Fin ray sensation participates in the generation of normal fin movement in the hovering behavior of the bluegill sunfish (Lepomis macrochirus)

In the unstable, three-dimensionally complex world of many fishes, the pectoral fins play critical roles in movement and posture. Pectoral fins power swimming and often serve as primary propulsors (e.g. Webb, 1973; Drucker and Jensen, 1997; Westneat and Walker, 1997). They function in maneuvering and other transitions in behavior such as swim initiation and braking (Higham et al., 2005). The pectoral fins are particularly important for maintaining posture in response to the body's own instability and to perturbation from the surrounding environment. In hovering behavior, these fins have a focused function in stability and the precise positioning of the body. Fine motor control of complex movements and fin ray bending (Geerlink and Videler, 1987; Lauder and Madden, 2006) suggest a tight association of fin movement with sensory input and raise questions of whether fin-based sensory feedback plays a role in fin movement control and coordination.

In tetrapods, sensation by the limb provides drive for behavior and feedback to hone its constituent movements. During limb-based behaviors, local sensory input has been shown to be critical for normal function. This has been demonstrated in human movements where cases of large fiber neuropathy illustrate how loss of afferent nerve fibers from the limbs greatly impairs multi-joint arm behaviors (e.g. Sanes et al., 1985; Ghez et al., 1990; Gordon et al., 1995; Sainburg et al., 1995), with a significant decrease in movement coordination and accuracy. Functioning with such loss requires other sensory modalities, particularly vision, to provide compensatory information on limb posture and movement (e.g. Rothwell et al., 1982; Sanes et al., 1985).

The role of proprioception in limb rhythms has been debated since early work by Brown and Sherrington (e.g. Brown and Sherrington, 1912; Sherrington, 1913) showed that basic limb rhythms could be generated without proprioceptive input in mammals. Since then, studies have aimed to address the effects of proprioceptive loss. For example, Gray and Lissmann (1940) showed that deafferentation of multiple limbs had a dramatic negative effect on locomotor ability in frogs. More recently, loss of limb afferents has also been shown to disrupt coordinated limb movement in cats (e.g. Abelew et al., 2000) and the early growth response three (Egr3) mouse mutant, for which the phenotype includes proprioceptive loss and, likely associated, ataxia (Tourtellotte and Milbrandt, 1998).

Work on mechanosensation of insect wings may provide the best insight into the role of fin sensation in fish. As with fin ray and membrane structure, wings are not muscular, they bend continuously along a relatively planar structure and beat rhythmically to generate locomotor and stabilizing movements. Mechanosensors located in sensilla on the wing cause afferent spiking during wing deflection (Dickinson, 1990,b,, 1992; Dickerson et al., 2014) and responses reflect activity of rapidly adapting and slowly adapting cells (Dickinson and Palka, 1987; Dickinson, 1990a,b). Afferents spike phasically with the wing beat cycle at typical wing beat frequencies (Dickinson, 1990b). Both of these features, adaptation properties and activity within a movement cycle, also appear to characterize mechanosensation by fish fins (Williams et al., 2013).

In insects, it has also been possible to dissect sensorimotor integration associated with wing mechanosensation. In the blowfly (Calliphora vicina), it is thought that afferent input from wing mechanosensors establishes the phase of motoneuron firing relative to the wing stroke (Fayyazuddin and Dickinson, 1999), and recent work in the hawkmoth (Manduca sexta) suggests that wing sensation can be used to initiate corrective movement reflexes (Dickerson et al., 2014), a function previously attributed to the halteres in the blowfly (Fayyazuddin and Dickinson, 1999). Deafferentation of wing sensilla of locusts results in the loss of early motoneuron activity of the elevator phase of wing movement, disruption and delay of phasic wing activity, and markedly decreased wing beat frequency (Pearson and Wolf, 1987; Wolf and Pearson, 1987). Subsequent work in locust flight ganglia has supported the observation of decreased frequency associated with sensory loss while confirming that the underlying rhythm can be maintained (Stevenson and Kutsch, 1986, 1987, 1988) without ongoing sensory input.

Mechanosensation has been shown to provide feedback to rhythm-generating circuits in diverse organisms. In a number of species, rhythmic movements have been shown to be controlled by central pattern-generating circuits that create the frequency and fundamental phasic structure of movement and are not dependent on sensation or input from higher centers for their ongoing activity after initiation (e.g. Grillner, 1985; Marder and Calabrese, 1996). While not necessary for the generation of the rhythm itself, sensation from mechanoreceptors and proprioceptors serves other functions for central pattern generators, including providing feedback modulation (e.g. Katz and Harris-Warrick, 1990), reinforcing rhythmicity (e.g. Fuchs et al., 2012) and, as discussed above for locusts, contributing to coordination of limb elements (e.g. Grillner and Zangger, 1979) or axial segments (Wen et al., 2012).

