusoidal water motions generated by a stationary wave source.

Figure 5. Responses of lateral line units in the medial octavolateral nucleus of goldfish to the water motions generated by a sphere moving alongside the fish (left) and a stationary, sinusoidally vibrating sphere (right). Each graph shows a dot raster plot (black) of the responses to 10 stimulus repetitions and the corresponding peri-stimulus time histogram (red). A: Example of a unit that responded to the moving but not the vibrating sphere. B: Example of a unit that responded to both the moving and the vibrating sphere. Speed of moving sphere was 8 cm s−1, displacement amplitude of sphere vibration was 270 μ‎m, and vibration frequency was 50 Hz.

Processing of Spatial Information

One of the fundamental tasks of a sensory system is to determine the spatial location of a stimulus source. Modelling data and neurophysiological recordings of peripheral lateral line responses to dipole stimuli originating from discrete spatial locations alongside the fish have shown that an array of neuromasts can encode amplitude, location, and vibration direction of a dipole source (Coombs & Conley, 1997; Coombs et al., 1996, 1998; Ćurčić-Blake & van Netten, 2006; Goulet et al., 2008). A vibrating sphere generates a distinct pressure gradient pattern or velocity field, the pattern of which changes in a predictable way with location (azimuth and elevation), distance, and orientation (vibration axis) of the sphere with respect to a given location on the fish surface (Figure 6). Size and shape of the receptive field of a single afferent nerve fiber (i.e., discharge rate, strength of phase coupling and phase angle) change exactly as predicted from the location, distance, and vibration axis of the sphere. In other words, the pressure gradient pattern or velocity field generated by a stationary dipole source is encoded by the pattern of excitation across an array of linearly arranged neuromasts (Coombs et al., 1996, 1998; Ćurčić-Blake & van Netten, 2006; Goulet et al., 2008).

Figure 6. Predicted and measured spatial excitation patterns of lateral line nerve fibers in response to a sphere vibrating sinusoidally either parallel (0°) or orthogonal (90°) to the fish. Upper: Modelled pressure gradient patterns across a linear array of lateral line receptors and resulting predictions for excitation patterns. Lower: Measured excitation pattern of a goldfish lateral nerve fiber. Red: Predicted and measured discharge rates. Blue: Predicted and measured phase angles. Insets indicate direction of sphere vibration and corresponding iso-pressure contours.

Modified from Künzel et al. (2011).

In terms of receptive field shape and size, only a proportion of the units on the MON and the TS are primary-like, exhibiting receptive fields that are in agreement with predictions based on the spatial excitation pattern generated by a stationary dipole source (Figure 7). Most central units have receptive fields that are quite different from those of afferent nerve fibers and are thus more difficult to relate to the stimulus field (Coombs et al., 1998; Künzel et al., 2011; Meyer et al., 2012; Mogdans & Kröther, 2001). For example, many MON and TS units have fairly broad receptive fields (Figure 7), extending across much of the fish surface, which is indicative of integration across large portions of the lateral line periphery. Moreover, receptive fields of central units may consist of excitatory and/or inhibitory regions, a finding that is consistent with lateral inhibition underlying the shaping of these fields. Yet other units may have receptive fields that consist of two or more areas from which stimulation with the vibrating sphere causes an increase or a decrease in discharge rate.

Figure 7. Measured spatial excitation patterns of lateral line neurons in the brainstem medial octavolateral nucleus of goldfish in response to a sphere vibrating sinusoidally either parallel (0°) or orthogonal (90°) to the fish. Upper: Example of a primary-like excitation pattern (see lower part of Figure 6). Lower: Example of a non-primary-like excitation pattern. Red: Discharge rates. Blue: Phase angles. Insets indicate direction of sphere vibration and corresponding iso-pressure contours.

Modified from Künzel et al. (2011).

In summary, most central lateral line units do not encode location and vibration direction of a stationary vibrating sphere. If this was the case, receptive fields should be small and tuned to a particular sphere location, distance, and/or motion direction. Exactly how and where in the brain spatial information is represented by the lateral line system is presently unknown. Perhaps this information is represented by a population code (i.e., by the joint activities of a number of neurons within a given brain nucleus).

