State-dependent sensorimotor processing: gaze and posture stability during simulated flight in birds

Abstract

Vestibular responses play an important role in maintaining gaze and posture stability during rotational motion. Previous studies suggest that these responses are state dependent, their expression varying with the environmental and locomotor conditions of the animal. In this study, we simulated an ethologically relevant state in the laboratory to study state-dependent vestibular responses in birds. We used frontal airflow to simulate gliding flight and measured pigeons′ eye, head, and tail responses to rotational motion in darkness, under both head-fixed and head-free conditions. We show that both eye and head response gains are significantly higher during flight, thus enhancing gaze and head-in-space stability. We also characterize state-specific tail responses to pitch and roll rotation that would help to maintain body-in-space orientation during flight. These results demonstrate that vestibular sensorimotor processing is not fixed but depends instead on the animal's behavioral state.

sensorimotor circuits adjust to changes in the relationship between a specific pattern of sensory input and the optimal motor output. Even highly stereotyped motor reflexes must be able to alter the input-output mapping when appropriate. One well-studied phenomenon that can adjust reflex output is adaptation. In this case, the goal of the reflex does not change, but the relationship between the physical sensory world and the motor plant does change. This necessitates a compensatory adjustment in how the motor plant responds to a given sensory stimulus, driven primarily by error detection from sensory feedback. For example, the vestibuloocular (VOR) and vestibulocollic (VCR) reflexes participate in gaze stabilization during head and body motion, by driving compensatory eye and head responses to minimize movement of the visual image on the retina. Experimental manipulations of the relationship between head movement and visual feedback induce adaptive changes in the amplitude and timing of the VOR (Blazquez et al. 2004; Boyden et al. 2004; Schubert and Zee 2010). Adaptation can also adjust the amplitude of the VCR in response to changes in head inertia (Goldberg 1999; Reynolds et al. 2008; Reynolds and Gdowski 2008). However, some changes in reflex input-output relations are driven not by sensory feedback or detection of errors but by discrete changes in environmental and physiological conditions, which may change the goal of the reflex or the way in which it should be performed. Over time, the animal learns to associate recurring conditions (and the corresponding sensorimotor mapping) with specific combinations of sensory and motor cues. Thus these cues may be used to retrieve the appropriate input-output mapping immediately upon detection of a change in conditions. We refer to this phenomenon as state-dependent sensorimotor processing.

Gaze and posture stabilizing responses can be strongly affected by factors related to behavioral state, such as mental set (Baloh et al. 1984; Barr et al. 1976; Keshner 2000) and locomotion context (Xiang et al. 2008). In fact, these responses can even be gated by behavioral state, as when the VOR is suppressed during voluntary gaze shifts or pursuit of a visual target (Belton and McCrea 2000a, 2000b; Lisberger 1990; Newlands et al. 2001). There is strong evidence in birds that gaze and postural responses are highly state dependent. For example, the dynamic range of the optocollic head response (OCR) to visual motion is expanded into higher stimulus velocities during flight (Bilo 1992, 1994; Gioanni and Maurice 2004; Gioanni and Sansonetti 1999; Maurice et al. 2006), suggesting a state-dependent enhancement of head-in-space stability. In addition, Bilo and Bilo (1978) observed a tail response to electrical vestibular stimulation that was specific to the simulated flight state (Rabin 1973, 1974, 1975a, 1975b). Tail movements make important aerodynamic contributions to postural orientation specifically during flight. Thus a state-specific tail response to motion would be appropriate for these animals.

In the current study, we characterized vestibular eye, head, and tail responses to rotational motion in the dark. We compared responses obtained during simulated flight with those observed during a resting or default condition. We conclude that the vestibular head response is significantly larger in amplitude during flight, thus improving head-in-space stability. In addition, the vestibular tail response is observed only during flight and is compensatory for whole-body tilt.

METHODS

Seven adult pigeons (Columba livia), ranging in weight from 400 to 700 g, were used in accordance with the guidelines set forth by the National Institutes of Health Guide for the Care and Use of Animals in Research; the study was approved by the Institutional Animal Care and Use Committee. The animals were housed and cared for in the Laboratory Animal Facilities under veterinary supervision.

Animal preparation.

Each bird was surgically implanted with a delrin stud, used to fix the animal's head position during experiments. General anesthesia was achieved using isoflurane gas (3–5% in O₂) via endotracheal intubation. Heart rate was monitored, and core temperature (40°C) was maintained using a heating pad. An incision was made along the midline of the skull, and the underlying periosteum was removed from the bone. The head stud was attached to the skull via titanium self-tapping screws and secured with dental acrylic. The head was positioned stereotaxically with the beak angled ∼12° downward, such that the upright orientation of the head stud corresponded on average to alignment of the horizontal semicircular canals with an earth-horizontal plane (Dickman 1996). After surgery, the wound margin was sutured closed.

After a 5- to 7-day recovery period, four birds were subsequently implanted unilaterally with an eye coil assembly, used to monitor three-dimensional eye movements. The coil assembly consisted of a large diameter (12 mm) direction coil and a small diameter (2 mm) torsion coil. The direction coil was constructed using three turns of multi-stranded, Teflon-coated, 41-gauge stainless steel wire (A-M Systems), and the torsion coil consisted of a 100-turn copper wire watchmaker coil (Imetra) attached perpendicularly to the direction coil. The three-dimensional (3D) coil assembly was coated in a thin layer of Araldite epoxy (Huntsman). For implantation, the animal was anesthetized using isoflurane gas (3–5% in O₂) via endotracheal intubation. A circumferential incision was made in the conjuctiva to allow visualization of the sclera. In pigeons, the sclera is calcified near the cornea, so the 3D coil assembly was attached to the posterior margin of the scleral globe with 8-0 prolene sutures. Coil leads were threaded through the bone of the posterior orbit and passed underneath the skin to exit near the head stud. The conjuctiva was then approximated and closed with 8-0 vicryl sutures. Lead wires were soldered to a nanoconnector (Omnetics), which was secured next to the head stud with dental acrylic. Following surgery, butorphanol (10 mg/kg) was administered for pain, and ophthalmic antibiotic ointment (bacytracin-neomycin) was applied to the operated eye. Pigeons were allowed to fully recover (7–10 days) before being used in experiments.

Behavioral recording.