While it is as yet unclear how local sensation from fins contributes to the neural control of fin movement and posture, physiological and behavioral data suggest capability. In bluegill sunfish (Lepomis macrochirus Rafinesque 1819), afferent nerves in the fin rays and membrane can convey proprioceptive feedback that reflects fin ray position and velocity at frequencies and bend amplitudes consistent with fin movement in swimming and hovering (Williams et al., 2013). Although not exploring movement or stability, several studies suggest the general importance of fin ray mechanosensation in behavior. Kinematic data have also shown that bluegills touch obstacles as they navigate a complex environment and that there are increased incidents of touch behavior when other sensory modalities are compromised (Flammang and Lauder, 2013). In other species, fin rays have been adapted for foraging, and mechanosensation is proposed to play a key role in that behavior (Bardach and Case, 1965; Silver and Finger, 1984).

Here, we focused on rhythmic fin movement, examining the role of pectoral fin sensation in generating hovering kinematics of the bluegill sunfish. The bluegill's pectoral fins function to stabilize the body, keeping the fish's center of mass balanced atop its center of buoyancy. During hover, the bluegill uses its pectoral fins to produce rhythmic beating that serves to maintain stable body orientation and position in the water column. We used fin ray nerve transection and behavioral experiments to explore how afferent information impacts fin movement. Based on work on mechanosensation in other taxa, we hypothesized that bluegills will continue to beat their fins rhythmically after transection but that the frequency of movement would decrease and the variability of stroke kinematics would increase. These experiments expand our understanding of sensorimotor integration in fin-based movements. They provide comparative data to previous studies on terrestrial vertebrates, invertebrates and insects, helping to fill a major phylogenetic gap in our understanding of limb mechanosensation and sensorimotor integration. In addition, sensation has been shown to be important for the control of robotic fins (e.g. Phelan et al., 2010). Understanding the role of sensation in the control of fin-based movements will inform efforts to generate fin-based propulsors for aquatic environments.

Both before and after fin ray nerve transections, fish hovered or swam slowly well above the tank floor and away from the sides, with the dominant behavior being hover. In control trials, hover uniformly included rhythmic, alternating pectoral fin movement (Fig. 1A), as previously described by Kahn and colleagues (2012). We tested for differences in hover behavior across trials of a given fish and condition. Adaptation of behavior is of particular concern with trials recorded after transection, where experience could affect movement patterns and performance. We found that hovering behavior did not differ significantly in fin beat frequency or amplitude of movement in trials from early in the recording period to those trials recorded later in the period (from 15 min up to 90 min post-recovery). Linear regressions of frequency and excursion against recording time showed near-zero slopes and low R² values (frequency control: −0.03x−1.10, R²=0.186; frequency bilateral transection: −0.08x−1.73, R²=0.194; excursion control: −0.86x−47.34, R²=0.012; excursion bilateral transection: 2.21x−18.86, R²=0.059).

Several aspects of timing and angular movement of the pectoral fins changed with bilateral transection of the fin ray nerves (Fig. 1B). The frequency of pectoral fin beats increased significantly in bilaterally deafferented fish as compared with pre-transection control behavior (Fig. 2A). The mean pectoral fin beat frequency increased from 0.96±0.13 Hz in control groups to 1.31±0.30 Hz after bilateral transection (ANOVA, P<0.001, F=17.75). Within the fin beat cycle, there was no significant change in the relative timing of abduction or adduction phases of movement. The timing of peak abduction for both pectoral fins relative to their respective fin beat cycles remained constant between trials of control and transected behaviors (right fin control: 46.27±5.65%, right fin transected: 45.99±5.39%; Watson–Williams test, P=0.71, F=0.14; left fin control: 47.68±4.57%, left fin transected: 47.07±4.74%; Watson–Williams test, P=0.97, F=0.33). The circular variance of the timing of peak abduction during the pectoral fin's own cycle was 0.063 for control fins and 0.057 for the bilaterally transected fins. Circular variance has a scale of 0 to 1, with lower values indicating tighter clustering about the mean. Comparison of the concentration parameters (a measure of dispersion about the mean) shows that this variability was not significantly different between the two groups (k-test, P=0.62, F=1.10).