Processing of Complex Hydrodynamic Stimuli

Most hydrodynamic stimuli are more complex than the single-frequency stimuli generated by a dipole source. For instance, a small object that is moved alongside the fish creates changes in water velocity consisting of an initial short transient and predictable component followed by an ill-defined long-lasting wake (Mogdans & Bleckmann, 1998). These changes are reflected in the discharges of afferent nerve fibers, which exhibit a characteristic succession of increased and decreased firing rates (Mogdans & Bleckmann, 1998; Mogdans & Geisen, 2009). As predicted from the intrinsic directional sensitivity of the hair cells within a neuromast, the order in which discharge rates of afferent fibers increase and decrease inverses when the direction of object motion is reversed (Figure 8). In addition to this transient response, many fibers discharge bursts of spikes even after an object has passed the fish, presumably in response to the wake generated by the object (Mogdans & Bleckmann, 1998). Afferents of this type most likely innervate SNs. In contrast, a smaller number of fibers do not respond after the object has passed the fish. These fibers presumably innervate CNs, which do not respond to the low-frequency, low-amplitude water motions generated by the object’s wake since they are filtered out by the lateral line canals (Denton & Gray, 1989; Engelmann et al., 2000).

Figure 8. Responses of a goldfish posterior lateral line nerve fiber to an object passing the fish laterally with a speed of 15 cm s−1. The graph shows peri-stimulus-time histograms (bin width 20 ms) of the responses to 10 stimulus presentations. Upper: Motion direction from anterior to posterior (AP). Lower: Motion direction from posterior to anterior (PA). The fish symbol indicates size, location, and orientation of the fish relative to the orbit of the moving object. Note that discharge patterns inversed when motion direction was reversed.

Central lateral line neurons also respond readily to the water motions generated by a moving object (Mogdans et al., 1997; Wojtenek et al., 1998). However, the responses are much more variable than those observed among afferent nerve fibers. Most MON units exhibit a single peak of increased discharge rate when the object passes the fish, others exhibit two response peaks separated by reduced spike activity, yet others show multiple response peaks, and in some units ongoing activity is even suppressed when an object passes the fish. Moreover, most units in the MON do not show any obvious responses to the water motions generated in the wake of a moving object. In addition, many responses are independent of the direction of object motion both in terms of discharge pattern and in terms of discharge rate (Mogdans et al., 1997).

In the midbrain TS, the number of units that respond with a different pattern or with a different rate to reversed directions of object motion is greater than in the MON (Wojtenek et al., 1998). While some TS units may respond with a transient discharge to only one motion direction but not the opposite, other units may exhibit complex temporal discharges, the patterns of which depend on the direction of object motion. Thus, these units clearly encode object motion direction but are not tuned to other aspects of a moving stimulus source, like speed or distance. Nonetheless, the finding that distinct response types in the TS are not found in the MON is indicative of neural computation between brainstem and midbrain. The causes and mechanisms underlying these computations are, however, not known.

Plachta et al. (2003) found TS units which responded to the passing of a sphere and to the wake caused by that sphere with increased spike rates, while other TS units responded to the passing sphere but not to the wake behind it. It is possible that units of the first type are processing input from SNs but that units of the second type are processing input from CNs. The same study suggested that the position of the moving sphere is represented systematically in the TS (Figure 9). That is, when an object was moved from anterior to posterior along the side of the fish, units in the anterior torus respond when the object passes anterior locations (e.g., near the head of the fish). As the object continues to move along the fish, units located at increasingly caudal positions in the torus respond to increasingly posterior locations stimulated by the object. Thus, object location is represented in a neural map in the TS. There is no evidence that other aspects of a lateral line stimulus, like speed or distance, are represented in neural maps in the TS or any other area in the fish brain studied so far.