For each experiment, the pigeon was secured to a restraint arm mounted to a six degrees-of-freedom hydraulic motion platform (Rexroth-Bosch, The Netherlands). The pigeon restraint included a plastic sling positioned around the animal's anterior torso, secured with a thin Velcro strip. Additional Velcro strips were wrapped around the wing joints and secured to the sling, and the sling assembly was secured to the restraint arm on the motion platform (Fig. 1). The restraint minimized movement of the pigeon's torso relative to the motion platform while leaving the wings, tail, and legs free to move. The head stud could be fixed to the restraint arm or freed, allowing for head-fixed or head-free experiments.

Fig. 1.

Pigeons change behavioral state in the presence of airflow. A: air off: default state. B: air on: flight state.

A three-field AC magnetic coil system (Riverbend Instruments, Birmingham, AL) was used to monitor the pigeon's rotational eye and head movements. The field coils were mounted to the motion platform and provided a 6-inch homogeneous magnetic field cube centered about the pigeon's head. Eye and head movements were monitored using a 3D eye coil and a separate head coil assembly attached to the stud. An infrared optical tracking system (Optotrak Certus, Northern Digital) was used to monitor tail movements in three birds. Since the primary concern in this study was to characterize the overall changes in tail-on-body angular position (rather than the more complex movements of individual feathers and muscles), the tail was experimentally defined by a 3D rigid body (∼8 g) attached to the primary feathers of the tail with wire twist ties and surgical tape. The rigid body was made of thin, stiff plastic arranged in two spatial planes: one flush to the tail surface (7 cm × 5.5 cm) and a second smaller piece angled 45° upward. This construct did add some weight to the tail and limited the individual movements of the feathers (relative to one another). However, observations of tail movement without attachment of a rigid body confirmed that the gross characteristics of the tail movement (i.e., state-dependent amplitude, amplitude range, and primary spatial components) were not affected by the rigid body. An additional rigid body with four infrared light-emitting diodes (LEDs) was secured to the motion platform. The position of the LEDs was recorded by a set of cameras mounted to the ceiling above the motion platform. System software (Tool Bench, Northern Digital) was used to compute the rotational orientation of each rigid body relative to the camera. Tail and platform position data were time-synced offline to eye and head coil data.

Motion stimuli.

The motion platform was driven by a personal computer and programmable interface (CED Model 1401plus, Cambridge Electronic Design). Stimulus control and data acquisition were performed using custom scripts written for the interface environment (Spike2, Cambridge Electronic Design). Stimulus deliveries were monitored using a rate sensor and three-axis linear accelerometer mounted to the motion platform. With the head fixed, the x-, y-, and z-axes corresponded to the pigeon's nasooccipital, interaural, and dorsoventral axes, respectively. Positive rotations were right ear down, nose down, and leftward. Sinusoidal rotations about the x- (roll), y- (pitch), and z- (yaw) axes at 0.25, 0.5, and 1 Hz and amplitudes of 5, 10, 15, and 20°/s peak angular velocity were delivered. Each experiment was performed with the pigeon either head-fixed (using the head stud) or head-free.

Experimental protocol.

All experiments were performed in darkness. A motion platform-fixed coordinate system (right-handed) was used to register eye, head, and tail movements and motion stimuli. Thus quantification of eye, head, and tail position relative to the body was derived. To test the effect of behavioral state on responses to motion, we repeated the protocol with and without the presence of airflow. Airflow was delivered using a commercial blower (hose diameter = 4 cm) secured to the motion platform ∼12 in forward of the pigeon's frontal breast surface and directed along the x-axis at the level of the pigeon's torso. Some airflow also reached the animal's head, but the primary airflow arrived at torso level. The speed of the airflow was ∼15 m/s (measured at the pigeon's frontal sufface), an appropriate air speed for slow to moderate gliding flight (Pennycuick 1968). In the absence of airflow (default state, Fig. 1A), pigeons were generally relaxed in the sling. When airflow was initiated, pigeons tensed their wings, lifted their tails, and pulled their legs up and extended them under the tail (flight state, Fig. 1B). This gliding flight posture was consistently observed across every animal tested and could be maintained for long periods of time (>30 min) without any observable change (after the initial 60-s transition period). Trials were blocked by stimulus axis, frequency, and behavioral state (default or flight). At least 60 s passed between blocks. The order of block presentation, as well as the order of velocity presentation within a single block, was varied across experiments. Throughout experiments, birds were monitored for alertness by visual inspection of posture and spontaneous eye/head movements.

For a subset of birds (n = 4), additional experiments were done to determine whether the noise associated with airflow (blower and air movement noise) was sufficient to produce the observed effects of state on behavior. For these experiments, the flight condition was replaced with a noise only condition, in which the blower was turned on but airflow was prevented from reaching the bird by a physical barrier secured to the motion platform. Experiments using the noise only condition were performed before experiments using the flight condition, to ensure that pigeons had not yet formed an association between blower noise and actual airflow.

Data analyses.

Eye and head movement signals were first converted to rotation vectors in Cartesian coordinates. Note that the animal's nasooccipital axis was directed collinear to the positive x-axis and that the primary visual axis was located ∼66° away from the bill tip, with the head fixed (Martin and Young 1983). Fast phases (including saccades, gaze shifts, and anomalies) were removed using custom scripts written in Matlab (Mathworks). Only slow-phase eye and head velocity were analyzed. Slow-phase rotation vectors were differentiated to produce rotation velocity vectors. From the rotation vectors, angular velocity vectors with components about the x-, y-, and z-axes were calculated (Haustein 1989; Hess et al. 1992; van Opstal 1993). The eye rotation velocity vectors corresponded to the net movement of the eye within the field coils (that is, within the platform-fixed coordinate system). During head-fixed trials, movement of the eye in the coils was exclusively due to movement of the eye in the head. During head-free trials, movement of the eye in the coils was due to movement of the head on the body and the eye in the head. To compute eye-in-head movement, the head coil response was vectorially subtracted from the eye coil response gain and phase (Haque and Dickman 2005).

Angular orientation of rigid bodies (RBs) on the pigeon's tail and the motion platform were also expressed as rotation vectors in Cartesian coordinates. Initially, these were expressed relative to a camera-fixed (space-fixed) coordinate system. Tail orientation relative to the motion platform-fixed coordinate system was computed by: [RB_platform in space]^T × [RB_tail in space] = [RB_tail in RB_platform], where []T indicates the matrix transpose. This formula uses the original angular positions of the tail and platform rigid bodies to express the angular position of the tail relative to the moving platform. The resulting orientations of the tail rigid body in platform-fixed coordinates were then expressed as rotation vectors, and the x-, y-, and z-axis components of angular position and velocity were calculated. To obtain absolute angular position of the tail relative to the pigeon's body, we expressed angular position relative to a reference with the tail oriented in a platform-horizontal plane, where the dorsal surface of the tail was approximately collinear with respect to the dorsal surface of the torso. Thus tail angular position was negative when angled downward below the torso, zero when parallel to the torso, and positive when angled above the torso.