After fin ray nerve transections, the pectoral fins exhibited a more splayed posture during rhythmic fin beats (Fig. 1, Fig. 2B–D). There was a decrease in the pectoral fin's overall angular excursion and both peak abduction and adduction differed significantly from controls. The mean angular excursion of the pectoral fins for control and transected trials was 45.67±9.83 and 31.82±13.83 deg, respectively, with the angle of fin excursion being significantly lower after bilateral fin ray nerve transection (ANOVA, P<0.001, F=7.95). The decrease in overall excursion is attributable to a significant decrease in minimum adduction of the pectoral fins after transection (control minimum adduction angle: 12.51±6.71 deg, post-transection minimum adduction angle: 25.82±12.79 deg, ANOVA, P<0.001, F=12.08). In contrast, the peak angle of abduction was not significantly different between control and post-transection trials (control peak abduction angle: 58.21±10.29 deg, post-transection peak abduction angle: 57.29±13.49 deg, ANOVA, P=0.31, F=1.01).

Peak angular speed of the pectoral fins increased after fin ray nerve transection (Fig. 3). Examining angular speed of the fin over time segments of abduction and adduction phases, between 0% and 33%, 33% and 66%, and 66% and 100% of both abduction and adduction phases of movement, showed that, except for 33–66% adduction, when the control fin reached its peak speed, angular speed was significantly higher post-transection (Table 1). Thus, the increase in fin beat frequency observed post-transection was due both to decreased fin beat excursion and to increased angular speed.

To explore how intrinsic movement of the fin changes as a result of deafferentation, we analyzed points along the leading edge of the fin from ventral view images and compared curvature at those points between control and post-transection trials. Although this only provides a two-dimensional perspective, it allows a first look at differences between control and post-transection fin ray bending. Mean curvature values were examined at six time points in the pectoral fin cycle, at 0% (peak adduction), 33% abduction, 66% abduction, 100% (peak) abduction, 33% adduction and 66% adduction phase, six time points in total (Table 2). Examples of typical curvature profiles for a control and post-transection trial are shown in Fig. 4. Comparisons of curvature between control trials and post-transection trials are presented in Table 2. Curvature in the control trials is typified by an increase in leading edge curvature toward the body (negative curvature) during abduction, and an increase in curvature away from the body (positive curvature) during adduction. This can also be seen in Fig. 1A. Notably, there was a consistent decrease in curvature at the base and middle of the fin at 33% of time to peak abduction, and points six to eight along the fin, in the distal end of the middle region of the ray, also regularly showed decreased curvature.

During hover, the left and right pectoral fins demonstrate an alternating gait (Fig. 5A). While alternation overall was maintained, the coordination of left and right pectoral fins changed noticeably after fin ray nerve transection. In the control trials, the pectoral fins were coordinated such that peak abduction of each fin occurred when the opposite side fin was approaching the end of adduction. After bilateral transection, there was a significant shift in the relative timing of the pectoral fins and an asymmetry between the sides. In Fig. 5B the relative timing of peak abduction of the left pectoral fin is expressed in relation to the normalized fin beat cycle of the right fin from the end of adduction of one cycle of the right pectoral fin (0%) to the complete adduction of the right pectoral fin in the next cycle (100%). The relative timing changes from a nearly anti-phase relationship in the control condition, with the average occurrence of left pectoral fin peak abduction at 97.28±5.80% of the cycle of the right pectoral fin, to 92.47±7.22% post-transection (Watson–Williams test, P<0.001, F=24.68). The circular variance of bilateral transection phase data (0.103) was greater than that of the control group (0.067). A statistical comparison of the concentration parameters of the control and bilateral transection groups showed that the circular variance of these groups differed significantly (k-test, P=0.038, F=1.55). An opposing shift was observed in the relative timing of the peak abduction of the right fin relative to the left pectoral fin beat cycle. This pattern of alternating fin beats changes from a near anti-phase in the control condition, with the average occurrence of right pectoral fin peak abduction at 96.07±6.84% of the cycle of the left pectoral fin, to 0.77±6.74% post-transection (Watson–Williams test, P<0.001, F=20.11). The circular variance of the control phase relationship (0.09) was not significantly different from the circular variance after bilateral transection (0.09; k-test, P=0.90, F=1.03). This change introduces an asymmetry into the left–right rhythm and a sidedness to the coordination.

Changes in anal fin movement were also observed as a consequence of transecting the pectoral fin ray sensory nerves. Anal fin frequency increased from 1.05±0.62 Hz to 1.40±0.46 Hz (ANOVA, P=0.01). This allowed the pectoral fins and anal fin to maintain the same coordination after transection as they had before it. The peak of anal fin movements to either side of the body preceded the peak abduction of the pectoral fin on the same side of the body in both control and bilateral transection conditions. Peak movements of the anal fin occurred at 41.24±14.77% of the pectoral fin cycle in control trials and 42.24±14.57% in bilateral transection trials, showing no significant change (Watson–Williams test, P=0.56, F=0.34). Maximum angular displacement of the anal fin was significantly different between the control and bilateral transected groups, with increased amplitude after transection (control: 8.93±3.39 deg, transected: 12.82±5.47 deg; Watson–Williams test, P<0.005, F=9.08). These patterns of frequency and amplitude change and fin coordination can be seen in the image series of Fig. 1.