Figure 9. Responses to a moving sphere of two neurons in the midbrain torus semicircularis of goldfish recorded simultaneously with electrode 2 (E2, red) and electrode 3 (E3, blue). Upper left: Responses to sphere movements in the anterior-to-posterior direction. Upper right: Responses to sphere movements in the posterior-to-anterior direction. The time axis was adjusted to match the distances travelled by the sphere. Lower: The positions of the electrodes relative to the fish midbrain surface. Red, E2; blue, E3. Colored circles below the fish drawing show sphere positions at times labeled 1, 2, 3, and 4 in the upper graphs. Arrows indicate direction of sphere motion.

Redrawn and modified from Plachta et al. (2003).

Vortex rings delivered from a pipette and passing the fish laterally have also been used to study the representation of complex hydrodynamic stimuli by the lateral line (Chagnaud et al., 2006). Similar to moving objects, vortex rings produce defined changes in water velocity and in water flow direction followed by less defined water movements. This is mirrored in the responses of afferent nerve fibers, which respond to vortex rings, just as they do to a moving object, with an initial strong and highly reproducible component consisting of either an increase or a decrease in discharge rate and a second weaker and less reproducible component. Again, these data can be explained by the hair cells’ intrinsic directional sensitivity. The responses of central neurons to vortex rings have thus far not been investigated.

Processing of Bulk Water Flow Information

Fish that are actively swimming or residing stationary in running water are constantly exposed to a bulk water flow across their body. In laboratory experiments in which fish are exposed to bulk water flow in a flow channel, two types of afferent nerve fibers can be distinguished: type I fibers, which respond to the flow with an increase in discharge rate for as long as the water flow was maintained, and type II fibers, which do not change their discharge rate, at least not up to 10 cm s−1 flow velocity (Engelmann, Hanke, & Bleckmann, 2002; Engelmann et al., 2000). The continuous firing of the first type of fibers suggests innervation of SNs, which are continuously stimulated by background flow. In contrast, fibers of the second type most likely innervate CNs, which are unresponsive to background flow due to the filter characteristics of the lateral line canals.

The functional difference between SNs and CNs is revealed if the lateral line is stimulated with sinusoidal water motions in still and running water. In still water, both flow-sensitive and flow-insensitive fibers exhibit sustained and phase-locked responses that can hardly be distinguished from each other. In a 10 cm s−1 background water flow, however, the responses of flow-sensitive fibers to a sinusoidal water motion are masked, whereas the responses of flow-insensitive fibers are hardly affected by the flow (Engelmann, Hanke, & Bleckmann, 2002; Engelmann et al., 2000). This demonstrates that fibers innervating SNs, which are constantly stimulated by the water flow, respond with high accuracy to a vibrating sphere only if the fish is not exposed to a background water flow. In contrast, fibers innervating CNs respond equally well to a vibrating sphere in still and running water.

In running water, many neurons in the lateral line brainstem exhibit firing behaviors that are similar to those of primary afferents. Again, flow-sensitive and flow-insensitive neurons can be found. However, in contrast to primary afferents, some flow-sensitive brainstem neurons may exhibit increased ongoing discharge rates, while others may exhibit suppressed ongoing rates. In some brainstem neurons, discharge rates may increase or decrease transiently when a water flow is turned on but thereafter decrease or increase again while the water flow is continued; that is, these fibers adapt to an ongoing water flow (Kröther et al., 2002). It is likely that flow-sensitive MON neurons receive input predominantly from fibers innervating SNs, and flow-insensitive MON neurons predominantly from fibers innervating CNs.

If stimulated with sinusoidal water motions produced by a vibrating sphere in a water flow, at least three types of MON neurons can be distinguished (Figure 10). Type MI neurons have properties similar to type I afferents; that is, they are flow-sensitive and their responses to a vibrating sphere are masked by running water. In contrast, type MII neurons have properties similar to type II afferents; that is, they are insensitive to water flow and their responses to a vibrating sphere are not masked by running water. Finally, type MII neurons are, like MII neurons, flow-insensitive. Nonetheless, their responses to a vibrating sphere are, just like those of type MI neurons, masked by running water either in terms of spike rate or in terms of phase coupling. These findings support the idea that type input from superficial and canal neuromasts is maintained to a large degree separate in the lateral line brainstem but that interactions between these two subsystems may also occur (Engelmann, Kröther, et al., 2002).