For each trial, the rate sensor and accelerometer signals, as well as the steady-state eye, head, and tail movement responses, were fit with sine curves at the fundamental stimulus frequency using a least-squares minimization algorithm. The fitted curves were used to calculate gain and phase values of the behavioral response components. For eye, head, and gaze responses, gains were expressed as the amplitude of the sinusoidal velocity response (in °/s) divided by the amplitude of the sinusoidal stimulus velocity (in °/s), and phase values were expressed relative to peak platform velocity. For tail responses, gains were expressed as the peak change in tail angular position (in °) divided by the peak change in platform orientation (in °), and phase values were expressed relative to peak platform angular orientation.

Mean gain and phase values were computed vectorially (see Wei and Angelaki 2001), and standard errors were computed using the error propagation formula (Bevington and Robinson 1992). All statistical analyses were based on univariate analysis of covariance (ANCOVA; Statistica, Statsoft), with either response gain or phase as the dependent variable and frequency and velocity of the stimulus as covariates. For all responses, both animal and behavioral state were included as factors. For the analysis of selected components of the tail responses, head-fixed/head-free was also included as a factor. In the initial analysis of tail responses in which all components (x/y/z) were analyzed, component was included as an additional factor. Post hoc comparisons were made using the Bonferroni method to adjust the significance level for multiple comparisons. Analyses were run separately for each stimulus axis. In Tables 1–3, statistical P values are stated as ranges (i.e., P < 0.001) when significance far exceeds the α-level of 0.05 and are stated as real values otherwise to permit independent evaluation of significance.

Table 1.

Statistical comparisons—head response

	Adjusted Mean			Interaction
	Air Off	Air On	Main Effect	×Amplitude	×Frequency
Pitch
Gain (°/s per °/s)	0.77	0.96	F(1,844) = 66.1, P < 0.001	F(1,844) = 16.2, P < 0.001	F(1,844) = 0.051, P = 0.822
Phase (°)	−175.8	−178.2	F(1,844) = 0.05, P = 0.822	F(1,844) = 1.30, P = 0.254	F(1,844) = 0.64, P = 0.425
Roll
Gain (°/s per °/s)	0.80	0.91	F(1,864) = 80.1, P < 0.001	F(1,864) = 24.5, P < 0.001	F(1,864) = 0.31, P = 0.576
Phase(°)	−181.0	−180.4	F(1,864) = 3.8, P = 0.0512	F(1,864) = 3.8, P = 0.0524	F(1,864) = 0.50, P = 0.482
Yaw
Gain (°/s per °/s)	0.89	0.98	F(1,902) =14.5; P < 0.001	F(1,902) = 2.0, P = 0.156	F(1,902) = 2.03, P = 0.155
Phase (°)	−176.8	−177.7	F(1,902) = 0.21, P = 0.645	F(1,902) = 0.31, P = 0.575	F(1,902) = 0.33, P = 0.567

Data included from 6 animals. Results from analysis of covariance (ANCOVA) (factors: air, animal; covariates: amplitude, frequency).

Table 2.

Statistical comparisons —head-free gaze and head-fixed eye responses

	Adjusted Mean			Interaction
	Air Off	Air On	Main Effect	×Amplitude	×Frequency
Head-free net gaze
Pitch
Gain (°/s per °/s)	0.83	0.92	F(1,279) = 35.8, P < 0.001	F(1,279) = 14.8, P < 0.001	F(1,279) = 2.5, P = 0.114
Phase (°)	−175.9	−178.8	F(1,279) = 1.1, P = 0.304	F(1,279) = 0.36, P = 0.549	F(1,279) = 0.43, P = 0.512
Roll
Gain (°/s per °/s)	0.90	0.96	F(1,293) = 36.9, P < 0.001	F(1,293) = 12.6, P < 0.001	F(1,293) = 11.2, P < 0.001
Phase (°)	−178.1	−179.9	F(1,293) = 5.9, P = 0.016	F(1,293) = 0.41, P = 0.521	F(1,293) = 0.43, P = 0.513
Yaw
Gain (°/s per °/s)	0.99	1.04	F(1,333) = 1.9, P = 0.167	F(1,333) = 0.42, P = 0.519	F(1,333) = 0.45, P = 0.503
Phase (°)	−176.6	−177.4	F(1,333) = 0.84, P = 0.360	F(1,333) = 0.34, P = 0.562	F(1,333) = 0.57, P = 0.449
Head-free eye response
Pitch
Gain (°/s per °/s)	0.26	0.35	F(1,589) = 4.9, P = 0.027	F(1,589) = 0.07, P = 0.79	F(1,589) = 1.6, P = 0.21
Phase (°)	−161.8	−161.0	F(1,589) = 0.30, P = 0.58	F(1,589) = 0.15, P = 0.70	F(1,589) < 0.001, P = 0.97
Roll
Gain (°/s per °/s)	0.46	0.58	F(1,549) = 5.7, P = 0.017	F(1,549) = 0.26, P = 0.61	F(1,549) = 6.5, P = 0.011
Phase (°)	−162.5	−162.2	F(1,549) = 7.6, P = 0.006	F(1,549) = 2.8, P = 0.094	F(1,549) = 7.8, P = 0.0055
Yaw
Gain (°/s per °/s)	0.64	0.71	F(1,631) = 0.10, P = 0.75	F(1,631) = 0.16, P = 0.69	F(1,631) = 6.2, P = 0.013
Phase (°)	−161.7	−158.9	F(1,631) = 8.9, P = 0.003	F(1,631) = 0.40, P = 0.53	F(1,631) = 6.0, P = 0.014

Data included from 3 animals per response. Results from ANCOVA (factors: air, animal; covariates: amplitude, frequency).

Table 3.