We did not observe a change in caudal fin movement between control and post-transection trials. Maximum angular displacement of the ventral lobe of the caudal fin did not increase significantly in fish with bilateral transections of the pectoral fin ray sensory nerves (control: 9.57±8.77 deg; transected: 13.74±12.23 deg; ANOVA, P=0.11). Peak movements of the ventral lobe of the caudal fin were variable within the cycle of the fin (control: 94.20±19.53%, transected: 36.52±18.72%). A Watson–Williams test was not performed on the caudal fin data, as the data did not meet the test's criterion of being von Mises distributed.

The frequencies of the pectoral fins and anal fin were compared to determine whether they differed significantly. In the control trials, the mean pectoral fin frequency (0.96±0.13 Hz) and the mean anal fin frequency (1.05±0.62 Hz) were not significantly different (P=0.07, F=3.51). Likewise, post-transection, no significant difference was observed (P=0.16, F=2.07) between the mean pectoral fin beat frequency (1.31±0.30 Hz) and the mean anal fin beat frequency (1.40±0.46 Hz).

Hovering behavior after unilateral nerve transection showed more trial-to-trial variation than that recorded after bilateral fin ray nerve transection. Behaviors ranged from complete adduction of the deafferented fin against the body (Fig. 6) (11/25 trials), to rhythmic movements of the nerve-transected fin (9/25 trials). One fish exhibited infrequent movements of the nerve-transected fin while the unoperated fin beat rhythmically (5/25 trials). Because of the variability and resulting small sample sizes of the classes of responses, we report the data but statistical analysis is limited.

In 11 trials of three fish, the unoperated pectoral fin produced rhythmic fin beats while the deafferented fin remained adducted to the side of the body. The unoperated fin beat at a frequency of 1.48±0.13 Hz, which was similar to the frequency observed in fish with bilateral sensory nerve transections (1.31±0.30 Hz). The average minimum adduction angle for the unoperated fins was 22.52±4.02 deg, greater than in the average observed control trials (12.51±6.71 deg), and less than the average minimum adduction angle in bilaterally transected preparations (25.82±12.79 deg). The maximum abduction angle of the unoperated fin was 60.89±13.84 deg, similar to that of the control trials (58.21±10.29 deg) and the bilateral transection trials (57.29±13.49 deg). Overall excursion of the fin was 38.37±14.93 deg in the unoperated fin. In comparison, total angular excursion was 45.67±9.83 deg for control preparations and 31.82±13.83 deg for bilaterally transected preparations.

Curvature of the leading edge of the fin, as observed from the ventral view, was compared between the unoperated fin and trials of control fish. In those unilateral transection trials where only the non-operated pectoral fin performed rhythmic fin beating while the operated pectoral fin remained adducted to the side of the fish (3 trials/11 fish), leading edge curvature of the non-operated fin was significantly increased compared with trials of the control fish (Fig. 7; Table 3).

In nine trials in two fish, both the unoperated and transected fins produced rhythmic fin beats. The unoperated fin beat at a frequency of 1.40±0.25 Hz, while the deafferented fin beat frequency was 1.29±0.36 Hz, frequencies that were greater than those observed in the control fish trials (0.96±0.13 Hz). The mean minimum adduction angle was 36.51±7.68 deg for the unoperated fin and 27.03±8.11 deg for the deafferented fin, both of which were greater than minimum control adduction angles. The mean maximum abduction angles were 71.71±12.33 and 65.33±11.22 deg for the unoperated and deafferented fin, respectively. Mean total angular excursion was 35.21±9.74 deg for the unoperated fin and 38.30±11.24 deg for the deafferented fin.

Trials of one fish (5 total trials) produced infrequent fin beats with the operated fin during hover trials while the unoperated fin beat rhythmically. The operated fin beat either once or twice for every five beats of the unoperated fin. The unoperated fin beat at an average frequency of 1.67±0.14 Hz, which was greater than control fins. Mean minimum adduction and mean maximum abduction angles for the unoperated fin were 5.61±3.50 and 55.06±12.84 deg, respectively. For the operated fin, the average minimum adduction angle was 10.61±9.71 deg and the average maximum abduction angle was 25.24±18.67 deg. The minimum adduction angles for both the operated and unoperated fins were greater than control angles, while the maximum abduction angle was similar to controls. Total angular excursion was 49.45±12.50 deg for the unoperated fin and 15.04±8.97 deg for the transected fin.