Figure 10. Responses of type MI, MII, and MIII neurons of in the brainstem medial octavolateral nucleus of goldfish to a 50 Hz vibrating sphere stimulus in still (left) and running 15.5 cm s−1 (right) water. Raster plots of the responses to ten stimulus presentations are shown. For details, see text.

Redrawn and modified from Kröther et al. (2002).

Neurons in the TS may also be either flow-sensitive or flow-insensitive, again pointing to the two input channels arising from the superficial and canal neuromast systems (Bleckmann et al., 2004). Surprisingly, responses to dipole stimuli not only of flow-sensitive but also of nearly all recorded flow-insensitive toral neurons were masked in running water. To account for this finding, neuronal interaction between the two submodalities, such as excitatory input from CNs and additional inhibitory input from SNs, have been proposed (Bleckmann et al., 2004). This argues for a discontinuation of the strict separation between superficial and canal neuromast information at the level of the midbrain.

In fish, the optic tectum is the most prominent midbrain structure, receiving sensory input from the visual, auditory, and lateral line system (e.g., Bartels et al., 1990; Echteler, 1985; Lowe, 1987; McCormick & Hernandez, 1996; Vanegas, 1983). While there is evidence for multimodal integration in the optic tectum from studies in amphibians (Bartels et al., 1990; Bastian, 1982), this has not been systematically investigated in fishes. Using calcium imaging, neuron ensembles responsive to water flow have been found in the optic tectum (Thompson et al., 2016). Exactly how tectal neurons are processing hydrodynamic information and how this information is integrated with other sensory modalities still need to be uncovered.

Determination of Bulk Water Flow Velocity

Flow-sensitive primary afferents respond to a water flow exclusively with an increase in discharge rate (Engelmann et al., 2000; Voigt et al., 2000). This is surprising since lateral line neuromasts contain two populations of hair cells with opposite orientation that are innervated by distinct fibers. Therefore, one would expect about equal numbers of the fibers that respond to water flow with an increase or a decrease in discharge rate, respectively. In contrast to this expectation, Chagnaud et al. (2008) showed that fibers increased discharge rate in response to both anterior–posterior and posterior–anterior water flow; that is, lateral line afferents do not encode the direction of a background flow by spike rate. Characterization of the water motions in the flow using particle image velocimetry (PIV) revealed the presence of flow fluctuations that increased with increasing flow velocity. Maximal spectral amplitudes of the flow fluctuations were similar to those of the firing rates of lateral line afferents, suggesting that the afferents mainly respond to the flow fluctuations but not to the direct current component of the flow.

The fluctuations that are contained in a background flow move downstream across the surface of the fish and will be sensed by consecutively arranged neuromasts. In fact, neurophysiological recordings from pairs of primary lateral line afferents showed that responses of many fiber pairs were highly correlated and that the time shift of the cross-correlation peak decreases with increasing flow velocity for flow velocities between 6.5 and 13 cm s−1 (Chagnaud et al., 2008). This spawned the notion of a cross-correlation mechanism to determine bulk flow velocity by comparing the inputs from an upstream neuromast with those of a downstream neuromast. However, the conduction velocities of primary afferents do not allow time delays that are sufficiently large to compensate for the time delay between the arrivals of a flow fluctuation at successive neuromasts. In any case, if such a mechanism was realized, central lateral line neurons should be found that are tuned to distinct flow velocities. However, such neurons have not yet been found.

Processing of Kármán Vortex Street Information

Any obstacle in a bulk water flow causes the shedding of vortices at the obstacles’ edges. Rheophilic fish (i.e., fish that prefer to live in fast-moving water) use flow regions that are disturbed by such vortices to hold station behind obstacles. If a cylinder or a half-cylinder of defined dimensions is placed in a flow, a Kármán vortex street develops, which consists of a defined staggered array of discrete columnar vortices which detach periodically with alternating sign from either side of the cylinder (Vogel, 1983). It has been shown that fish may use the upstream water motions in a Kármán vortex street to conserve locomotor energy (Bleckmann et al., 2012; Liao, 2007; Liao et al., 2003a; Przybilla et al., 2010). Vortex shedding frequency and wavelength (i.e., the spacing between successive vortices in a Kármán vortex street) can be changed systematically by changing flow velocity and/or cylinder diameter (Liao et al., 2003b). Thus, vortex shedding frequency and wavelength provide information about location, size, and form of the object that caused the vortex street.