Statistical comparisons - Tail responses

	Adjusted Mean			Interaction
	Air Off	Air On	Main Effect	×Amplitude	×Head	×Animal
Pitch Taily
Gain (°/s per °/s)	0.11	0.98	F(1,456) = 118.7, P < 0.001	F(1,456) = 0.16, P = 0.69	F(1,456) = 23.0, P < 0.001	F(1,456) = 12.4, P < 0.001
Phase (°)	139.2	9.89	F(1,456) = 185.8, P < 0.001	F(1,456) = 1.08, P = 0.30	F(1,456) = 8.13, P = 0.005	F(1,456) = 6.23, P = 0.002
Roll Tailz
Gain (°/s per °/s)	0.14	0.28	F(1,472) = 40.5, P < 0.001	F(1,472) = 2.81, P = 0.09	F(1,472) = 0.20, P = 0.65	F(1,472) = 5.9, P = 0.003
Phase (°)	−24.9	−18.6	F(1,472) = 11.5, P < 0.001	F(1,472) = 6.53, P = 0.01	F(1,472) = 1.04, P = 0.31	F(1,472) = 6.6, P = 0.0015

Data included from 3 animals. Results from ANCOVA (factors: air, animal, head; covariate: amplitude).

RESULTS

Behavioral state: default versus flight.

To study the dependence of motion responses on behavioral state, we simulated gliding flight by delivering a frontal airflow to the pigeon's ventral breast surface. This technique has been used in previous studies of state-dependent behavior in pigeons (Bilo 1992, 1994; Bilo and Bilo 1978, 1983; Gioanni and Maurice 2004; Gioanni and Sansonetti 1999, 2000; Maurice and Gioanni 2004), and our observations regarding the effect of airflow on the bird's posture are consistent with their descriptions. In the absence of airflow, pigeons held their wings relaxed at their sides, their tails angled downward (depressed below the torso), and their legs relaxed underneath their bodies (default state, Fig. 1A). Within the first minute after airflow onset, three distinct postural changes took place: 1) the wings were tensed and pulled upward against the body, 2) the tail was lifted to an initial position above horizontal that then settled to a steady-state horizontal position within 30–60 s, and 3) the legs were pulled up and extended underneath the tail. The resulting posture (flight state, Fig. 1B) was similar to that observed during midspeed level gliding flight in wind tunnels (Pennycuick 1968). These postural changes were consistent across birds, maintained throughout each trial, and evoked under both head-fixed and head-free conditions in the dark. Occasionally, pigeons would exhibit a brief startle response to airflow onset, including increased fast head movements and erratic tail movements; however, these responses were rare and did not persist for more than 30 s.

It was possible that differences in behavior observed in the presence and absence of airflow could simply result from startle responses to the blower and associated changes in arousal. To determine whether the startling stimulus alone was affecting behavior, pigeons were subjected to a third condition (noise only) in which the airflow source was turned on but a physical barrier prevented the airflow from reaching the bird. Initially, we attempted this condition with pigeons after weeks of exposure to the original default and flight conditions, in which blower noise and actual airflow were constantly paired. In response to blower noise alone, these pigeons exhibited a short-lived change in posture that was similar to simulated flight posture. This posture decayed within several minutes of noise onset, in contrast with the robust and long-lasting change in posture achieved during actual airflow. Furthermore, no such change in posture was observed in naïve animals. These observations are consistent with pigeons learning an association between airflow and blower noise that then allowed the blower noise alone to initiate (although not to sustain) flight posture. Thus all data reported here for the control condition were collected from pigeons before experiments using actual airflow. During noise only trials, these naïve pigeons did not exhibit any of the changes in posture that were consistently observed in the presence of airflow, although they would occasionally display a brief startle response. In fact, birds′ behavior during noise only trials was consistent with their still being in the default state.

Head responses: state-dependent gain.

Head responses to steady-state sinusoidal rotation in the dark were recorded from six alert pigeons during two behavioral states: default (air off) and flight (air on). During the default state (Fig. 2A), head responses were present and compensatory in phase (−180°), although the response amplitude was under-compensatory (gain <1). Consistent with previous studies, we found that head responses were spatially compensatory for the axis of the rotational motion (Dickman et al. 2000; Haque and Dickman 2005; McArthur and Dickman 2008). For instance, a pitch rotation elicited a head movement primarily along the animal's y-axis, with negligible (<10%) movement along orthogonal axes. These small noncompensatory response components were not analyzed further.

Fig. 2.

Head and tail responses to whole-body rotation with air off (default) and air on (flight). Compensatory components of head and tail responses to pitch rotation (0.5 Hz, 20°/s) are shown. A: in the absence of airflow (default state), head movements (Head_Y) are compensatory in phase (−180°) but low in amplitude. Tail position is depressed, and tail movement (Tail_Y) is negligible. B: during airflow (flight state), head movements are larger in amplitude. Tail movements increase and are compensatory (0°) in phase. Gray curves = position; black curves = velocity.

During the flight state (Fig. 2B), head responses to rotational motion were still spatially and temporally compensatory but were higher in amplitude than those observed during the default state. A perfectly compensatory head response would have unity gain and opposite (−180°) phase relative to the platform motion stimulus, to maintain head-in-space stability. As shown in Fig. 3, head responses recorded during the default state (black) had under-compensatory gains to all three axes of stimulus motion (0.5 Hz), across all birds tested (n = 6). Head gains during yaw rotation were generally higher than during pitch and roll, particularly for lower velocity stimuli. Head gains did not vary as a function of stimulus frequency (ANCOVA: P > 0.05; data not shown). However, head gains increased with increasing peak stimulus velocity, particularly for pitch and roll. At the lowest velocity (5°/s), the head response compensated for only about half of the platform rotation, whereas at the highest velocity (20°/s) the head response compensated for ∼80% of the motion. Head gains increased during the flight state (red) to near unity across all stimulus velocities for all three stimulus axes. There was also a modest trend for improved compensatory phase in the flight state (ANCOVA: pitch and yaw, P > 0.5; roll, P = 0.05).

Fig. 3.

Mean head responses for each animal (n = 6) during 0.5 Hz head-free rotational stimuli. Gain values are expressed as the ratio of response slow-phase angular velocity to rotational stimulus velocity, and phase values are expressed relative to peak rotational stimulus velocity. Responses are shown for three axes of stimulation (A: pitch; B: roll; C: yaw), plotted against stimulus velocity. The dashed lines indicate the gain (unity) and phase (−180°) required for a perfectly compensatory response. Symbol shapes indicate different animals. Black lines, solid symbols = air off (default state); red lines, open symbols = air on (flight state).