These results indicate that local sensory feedback provides important modulation of pectoral fin movement in fish. Previously, sensation has been shown to provide critical feedback for limb movement of tetrapods and insects. This study adds taxonomic breadth to the body of work on the role of sensation in movement of vertebrate limbs and shows that sensorimotor integration in limb movement is a more general feature of animals than previously known. Although deafferentation removes all modes of sensory feedback from the rays and membrane of the fins, given the known proprioceptive capacity of fin ray nerves in fish (Williams et al., 2013) and the importance of mechanosensation in limb movement of other taxa (e.g. Gray and Lissmann, 1940; Abelew, 2000; Stevenson and Kutsch, 1986, 1987, 1988), we believe that the changes we see primarily result from the loss of mechanosensory feedback. In fish species, work on the evolution of fin structure and function has focused on the musculoskeletal system and motor function. We suggest the co-evolution of fin-based mechanosensation is also critical and necessary for understanding the diverse form and abilities of fins.

In these experiments, we focused on nerves with endings distal to the muscular base of the fins. It is as yet unclear whether more proximal somatosensory structures are also present in the pectoral fins. There are putative tendon organs at the base of the fin rays (Ono, 1979) but these have not been verified physiologically. It is also not known whether fish have muscle proprioceptors that are analogous or homologous to the muscle spindles of tetrapods. A muscle spindle-like structure has been described morphologically in a jaw muscle of the Japanese salmon (Maeda et al., 1983), but it has not been verified or examined physiologically. In interpreting the results of this study, the differences in proprioceptive system and transections are important to consider. As we are examining proprioceptive loss from a non-muscular, distal and un-jointed structure, insect wings may be a better analogy than the muscular and jointed tetrapod limbs for understanding the roles of proprioception in fish fin rays and membranes.

In our experiments we were careful to limit the amount of time over which we captured behavior. Our primary concern was that, with experience, the fish might learn how to compensate for the sensory loss through feedback from other sensory modalities or from proprioception from the more proximal musculoskeletal region of the fins. However, a competing concern was that the fish should be fully recovered from the anesthetic used in the transection protocols and we did not start recording data reported here until the fish demonstrated upright and stable body posture in mid-water. Over the time period used, there was no observable change in behavior, suggesting that we were able to effectively control for these factors. The possibility of learning to compensate for such loss is itself an interesting question. As fins of fishes are often damaged and can regenerate (Morgan, 1900), fish are regularly faced with the need to produce effective movement with a fin that has changed in its sensory and mechanical abilities. This might suggest that they would be unusually capable of making adaptive adjustments to how they weight sensory input in the control of motor output.

Our analysis focused on hover, a postural behavior that shares features with rhythmic fin-based locomotion. Bluegills are inherently unstable with the center of mass located over the center of buoyancy and will flip without active control of posture. Hovering fin movements serve a significant role in this stabilizing function and in that way are similar to postural adjustments in other organisms. The bluegill body can appear to be completely still during hover, a remarkable achievement given that hover is generated by high amplitude, multi-fin movements. In contrast, for terrestrial taxa, stability often results from fine, behaviorally imperceptible adjustments of the limb musculoskeletal system to maintain a particular posture. A characteristic of the hover-associated fin movements of bluegills is that they are composed of left–right alternating pectoral fin movement (Flammang et al., 2013; present study). Unlike hover behavior of insects, where lift generation is imperative, the bluegill has some ability to generate lift with inflation of the swim bladder and it is unclear how much fin movements in hover contribute to this role. In their rhythmicity, the pectoral fin movements of hover are similar to rhythmic central pattern generator-driven locomotion. As with postural control, proprioception has been shown to modulate rhythmic limb movement in locomotion of insects (e.g. Pearson and Wolf, 1987; Wolf and Pearson, 1987; Stevenson and Kutsch, 1986, 1987, 1988) and mammals (e.g. Gray and Lissmann, 1940; Abelew et al., 2000). Because of the shared characteristics of hover with both postural stability and rhythmic movements, we compare our data on proprioceptive loss in both of these contexts.

While fish were able to hover effectively after transection of pectoral fin ray nerves, the pectoral fin movements with which they accomplished hover changed. The frequency of fin's beats and the overall speed of fin movement increased while fin angular excursion decreased. In general when we think about pectoral fin's amplitude differences, they are coincident with the amount of abduction. Here, amplitude change was not a consequence of an increase in abduction angle but resulted from an increased angle of minimum adduction. Thus, throughout the fin beat cycle the fins maintained a more splayed position.

