Sensory convergence in the parieto-insular vestibular cortex

doi:10.1152/jn.00731.2013 111:2445-2464, 2014. First published 26 March 2014;J NeurophysiolMichael E. Shinder and Shawn D. NewlandscortexSensory convergence in the parieto-insular vestibular

You might find this additional info useful...

82 articles, 27 of which can be accessed free at:This article cites /content/111/12/2445.full.html#ref-list-1

including high resolution figures, can be found at:Updated information and services /content/111/12/2445.full.html

can be found at:Journal of Neurophysiologyabout Additional material and information http://www.the-aps.org/publications/jn

This information is current as of November 3, 2014.

American Physiological Society. ISSN: 0022-3077, ESSN: 1522-1598. Visit our website at http://www.the-aps.org/.(monthly) by the American Physiological Society, 9650 Rockville Pike, Bethesda MD 20814-3991. Copyright © 2014 by the

publishes original articles on the function of the nervous system. It is published 12 times a yearJournal of Neurophysiology

on Novem

ber 3, 2014D

ownloaded from

on Novem

ber 3, 2014D

ownloaded from

http://www.the-aps.org/publications/jn

Sensory convergence in the parieto-insular vestibular cortex

Michael E. Shinder and Shawn D. NewlandsDepartment of Otolaryngology, University of Texas Medical Branch, Galveston, Texas

Submitted 9 October 2013; accepted in final form 21 March 2014

Shinder ME, Newlands SD. Sensory convergence in the parieto-insular vestibular cortex. J Neurophysiol 111: 2445–2464, 2014. Firstpublished March 26, 2014; doi:10.1152/jn.00731.2013.—Vestibularsignals are pervasive throughout the central nervous system, includingthe cortex, where they likely play different roles than they do in thebetter studied brainstem. Little is known about the parieto-insularvestibular cortex (PIVC), an area of the cortex with prominentvestibular inputs. Neural activity was recorded in the PIVC of rhesusmacaques during combinations of head, body, and visual targetrotations. Activity of many PIVC neurons was correlated with themotion of the head in space (vestibular), the twist of the neck(proprioceptive), and the motion of a visual target, but was notassociated with eye movement. PIVC neurons responded most com-monly to more than one stimulus, and responses to combined move-ments could often be approximated by a combination of the individualsensitivities to head, neck, and target motion. The pattern of visual,vestibular, and somatic sensitivities on PIVC neurons displayed acontinuous range, with some cells strongly responding to one or twoof the stimulus modalities while other cells responded to any type ofmotion equivalently. The PIVC contains multisensory convergence ofself-motion cues with external visual object motion information, suchthat neurons do not represent a specific transformation of any onesensory input. Instead, the PIVC neuron population may define themovement of head, body, and external visual objects in space andrelative to one another. This comparison of self and external move-ment is consistent with insular cortex functions related to monitoringand explains many disparate findings of previous studies.

insula; multisensory; self-motion; sensory integration; spatial orien-tation

THE INSULAR CORTEX has for a long time been difficult to studyand has therefore been poorly understood. The absence ofclarity can also be attributed to the insular cortex itself. Unlikeother brain regions, insular damage does not always producedefined or lasting deficits in behavior (Cereda et al. 2002;Duffau et al. 2006; Baier et al. 2013). The advent of brainimaging has provided many new insights into insular function.The posterior insula has recently been shown to be activatedwith a number of perceptual conditions including out-of-bodyexperience, subjective awareness of bodily states (Simmons etal. 2012), subjective imagery of body orientation (Arzy et al.2006), body ownership (Karnath and Baier 2010), and theinterpretation of others’ actions (Abreu et al. 2012). Theculmination of brain imaging has led to the view that the post-erior insula contributes to interoceptive awareness or morespecifically monitoring body states (zu Eulenburg et al. 2013).

This same cortical area is also viewed as the primaryvestibular sensory cortex (Eickhoff et al. 2006; Lopez et al.2012a). This perception of the posterior insula has a notable

history, as stimulation in the vicinity of the posterior insulaproduces the sensation of movement (Penfield and Boldrey1937; Kahane et al. 2003). Sensory information may be essen-tial to interoceptive monitoring, but it is unclear whether theposterior insula also acts in sensory perception.

Physiological and anatomical studies in primates haveshown that the posterior insula is anatomically connected to thevestibular nuclei and is defined by physiological responses tovestibular stimulation (the posterior insula is better known inthe primate as the parieto-insular vestibular cortex, PIVC)(Akbarian et al. 1988; Grusser et al. 1990b). However, theposterior insula is not always activated by moving stimuli thatproduce perceptions of self-motion (physiology: Chen et al.2010; imaging: Becker-Bense et al. 2012). Thus functions ofvestibular signals in this region and how these signals areprocessed remain open questions.