The effect of state on head gain across all stimulus conditions (0.25–1 Hz, 5–20°/s) is summarized in Fig. 4A. Head gains were significantly higher during flight than during the default state (ANCOVA: P < 0.001; Table 1). It was possible that this increase in head gain was solely due to changes in arousal produced by blower noise. Thus we also compared head gains recorded during default (blower off) and noise only (blower on) conditions (Fig. 4B) for four animals. Unlike the flight state, blower noise alone was not associated with a significant increase in head gain (ANCOVA: pitch and roll, P > 0.4; yaw, P < 0.001 corresponding to a decrease in gain of <0.01 during noise only trials).

Fig. 4.

Head response gain and the effect of behavioral state. Each data point corresponds to the mean head response gain measured for a single animal during pitch (black), roll (gray), and yaw (open) rotation at a specific velocity (5–20°/s) and frequency (0.25–1 Hz). Symbol shapes indicate different animals. A: comparison of head gain measured with air off (default state) and air on (flight state), n = 6. B: comparison of head gain measured with blower off (default state) and blower on (noise only), n = 4.

Gaze-related responses: net gaze and eye movements.

Under head-free conditions, gaze (eye-in-space) is the net result of both head and eye component responses. Previous studies have demonstrated that pigeon gaze under head-free conditions is dominated by the head component (Gioanni 1988b; Haque and Dickman 2005). Thus we predicted that gaze state-dependence would be similar to that observed for the head. A comparison between head and gaze gains across behavioral states is shown in Fig. 5A. First, gaze gain was higher than head gain overall. Second, both gaze and head gains were higher during simulated flight (red) than during the default condition (black), specifically for pitch and roll rotation (ANCOVA: P < 0.001; Table 2). Note that the gains during yaw were already closer to unity during the default state, so a smaller increase in gain during flight was appropriate for an optimal response. However, the eye component of the head-free gaze response was not affected by behavioral state (ANCOVA: P > 0.05), probably because head-free eye gains were generally small (< 0.3) and variable (Fig. 5B).

Fig. 5.

Gaze response gains and the effect of behavioral state. Each data point corresponds to the mean response gain measured for a single animal during pitch (black/red), roll (gray/pink), and yaw (open) rotation at a specific velocity (5–20°/s) and frequency (0.25–1 Hz). Symbol shapes indicate different animals (n = 3). Black symbols = air off (default state); red symbols = air on (flight state). A: comparison of head and net gaze (eye + head) gain. B: comparison of head-fixed and head-free eye gain.

Previous studies have shown that eye gains are higher under head-fixed conditions, where only eye-in-head movements contribute to gaze (Gioanni 1988b; Haque and Dickman 2005). Here we observed similar results in that head-fixed eye gains were higher than head-free gains (Fig. 5B, black). These head-fixed eye responses also increased in gain during simulated flight (Fig. 5B, red), specifically for pitch and roll rotation (ANCOVA: P < 0.001; Table 2). Like the head response, blower noise alone was not sufficient to significantly increase head-fixed eye gain (ANCOVA: pitch, P = 0.68; roll, P = 0.45).

Tail responses: pitch Tail_Y and roll Tail_Z.

Previous studies have reported a tail movement response to galvanic vestibular stimulation that only appears during simulated flight (Bilo and Bilo 1978). Here we characterized compensatory tail movements during rotational motion about the three cardinal axes in three birds. During the default state, the tail remained immobile on the body even during rotational stimuli (Fig. 2A). Small-amplitude tail movements during respiration and rare transient startle responses were sometimes observed. In the flight state (Fig. 2B), the tail moved substantially on the body, with variable amplitude but a consistent phase relationship to the motion profile. Representative tail responses and gain data for the three birds are shown in Fig. 6. The largest and most consistent response across animals was a pitch tail movement (Tail_Y) in response to pitch rotation (Fig. 6A, red). When the animal was pitched nose-down in the flight condition, elevation of the tail above the body was observed. This pitch response varied in amplitude but was consistently observed in all animals. During roll rotation, a yaw tail movement (Tail_Z) was observed (Fig. 6B, green). This response corresponded to a twisting of the tail relative to the body that also varied across birds but was generally of smaller amplitude than the Tail_Y response to pitch rotation. During yaw rotation, only very small tail movements were observed (Fig. 6C). All three components of the tail response increased in amplitude during flight, for pitch and roll but not yaw rotational stimuli (ANCOVA: pitch, P = 0.003; roll, P = 0.04; yaw, P = 0.2; confirmed for each component with post hoc comparisons). We chose the largest component for each stimulus (Tail_Y during pitch rotation; Tail_Z during roll rotation) for further analyses.

Fig. 6.

Three-dimensional tail position during pitch (A), roll (B), and yaw (C) rotation. For each stimulus, representative tail position traces in response to rotation during simulated flight are shown. Blue, red, and green traces correspond to x-, y- and z-axis tail position. Gray traces correspond to platform motion (0.5 Hz, 20°/s), equal to the displacement of the bird's body in space. Each bar graph plots mean gain of the 3 tail response components for a single behavioral state (top = flight state; bottom = default state) to the designated stimulus axis, shown separately for each of 3 animals. Error bars = SE.

As shown in Fig. 7A, the Tail_Y response to pitch rotation was negligible in the absence of airflow and increased by an order of magnitude during flight (ANCOVA: P < 0.001; Table 3). The response gain did not vary as a function of stimulus frequency (ANCOVA: P = 0.8) and showed only a weak dependence on stimulus peak velocity (ANCOVA: P = 0.1). State-dependence of the Tail_Y response was further reinforced by the observed change in phase (ANCOVA: P < 0.001). During the default state, the small-amplitude Tail_Y response tended to be opposite in phase (180°) relative to the stimulus. However, the response became nearly in-phase with the stimulus (near 0°) during simulated flight. For both gain and phase, there was a significant interaction between state (flight/default) and animal (P < 0.01). Post hoc tests confirmed that the effect was significant for each animal (P < 0.001 for both gain and phase). Gain and phase of the Tail_Z response are summarized in Fig. 7B. Like the Tail_Y response, the gain of the Tail_Z response was significantly higher during simulated flight (ANCOVA: P < 0.001; Table 3) for each animal tested (ANCOVA, state × animal, P = 0.003; post hoc, P < 0.001). Response gain did not vary with stimulus frequency (ANCOVA: P > 0.05) but did increase with peak stimulus velocity (ANCOVA: P = 0.04). There was a small difference (<10°) in Tail_Z response phase during flight in one of the three animals (ANCOVA: P < 0.001), but the difference was very small.