Comparison across studies showed variation in the effects of deafferentation on frequency and amplitude measures. Comparing frequencies is challenging as in some preparations there is such a deficit in post-transection behavior that movements may not be truly rhythmic. Our results on frequency are opposite to what has been observed in insects, where the frequency of the locust wing beat has been shown to decrease with loss of proprioceptive input (Pearson and Wolf, 1987; Wolf and Pearson, 1987; Stevenson and Kutsch, 1986, 1987, 1988). However, comparable changes of increased frequency have been observed in locomotor rhythms in other deafferented vertebrate preparations. For example, experiments with neonatal rats show that the frequency of hindlimb activity during locomotor behavior is significantly increased (Yakhnitsa et al., 1987) after nerve transection. Our result of decreased fin excursion after transection of the fin ray nerves is consistent with reduction of limb excursion during locomotion, after deafferentation in terrestrial vertebrates (e.g. Gray and Lissmann, 1940; Miller et al., 1975; Yakhnitsa et al., 1987; Goldberger, 1988).

We suggest that both the more abducted posture of the fins and the increased fin beat frequency and speed may aid stability following the loss of feedback from the fin. In humans, the size of the base of support of the limbs and relation to the center of mass have been shown to be fundamental to stability and it has been observed that a broad base of support for the legs results in better reaction times to destabilization (e.g. Kerr et al., 1985). We suggest that the more lateral fin beats may be acting in a similar way in the bilaterally transected fin ray nerve condition. While they are not planted as in tetrapod standing, the lateral position may improve the effectiveness of force generation for stability and the responsiveness to perturbation. The three-dimensional fin kinematics of hover have been shown to be complex and variable (Kahn et al., 2012). If the deafferented fins are less able to perform nuanced movements in response to subtle disturbances, such kinematics may provide an alternative approach to making the fish robust to postural instability. Similarly, increased frequency and speed of fin movement in hover may increase force production. Kahn et al. (2012) showed that on a robotic fin, decreasing the duration of the fin stroke and increasing velocity resulted in greater force production over fin beat frequencies comparable to those recorded here. Although hover can be maintained after fin ray nerve transection, such increased frequency and speed would be predicted to make a given kinematic gait more energetically expensive; thus, sensory feedback may improve efficiency, contributing to effective movement generation. Assessment of the orientation and magnitude of forces produced by the pectoral fins before and after transection would be important to address these ideas.

Pectoral fin ray curvature can be generated by passive and active mechanisms. The mechanical properties and movements of the fin and their interaction with the surrounding water will cause passive bending of the rays and membranes during the fin stroke. However, in fish it has also been shown that there is the potential for the fins to be actively curved or stiffened with muscles at the base of the fin rays (McCutchen, 1970; Geerlink and Videler, 1987; Lauder, 2006). The division of the rays into the two hemitrichia allows for differential force to be applied on the two sides of the ray. Depending on the muscle force applied to each hemitrich and the pattern of co-contraction, such activity could cause bending to either face of the fin, stiffening or both. It is as yet unclear how such active control may be implemented in the fin beat cycle and thus it is not possible to dissect the active and passive components of bending. In the comparison of fin ray curvature in hover between control trials and following bilateral fin ray nerve transection, we found that curvature was decreased in both the abduction and adduction phases of movement. This may be purely a consequence of differences in the frequency, speed and/or excursion of the fins and may not reflect additional changes in how the fins are being used. Alternatively, it may be that the curvature of the fin rays is being controlled differently before and after fin ray nerve transection. Assessment of activity of fin ray muscles would be necessary to investigate this further.

In the unilateral transection trials, we examined the curvature of the unoperated fin after transecting the fin ray nerves on the other side. We focused on trials in which the opposite-side fin was tucked against the body during hover. We found changes in curvature compared with the control data and suggest that the pectoral fin movement is changing to compensate for the loss of use of the opposite side fin and the asymmetric generation of forces. It is possible that increased curvature of the fin during adduction decreases force production and moments that would unbalance the fish.

The hover behavior we recorded in control videos was characterized by rhythmic abduction–adduction cycles for each fin and left–right alternation between the fins (Fig. 1). After transection of both left and right fin ray nerves, abduction–adduction fin rhythms were retained, indicating that feedback from the rays and membrane of the fin is not necessary to generate rhythmic movement. However, as feedback from the proximal region of the fin is likely, our experiments do not demonstrate that rhythmic fin movement is generated without proprioceptive feedback from the fins.