Flow-sensitive afferent nerve fibers increase their discharge rates in running water (Chagnaud et al., 2008; Engelmann et al., 2000; Voigt et al., 2000). However, the average discharge rates in running water with and without Kármán vortex street do not necessarily differ. Rather, when exposed to a Kármán vortex street, the temporal pattern of the discharge follows the vortex shedding frequency, in particular if the fish is located at the edge of the vortex street (Chagnaud et al., 2007). The precise pattern of firing changes in a predictable way if cylinder position, cylinder size, or flow velocity is altered, thus providing to the brain the information that is needed to estimate location and 3D structure of the cylinder.

Like afferent nerve fibers, MON units may discharge bursts of spikes in a Kármán vortex street (Klein et al., 2015; Winkelnkemper et al., 2018), and the burst pattern matches the vortex shedding frequency (Figure 11). The representation of the vortex shedding frequency may be even better in the MON than in primary lateral line afferents (Bleckmann et al., 2012); that is, central mechanisms lead to an improved signal-to-noise ratio. It is, however, not known how this is achieved.

Figure 11. Representation of the vortex shedding frequency by central lateral line neurons. A: Schematic drawing of the vortices generated behind a half-cylinder in a water flow, passing the fish from anterior to posterior. Note the alternating sense of rotation of successive vortices. B: Original recording of a medial octavolateral nucleus neuron in the nase (Chondrostoma nasus) with the fish exposed to a Kármán vortex street generated behind a half-cylinder in a water flow. C: Probability density function of the spikes in the spike train shown in B. D: Cross-correlation function of the function shown in C. Note that periodic peaks occur at temporal separations corresponding to the vortex shedding frequency (arrowheads).

Furthermore, in a Kármán vortex street, the cross-correlation of the information provided by two consecutively arranged neuromasts about sequentially received vortices could be used to determine bulk flow velocity (Chagnaud et al., 2008). Cross-correlation of temporally distant events like these requires the use of neural delay lines, similar to those found in the auditory system of birds (Carr, 1993; Carr & Konishi, 1988). If delay lines were implemented in the fish lateral line system, flow-sensitive central units should be found, which respond exclusively to a head-to-tail flow, are tuned to a particular bulk flow velocity, and are insensitive to flow fluctuations. Some MON and TS units may indeed be highly sensitive to the direction of water flow. However, neurons tuned to a distinct flow velocity have not yet been found. Thus, at present, there is no physiological evidence for a cross-correlation mechanism in the lateral line system, and the question of how the lateral line system computes flow velocity remains unsolved.

Contribution of Efferent Connections to Sensory Processing

The hair cells of the lateral line neuromasts are innervated by efferent nerve fibers (Hama, 1978), which can be distinguished morphologically from afferent nerve fibers by their smaller axon diameters (Münz, 1985; Northcutt, 1992). A single efferent fiber may innervate all hair cells within a neuromast independent of their polarity (Faucherre et al., 2009). In addition, an efferent fiber may innervate more than one neuromast (Köppl, 2011), hair cells in both SNs and CNs (Münz, 1985), on different sides of the body (Roberts & Meredith, 1989), and even in other modalities (e.g., the vestibular and auditory systems) (Münz, 1985).