Fig. 7.

Mean tail responses for each animal (n = 3) during 0.5 Hz head-free rotational stimuli. A: change in tail y-axis position (Tail_Y) during pitch rotation. B: change in tail z-axis position (Tail_Z) during roll rotation. Gain values are expressed as the ratio of change in tail angular position to stimulus angular position, and phase values are expressed relative to peak stimulus angular position. Symbol shapes indicate different animals. Black lines, solid symbols = air off (default state); red lines, open symbols = air on (flight state). Error bars = SE.

The effect of state on the gain of Tail_Y and Tail_Z responses to pitch and roll rotation, respectively, was observed under both head-free and head-fixed conditions (compare black and red curves in Fig. 8). Overall, state-dependent increases in tail gain tended to be higher under head-free conditions, particularly for the Tail_Y response to pitch rotation (ANCOVA: Tail_Y, P = 0.005; Tail_Z, P = 0.62). Smaller state-dependent increases in gain under head-fixed conditions make sense in the context of flight muscle reflexes, given that there are additional tail muscle responses evoked by neck flexion (Bilo and Bilo 1983; Bilo 1994) that complement vestibular responses during head-free conditions. However, we question the robustness of this finding, since it was variable across animals (Fig. 8A).

Fig. 8.

Mean tail responses for each animal (n = 3) during head (HD)-fixed and head-free rotational stimuli (0.5 Hz, 10°/s). A: mean gain values for Tail_Y responses to pitch rotation. B: mean gain values for Tail_Z responses to roll rotation. Black lines, solid symbols = air off (default state); red lines, open symbols = air on (flight state). Error bars = SE.

As with eye and head responses, substituting the noise only condition for the flight condition did not produce significant increases in tail response gain [Tail_Y during pitch: F_(1,173) = 1.40, P = 0.24; Tail_Z during roll: F_(1,206) = 1.10, P = 0.30], indicating that blower noise alone was not sufficient to evoke flight-related tail responses. In Fig. 9, we compare head-free gain values for the original experimental conditions (default vs. flight) and the control conditions (blower off vs. noise only) during 0.5 Hz rotation. Each data point corresponds to gain values obtained from matched trials during a single experiment, rather than means across experiments, to illustrate the response variability. Although head and head-fixed eye gains were generally consistent across trials and across cycles within a single trial, tail movements (particularly Tail_Y recorded during pitch rotation) exhibited greater variability. The trends in the data agree with the statistical comparisons: tail gain increased during flight but did not increase in the presence of blower noise alone. However, plotting the data from individual trials reveals that a small subset of control (noise only) trials display higher Tail_Y gains (see outlying black empty triangles in Fig. 9A). Note that the gain values for these data points are higher during both blower off and blower on conditions. We attribute these outlying, high-gain control trials to transient tail movements associated with arousal, which likely contribute to the variability in tail response overall.

Fig. 9.

Paired single-trial tail responses to head-free rotational stimuli (0.5 Hz), recorded during different behavioral states. Each data point corresponds to the response gain measured for a single animal at a single stimulus velocity. Symbol shapes indicate different animals (n = 3). Solid red symbols = comparison of tail gain measured with air off (default state) and air on (flight state). Open black symbols = comparison of tail gain measured with blower off (default state) and blower on (noise only). A: Tail_Y response to pitch rotation. B: Tail_Z response to roll rotation.

DISCUSSION

Here, we demonstrated that nonvisual gaze and posture stabilizing responses to whole-body rotation are state dependent in pigeons. We found that the flight state enhanced vestibular gaze and head-in-space stabilizing responses by increasing the gain of eye and head movements during rotational motion. We also observed two state-specific compensatory tail responses to rotation: Tail_Y movements during pitch rotation and Tail_Z movements during roll rotation. Both responses are likely to contribute to body-in-space stabilization during flight. These results have important implications not only for nonvisual behaviors implemented during bird flight but also for state-dependent modification of reflexes in general.

State-modulated responses: head-in-space and gaze stability.

We demonstrated that vestibular head responses to rotational motion were significantly larger and more compensatory during simulated flight, particularly for low velocity stimuli. These results are complementary to those previously shown for head responses to visual motion, which have near-unity gains for higher stimulus velocities during simulated flight (Bilo 1992, 1994; Gioanni and Maurice 2004; Gioanni and Sansonetti 1999; Maurice et al. 2006). In both cases, the flight state extends the working range of the head-stabilizing response, improving it for the velocity range at which it tends to be weakest (low velocities for nonvisual vestibular responses, high velocities for visual responses). Furthermore, flight state has the strongest effect on head responses to pitch and roll rotation, which have smaller default gains than yaw rotation for both vestibular (Haque and Dickman 2005) and optokinetic (Gioanni 1988a) reflexes over the frequency range used here.

Although head-free eye responses were small and variable (independent of state), fixing the head increased the gain of the eye response, as previous studies reported (Gioanni 1988b; Haque and Dickman 2005). Furthermore, we saw a state-dependent eye gain increase during simulated flight in the head-fixed condition. A previous study failed to demonstrate state-dependent improvements in the optokinetic response, although coordination between eye and head movements during flight was improved (Gioanni and Maurice 2004). This might be related to a change in the phase of the cervico-ocular reflex, which follows head-on-body velocity during walking but follows head-on-body position during flight (Maurice and Gioanni 2004). Under ethological conditions, gaze stability in birds is dominated by head responses, and head-in-space stability is further improved during flight. Thus the need for compensatory eye movements during flight is lessened, as reflected by the low head-free eye gains observed here. However, stimuli were restricted to a relatively slow velocity range (5–20°/s) at moderate frequencies (0.25–1 Hz). It is possible that eye gains could increase with higher velocities or frequencies, given that the inertia of the head could slow its response under those conditions.

We also observed improved gaze stability during simulated flight, which likely serves to minimize blurring of the visual scene during rotational motion. However, head-in-space stability appears to be critical for its own sake as well. Pigeons tend to maintain a stable head orientation across behavioral contexts such as standing, walking, and flying (Erichsen et al. 1989). In addition, pigeons isolate their head position from body movements, both translational and rotational, during slow flapping flight (Warrick et al. 2002). Even when the animal's goal is to change its azimuth orientation, as during a banking turn, changes in head orientation are restricted to quick fast-phase head movements alternated with very stable hold-phases (Bilo 1992, 1994). In fact, if a pigeon's head is locked to the body using a cervical collar, the bird is unable to maintain equilibrium during flight (Warrick et al. 2002). Warrick et al. (2002) provide an explanation for the importance of aerial head stability in terms of sensory conflict. During a maneuver such as a banking turn, the visual system will respond to motion in the visual scene throughout the movement, but the vestibular system responds specifically to linear and angular accelerations. If the head and body moved together during a rotational maneuver, transient conflicts between visual and vestibular sensory input could be catastrophic for the animal.