While left and right fins remained coordinated from cycle to cycle after transection, the coordination pattern changed from the pre-transection behavior. We observed an asymmetry between the two sides that was not present in the control data such that the fins were slightly out of phase, one following the other. This pattern is similar to, but not as extreme as, the coordination of forelimbs or hindlimbs in a tetrapod gallop. Across the five fish examined, the change consistently involved the left fin cycle following more closely the right, suggesting some sort of asymmetry in the underlying control of left–right coordination.

Little is known of how fish drive coordinated, rhythmic limb movement, which limits our ability to interpret the neural mechanisms behind changes in the fin rhythm with transection. In larval zebrafish, rhythmic fin movement can be generated with a small region of the hindbrain and rostral spinal cord (M.E.H., personal observation), suggesting a local central pattern generator. If this is the case for bluegills, it is possible that the change in the frequency and other parameters of the fin beat rhythm could be a consequence of the decreased sensory feedback on a local rhythm-generating circuit in the spinal cord and/or caudal hindbrain. Alternatively, changes could be actively driven through increased descending drive from more rostral regions of the brain, in response to sensory measures of stability such as vestibular input or vision.

With the unilateral transections, we saw more variable responses in the transected fin. In some cases, fins beat relatively normally but in other cases fin movement was sporadic or the fin did not move at all but remained near the body. When one fin was tucked, the fish was still able to generate hover but the movement of the intact fin was altered compared with control trials. It is possible that the sensory dissonance between the left and right sides challenges the fish's ability to coordinate the two fins. It also clear that the fish can generate hover using one pectoral fin, in combination with median fins; this mode may be more efficient or simpler to control than trying to incorporate the transected fin.

With the bilateral transection of the pectoral fin ray nerves, we observed changes not only in the movement of the pectoral fins but also in the movement of the anal fin. One notable change was an increased amplitude of anal fin movement, suggesting that the anal fin may be helping to compensate for decreased pectoral fin performance. Another was that the beat frequency of the anal fin matched that of the pectoral fins, indicating coordination among their neural drivers. Clear coordination among the fins, including the pectoral fins and anal fin, has previously been shown with steady swimming (Hove, et al., 2001; Arreola and Westneat, 1996) and the changes we see in hover may result from the same neural control mechanisms.

This work demonstrates the importance of the somatosensory roles of the fins of fish, and provides additional evidence that the fins of fish are not merely propulsive surfaces but also contribute to intricate sensorimotor networks. In addition to sensory input from the pectoral fin rays and membranes, sensation from the musculoskeletal base of the pectoral fin, from other fins and from other sensory systems (e.g. lateral line, vestibular and vision) is likely integrated for hovering and locomotor behaviors. Given the relatively rich literature on fin-based swimming movements in fishes, the field seems primed for greater focus on the role of sensation in these behaviors.

This research has implications for the study of sensorimotor systems evolution. The paired fins of actinopterygians (the pectoral and pelvic fins) and the limbs of terrestrial vertebrates are homologous, like vertebrate forelimbs and hindlimbs. Fossil records show fin rays to be a primitive feature of sarcopterygian fins (Coates et al., 2002). Investigating proprioceptive mechanisms in the fin rays of actinopterygian fish may provide insights into the evolution of limb proprioceptive systems. In particular, exploring the homologous proprioceptive systems of fish may provide insights into how the proprioceptive systems of terrestrial vertebrates evolved from aquatic to land-based functionality.

These experiments may also contribute to the design of biologically inspired swimming devices, as engineers seek to utilize feedback from sensors to modulate and optimize swimming movements in such devices. The fins of many actinopterygian species, such as the bluegill, combine maneuverability and efficiency in a hydroacoustically quiet design. In an effort to produce aquatic devices with similar properties, engineers have begun to emulate the propulsive designs present in fishes. Knowledge of how fish neuromechanical systems utilize proprioceptive input to modulate and optimize their movements can provide highly valuable information for improving the performance of human-made aquatic devices by incorporating appropriate sensory feedback mechanisms into their design.

Fish were obtained from Keystone Hatcheries (Richmond, IL, USA) and housed in aquaria at 20–23°C with a standard seasonal light/dark cycle. Ten adult bluegill sunfish (Lepomis macrochirus) were used in these experiments. Fish ranged in size from 12.7 to 15.5 cm total length (14.28±0.90 cm). All experiments and procedures were approved by the University of Chicago's Institutional Animal Care and Use Committee.

Experiments on each fish occurred over 2 days. Control kinematics of hovering were recorded on the first day. On the second day, two sets of kinematic data were recorded. Behavioral trials were recorded after a sham operation, in which the full surgical procedure was performed except for nerve transection, and again after the fin ray nerves were transected. Sham operations and nerve transections were performed bilaterally in five fish [14.12±1.15 cm total length (TL), mean±s.d.] and unilaterally in five fish (14.44±0.49 cm TL). The sham experiments provided important controls for possible effects of the surgery on fin movement and performance. Video recording, described below, was consistent between the control and treatment conditions.