Efferent fiber activity typically reduces both spontaneous and evoked discharges of afferent nerve fibers (Lunsford et al., 2019; Russell, 1971; Tricas & Highstein, 1991). There are, however, some studies in which efferent fiber activity caused increased discharge rates of vestibular (Highstein & Baker, 1985) and lateral line afferents (Flock & Russell, 1973). The efferent system can be activated by stimulation of the visual (Tricas & Highstein, 1990) or somatosensory (Roberts & Russell, 1972) system or by motor acts like vocalizations or gill and/or swimming movements, which may cause self-stimulation of the lateral line receptors (Ayali et al., 2009; Palmer et al., 2005; Roberts & Russell, 1972; Russell & Roberts, 1974; Tricas & Highstein, 1991; Weeg et al., 2005). In summary, the present evidence suggests that the efferent system modulates the “gain” of the sensory hair cells to prevent them being overstimulated and exhausted by the water disturbances resulting from powerful body movements or to suppress or enhance afferent nerve fiber sensitivity depending on behavioral context.

In addition to the efferent innervation originating in the octavolateral efferent nucleus, the peripheral lateral line system also receives efferent input from the diencephalon in zebrafish (Bricaud et al., 2001; Metcalfe et al., 1985), goldfish (Puzdrowski, 1989; Zottoli & Horne, 1983), and catfish (New & Singh, 1994). The function of these inputs is not known.


Fish detect water motions with hundreds, in some species thousands, of lateral line neuromasts distributed across the body and within subdermal canals. Morphological properties of the peripheral lateral line like shape, size, and number of neuromasts and lateral line canals cause the received sensory stimuli to be filtered in terms of space and time (frequency). In other words, there is substantial preprocessing of hydrodynamic information already in the lateral line periphery. Consequently, the peripheral nerve fibers provide the fish brain with information about water flow direction, spatial location of stationary wave sources, and an improved signal-to-noise ratio.

Within the brain, the various input channels arising from the lateral line periphery are, at least in part, kept separate. For instance, information about opposite water flow direction, which is resulting from the separate innervation of oppositely oriented hair cell populations within single neuromasts, is maintained in the brainstem and midbrain and may even be enhanced as suggested by the different patterns of discharges of central neurons to different water flow directions. Similarly, inputs from SNs and CNs are also at least partially kept separate in the brainstem, allowing the brain to analyze information arising from these two subsystems independently. On the other hand, the large receptive fields of many brainstem and midbrain neurons clearly demonstrate that there is also substantial convergence of inputs from large portions of the fish surface. There may also be integration across inputs from the SN and CN system, but this has not been demonstrated unequivocally.

While the spatial location of a stationary vibrating source is unambiguously represented by the peripheral lateral line, only some central neurons exhibit primary-like excitation patterns in response to vibratory stimuli. Most central lateral line units have spatial responses that are completely different from those of afferent nerve fibers and thus do not encode the spatial location of a wave source in the same way. Therefore, space coding by the central lateral line must involve other mechanisms and may be achieved, for example, by a population code.

One apparent feature of the lateral line system is that central neurons are more responsive to complex water motions generated, for instance, by a moving source as compared with rather simple oscillatory water motions generated by a stationary dipole. Presumably, this is due to fact that many central neurons intergrate across multiple inputs, thereby abandoning information about local sources. This argues for the importance of large-scale water motions for the lateral line system, which stimulate different parts of the sensory surface, possibly even in different ways. In this case, the task of the central lateral line system is to analyze the information arising from different areas of the sensory periphery and bind them together to create a coherent image of the sensory stimulus which in turn allows conclusions about the stimulus source.

Finally, primary afferent nerve fibers clearly respond to bulk water flow and represent information about Kármán vortex streets (e.g., vortex shedding frequency). In the lateral line brainstem, at least a proportion of the neurons also represent vortex shedding frequency in their discharges. Responses of higher-order neurons to vortex streets have thus far not been investigated. In addition, it is not known whether central brain circuits use the information about Kármán vortex streets provided by the lateral line periphery to calculate water flow velocity or the locations and structural differences between the objects that are generating different types of vortex streets.

In summary, there is already substantial knowledge on how the peripheral lateral line represents various types of hydrodynamic stimuli and how central lateral line neurons respond to these stimuli. We are, however, just beginning to uncover the mechanisms that underlie the coding of hydrodynamic information by central lateral line neurons. The unravelling of these mechanisms is fundamental to understand how the fish brain uses hydrodynamic information to adequately guide behavior.


The original work of the authors was supported by the DFG, DARPA, BMBF, and the EU.