State-specific responses: body-in-space orientation stability.

During simulated flight, pigeons generated a tail-on-body pitch movement in response to pitch rotation, where the tail was elevated or depressed when the animal was pitched nose down or nose up, respectively. This result is consistent with previous observations of tail movements in pigeons during actual flight (Brown 1963; Warrick et al. 2002). Furthermore, the tail response appears to be aerodynamically compensatory. For example, if a flying pigeon experienced a nose-down pitch rotation (e.g., due to air turbulence), elevating the tail would decrease the lift acting on the posterior part of its body, thereby pitching its body upright again. Conversely, in response to a nose-up pitch stimulus, depressing the tail would increase posterior lift, thereby pitching the body back down to its original orientation. In addition, we observed a flight-specific tail yaw movement in response to roll rotation, where the tail twisted right or left during right ear down and left ear down rotations, respectively. These are in fact spatially compensatory responses. As noted by Brown (1963), the body tends to be more unstable in pitch than roll in turbulent air, with the wings providing most of the necessary active and passive roll stability. However, during banking turns, the bird's body actively rolls in the direction of the turn, lifting the outside wing to decrease drag. The asymmetrical drag across the right and left sides of the body produces a yawing movement in the opposite direction of the turn (i.e., banking roll right ear down results in a wing drag-associated yawing movement to the left). It is this yawing movement during active rolling maneuvers that is counteracted by twisting and depressing of the tail (Warrick et al. 2002).

The state-dependent tail responses we observed were consistent with previous results using galvanic vestibular stimulation to elicit tail responses in pigeon simulated flight (Bilo and Bilo 1978). Some transient elevations of the tail were observed outside of simulated flight, although these were rare and seemed more closely tied to the bird's level of arousal than to the strength of the rotational stimulus. These transient responses might be similar to nonvestibular tail reflexes reported by others (Biederman-Thorson and Thorson 1973; Delius and Vollrath 1973), which occurred in the absence of airflow and were possibly driven by mechanosensors in the mesentery. We cannot exclude the possibility that these responses also contributed to the tail movements analyzed in the current study. However, given that these nonvestibular responses were observed in the absence of airflow, it is unlikely that they contributed to the dramatic state-dependent increase in tail movement amplitude evoked by the flight state.

There was a significant amount of interindividual variability in the amplitude of the state-specific tail responses. Although the current study included a detailed analysis of responses from only three birds, qualitative observation of tail movements in additional animals indicates that the variability reported here is representative of real variability across pigeons. These tail movements function to counteract air turbulence during flight, thus maintaining upright body orientation in space. The amplitude of tail movement needed to counteract a given passive whole-body rotation will depend on the size and shape of the animal, as well as the size and shape of the animal's tail. An analysis of how these factors covary with tail response amplitude is beyond the scope of this study; however, it is likely that these factors are responsible for the interindividual tail movement variability.

Vestibular responses: modulation by behavioral state.

Previous studies have demonstrated long-term modification of vestibular reflexes, evoked by manipulation of the sensorimotor mapping. Detection of errors in the motor output drives a gradual change in the sensorimotor reflex to reduce the error, by processes collectively referred to as adaptation. It is well documented that the vestibular eye response (VOR) adapts to manipulation of visual feedback during whole-body rotation (Blazquez et al. 2004; Boyden et al. 2004; Schubert and Zee 2010). Vestibulospinal responses also adapt to changes in the sensorimotor mapping. For instance, experimental increases in head inertia are associated with increased strength of head responses to horizontal whole-body rotation, consistent with adaptation of the vestibular and flexion-related head responses (Goldberg 1999; Reynolds et al. 2008; Reynolds and Gdowski 2008). Another example of vestibulospinal response adaptation was found for forelimb vestibular reflexes in cats, which adapt in response to combinations of vestibular and joint stimulation that signal an error in posture stabilization (Andre et al. 2005).

Although adaptation requires feedback and error detection, the state-dependent changes we observed required instead a change in sensorimotor cues to trigger a transition between distinct behavioral states. The vestibular stimuli were identical across conditions, and there was no manipulation of the relationship between vestibular input and motor output. Therefore, no additional state-specific errors were present. However, it's possible that state-dependent responses could result from cue-driven retrieval of distinct adapted states that were acquired during the bird's development. The flight state experienced under ethological conditions introduces a distinct sensory environment with constant visual flow and forward linear movement. That change in the environment could, in fact, favor an adjustment in the sensorimotor mapping. Birds might develop two distinct neural mappings associated with flight-related cues such as frontal airflow and lack of limb contact with a terrestrial surface.

We can only speculate on the role of development and flight training on the resulting state-dependent responses described here, but there is a growing body of evidence that supports state-specific retrieval of adapted states as a general property of motor systems. In the seminal behavioral model of visuo-motor adaptation, prism goggles displace the visual field relative to the subject, and subjects gradually modify their motor output to reduce the resulting error in visually guided behaviors. Interestingly, subjects that are required to alternate between adapting and readapting become more efficient at the transition (Cunningham and Welch 1994; Martin et al. 1996; Welch et al. 1998). These dual-state adaptations are consistent with a cue-based retrieval of the prism mapping, such that training is dependent on error detection but retrieval relies on the sensation or knowledge of putting on prism goggles. Dual-state adaptations have also been shown in humans for perception of apparent concomitant motion, where states were associated with the frequency of head movement (Dumontheil et al. 2006), and for transitions between refractive lenses such as spectacles and contacts (Tuan and Jones 1997).