Prior to surgery, fish were anesthetized in a solution of MS-222 (0.25 g l⁻¹) in water. Sham surgeries consisted of making an incision in the skin and the underlying connective membrane on the medial side of the fin to expose the sensory nerves innervating the pectoral fin rays. The same anesthesia procedures were followed in transection procedures but the branches of the sensory nerves exposed by the sham procedure were cut. There was no disruption of muscle during either procedure as these nerves lie superficial to the fin muscle and the skin does not adhere tightly to the structures beneath. Immediately after sham or transection surgery, the fish were moved to the filming tank for recovery.

Behavior was recorded in a tank with working area dimensions of 33.02×33.02×20.32 cm. A mirror oriented at a 45 deg angle beneath the filming tank was used to capture ventral views of movements. A blind around the filming space prevented the fish from seeing our movement during filming. Through a small slit in the blind we observed the fish during recording to ensure that it was positioned well above the bottom of the tank. Behavior was video recorded at 15 min intervals for 2 h. Data used in the analysis were taken from trials that were collected between 15 and 90 min post-surgery. In all cases, fish were oriented upright, maintained their position in the water column and did not exhibit signs of stress such as erratic swimming or rapid respiration prior to behavioral recordings. We limited data to that narrow time frame in order to control for possible changes in behavior with experience following transection. Each recording was 49.15 s in duration. The behavior videos were recorded using a Fastcam APX RS camera (Photron, San Diego, CA, USA) at a frame rate of 125 frames s⁻¹. From these videos, the first five segments containing hovering bouts were selected for analysis. Hovering bouts included at least five cycles of fin beats of at least one pectoral fin while the fish held a stationary position, without visually observable yaw, pitch or roll.

ImageJ 1.48 (NIH) was used to digitize points along the fins and body of the fish. For each fin stroke cycle, seven frames were sampled from minimum adduction (minimum angle of the fin to the body wall caudal to the fin base), to peak abduction (maximum angle of the fin to the body wall caudal to the fin base), back to minimum adduction for both left and right fins. We considered minimum adduction as the point at which the fin reached its closest position to the body as assessed in ventral view. This point is analogous to peak flexion of legged movement. Maximum abduction was the point at which the fin reached its maximum excursion from the body as assessed in ventral view, analogous to peak limb extension. For unilateral nerve transection trials, five complete fin strokes of the fin with intact fin ray nerves were sampled. Each abduction and adduction phase of every fin beat cycle was sampled at 0%, 33%, 66% and 100% of abduction/adduction in order to observe pectoral fin kinematics throughout the course of abduction and adduction. Thirty-three points were digitized from each frame. Eleven equidistant points were collected along the arc length that represents the leading edge of both pectoral fins, from the fin tip to the fin base. Nine of these leading edge points were used for measurements of fin curvature. Fin angle measurements were taken as the angle made between the line segment calculated from the from the tip of the fin to the base of the fin at its rostralmost margin to the line segment calculated from that same point at the base of the fin to the margin of the body at the rostrocaudal level of the pelvic fin base. Additionally, points at the distal tips and bases of the anal and caudal fins were digitized as well as the tip of the snout and a midline point at the same rostrocaudal level as the tip of the anal fin. Kinematic data were analyzed using MATLAB 8.4.0 (MathWorks, Natick, MA, USA).

Several pectoral fin kinematic parameters were examined for each of the hovering bouts: stroke onset; time to maximum abduction angle for each cycle; cycle duration; maximum abduction angle; minimum abduction angle; angular displacement of the fin across cycles; angular speed of the pectoral fins; and leading edge curvature. Leading edge curvature was calculated as k=dT/ds, where T is the unit tangent and s is arc length. With the exception of curvature, similar measurements were made for the non-operated anal and caudal fins. Kinematic measurements were compared between transected behavior trials and measurements taken in control behavior trials. The Shapiro–Wilk W-test was used to test for normality in measurements of non-cyclic fin kinematic parameters, and one-way ANOVA was performed on normally distributed data. The Watson–Williams test was used to determine significant differences in the means of circular data. Statistical comparisons of the distribution of circular data were performed with a k-test (Berens, 2009). A significance level of α=0.05 was applied for all statistical tests and P-values are reported in the text.

We thank Meera Patel, Nishil Patel, David Rodgers and Noah Sawyer for their assistance with digitization. We also thank Drs Sliman Bensmaia, Nicho Hatsopoulos and Mark Westneat for feedback on this research.