A more specific term for such phenomena is context-specific adaptation (CSA): the ability to maintain two different adapted states for a given behavioral response. Here, each state is associated with a specific set of cues (the context), and subjects switch between adapted states immediately upon a change in the cues-without the need to detect an error or to readapt (Shelhamer and Zee 2003). CSA has been demonstrated in monkeys making visually guided arm movements, where the viewing eye served as the context cue (Lewis and Tamargo 2001; Lewis 2003). Subjects can also be trained to exhibit CSA for VOR gain (monkeys: Lewis 2003; Yakushin et al. 2000) and phase (humans: Kramer et al. 1998), using a variety of sensorimotor cues such as vergence angle and head orientation to signal the appropriate adapted state. Similar mechanisms might be in play in primates during compensatory eye movements, when the gain of the VOR varies with viewing distance (Schwarz and Miles 1991). Studies of both rotational (Snyder et al. 1992) and translational (Semrau et al. 2006) responses reported that rescaling of VOR amplitude preceded changes in vergence angle, consistent with VOR modification that was driven by a central sensorimotor signal rather than the detection of errors. However, we might imagine that the associations between a given context (i.e., viewing distance) and the appropriate VOR output were initially learned using an adaptive process, and then became associated with specific internal cues (i.e., motor commands) via CSA.

In addition to sensory and motor cues, the subject's internal state, previous conditioning, and behavioral goals can also trigger a change in the desired sensorimotor mapping between sensory processing and motor plant. For example, the gain of the VOR in darkness can be modulated by changes in mental set evoked by experimenter instructions (for example, distraction with mental math or imagining a head-fixed target; Baloh et al. 1984; Barr et al. 1976). Arousal can also have a strong impact on VOR gains. In fact, pigeon VOR gains were significantly higher when amphetamine injections were used to artificially maintain arousal level (Anastasio and Correia 1988; Gioanni 1988b). Head response (VCR) gains were not affected or exhibited a smaller effect. However, head response gains have been shown to be affected by mental set (Keshner 2000).

These examples demonstrate that even relatively simple sensorimotor processing can be modified by the conditions under which it is performed. What, if anything, is special about state-dependent responses during flight? We have already emphasized that although adaptive modifications are driven by feedback about performance (i.e., error detection), changes in the VOR, VCR, and tail response during flight are driven by sensorimotor cues such as airflow that are unrelated to performance. Adapted states can become associated with sensorimotor cues (via CSA) so that these cues can trigger anticipatory changes in behavior independent of feedback, as has been shown for the VOR using head orientation and vergence angle as cues. However, these cues vary along a continuous scale, as do the associated changes in reflex performance. Contrast this with the difference between default and flight states in the current study, where there is no apparent gradation of the behavioral state-the bird is either in flight or not, and these two conditions are clear and discrete. It is feasible that both continuous and discrete shifts in state reflect the same fundamental mechanism, implemented using different kinds of cues, but it is also possible that these represent two fundamentally different ways of modifying motor output-one that shifts output along a continuum based on properties of the motor plant and another that associates specific sensory cues with jumps between distinct behavioral states.

Simulated flight: what is the state?

What are the critical features of the simulated flight condition that signify a distinct behavioral state? Before we define what flight state is, we should consider what flight state is not. Based on our observations, flight state is not a purely cognitive (top-down) process. In each experiment, the bird's body was fixed in space during the transition into simulated flight, and no active change in motor behavior occurred (i.e., the bird did not initiate flight). Second, the transition between behavioral states could not have been driven by an adaptive process requiring error detection. Rotational stimuli were identical across conditions, so there was no manipulated change in performance that the animal could use to distinguish one state from another. Finally, although increased arousal appears to be a component of flight state, we do not believe that changes in arousal alone were sufficient to explain the observed changes in posture and performance, since blower noise associated with airflow did not increase the behavioral response.

The flight state could be described as a locomotor condition. The postural changes evoked by frontal airflow place the wings, legs, and tail in positions that are appropriate for flight-related locomotion (Pennycuick 1968). Previous studies have shown that locomotor state can have a profound effect on posture-stabilizing reflexes (Xiang et al. 2008). In fact, although flying pigeons have the best head-in-space stability, walking pigeons exhibit higher head response gains to visual motion than standing or fully restrained pigeons (Maurice et al. 2006). Others have considered the flight state to be an internal global sensory-motor pattern (Maurice et al. 2006), signified by a combination of cues including frontal airflow and lack of terrestrial substrate. They speculate that the brain recognizes the change to flight state and then modulates eye, head, and tail reflexes accordingly. This description of transitions in and out of the flight state is similar to the definitions of dual-state and context-specific adaptation. In fact, models of dual-state adaptation suggest the coexistence of predictive (cue-based) and postdictive (error-based) cerebellar mechanisms for switching between internal models (Imamizu and Kawato 2008; Lee and Schweighofer 2009). Flight as described above might constitute an internal model of a locomotor state.

Neural substrates of state-dependent vestibular responses.

To implement the state-dependent responses observed in the current study, the brain region(s) responsible must receive convergent multisensory input and then modulate output in such a way as to broadly influence performance across a variety of motor responses. The cerebellum could likely be implementing state determination and modulating vestibular neurons responsible for generating eye, head, and tail responses to motion. Others have proposed that the cerebellum modifies reflexes (such as vestibular eye, head, and limb responses) according to behavioral context by reweighting or switching between populations of brainstem neurons (Shelhamer and Zee 2003). Each microzone in the cerebellar cortex receives a specific array of multisensory inputs, including somatosensory, visual, and vestibular signals that could regulate transitions between behavioral states according to specific combinations of sensorimotor cues (Manzoni 2005, 2007). In turn, cerebellar neurons project heavily to the vestibular nuclei (pigeons: Arends and Zeigler 1991). Previous studies have found vestibular neurons in the brainstem whose activity is modulated by behavioral state. Vestibulospinal neurons in the lateral vestibular nucleus increase their spontaneous activity during locomotion and respond rhythmically during the locomotor pattern (Orlovsky 1972; Orlovsky and Pavlova 1972; Marlinsky 1992). Specifically in pigeons, electrical stimulation of the vestibular labyrinth in a resting animal evokes activity in the vestibular brainstem and cervical spinal cord motoneurons but not in the lumbosacral spinal cord. However, direct electrical stimulation of the vestibular brainstem does excite lumbosacral motoneurons (Rabin 1973, 1974, 1975a, 1975b). These previous electrophysiological results and the behavioral results we have shown here strongly suggest some state-specific gating of vestibular inputs to brainstem vestibulospinal neurons, and further study is needed to determine whether this is the neural substrate for state-dependent vestibular behaviors in birds.

GRANTS

This work was supported in part by funding from National Institutes of Health Grants DC-010373, DC-007618, DC-006913, and DC-009734.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the author(s).