The posterior insula is multimodal, where vestibular signalsconverge with at least visual and proprioceptive information(physiology: Akbarian et al. 1988; Grusser et al. 1990b; imag-ing: zu Eulenburg et al. 2013). Proprioceptive sensitivity tostimulation of the neck, trunk, limbs, hands, and feet is foundin the posterior insula (physiology: Friedman et al. 1986;Robinson and Burton 1980; imaging: Lotsch et al. 2012;Macey et al. 2012). Furthermore, in imaging studies, somevisual stimuli are shown to evoke responses in the posteriorinsula (Kleinschmidt et al. 2002; Dupont et al. 1994; Sunaert etal. 1999; Antal et al. 2008). The posterior insula is also activewhile human subjects track moving objects (Indovina et al.2005; Nagel et al. 2006; Lindner et al. 2006) but not whilefixating a stationary object (Asahi et al. 2006). This combina-tion of responses is consistent with monitoring objects in theenvironment. However, although visual responses have beenseen in neural responses to rotations of the visual field (Grusseret al. 1990b), posterior insular neurons did not represent opticflow (Chen et al. 2010). Furthermore, the response of posteriorinsula neurons to visual targets has not been previously tested.To be consistent with monitoring one’s motion relative toobjects of interest, the posterior insula should contain neuralsignals that carry visual object information along with theinformation about self-motion. Vestibular signals carry infor-mation about the movement of head in space. Proprioceptivesignals represent movement of the body elements relative toone another.

To date, much of the speculation about posterior insulafunction is based on fMRI and lesion data. Only a handful ofpapers describing neuronal recording in the posterior insulahave been published. We explored responses of single neuronsduring combinations of head, body, and visual target move-ments in the horizontal plane. We sought to understand theinteraction of visual object and self-motion signals at the singlecell level in the posterior insula and to explore whether the

Address for reprint requests and other correspondence: S. D. Newlands,Univ. of Rochester Medical Center, Dept. of Otolaryngology, 601 ElmwoodAve., Box 629, Rochester, NY 14642 (e-mail: [email protected]).

J Neurophysiol 111: 2445–2464, 2014.First published March 26, 2014; doi:10.1152/jn.00731.2013.

24450022-3077/14 Copyright © 2014 the American Physiological Societywww.jn.org

on Novem

ber 3, 2014D

ownloaded from

mailto:[email protected]

mailto:[email protected]

posterior insular neurons might differentially encode active(self-generated) vs. passive (imposed) movements. Our resultsfound that PIVC neurons respond to combinations of head,body, and visual object motion. Although these neurons dem-onstrated heterogeneity in responses to multimodal stimuli, theresponses were consistent with perceptual monitoring of mov-ing objects in the environment and relative to our own move-ments whether passively applied or self-generated.

METHODS

Subjects and surgery. One female and one male rhesus macaque(Macaca mulatta), each weighing �4 kg, were used in these experi-ments. The details of the surgical implantation of the head restraint,recording chamber, and eye coils have been previously published(Newlands et al. 2009). Initially, a head-restraint post was implantedin dental acrylic that was attached to the monkey’s skull via surgicallyimplanted, transcranial stainless steel T-bolts. A recording well (cus-tom built, stainless steel, 1.5-cm diameter) was also attached to theskull embedded in the same acrylic. The recording grid that held theelectrode in the chamber allowed sampling a 10-mm circular anatom-ical region at 1-mm intervals. The chamber was implanted over thePIVC using stereotaxic coordinates, in which the chamber was angledlaterally by 25°. The electrode chamber was centered such that theelectrode at the center of the grid entered the brain at 18 mm lateralof midline and 5 mm anterior to the interaural axis. At a secondsurgery, 1.5-cm diameter three-turn stainless steel eye coils wereimplanted under the conjunctiva of one eye, and the eye coil lead wirewas tunneled subcutaneously to the head implant where the leads weresoldered to connectors. A second coil was attached to the head-restraint post, but not surgically implanted, and allowed the monitor-ing of head movement in space. All surgical procedures were per-formed in accordance with institutional and National Institutes ofHealth guidelines and under a protocol approved by the InstitutionalAnimal Care and Use Committee at the University of Texas MedicalBranch. At the termination of the experiments, both animals wereeuthanized by deep phenobarbital anesthesia followed by transcardiacperfusion with heparinized saline and 4% paraformaldehyde.

Equipment. For the experiments described, the monkeys sat in acustom-made Plexiglas primate chair. The chair restrained their shoul-der movements using a Plexiglas bar across the upper chest, limitingtheir ability to turn their shoulders. For experiments, the chair wasmounted to a computer-controlled 80 ft/lb Neurokinetics motor thatprovided passive yaw rotation. The head-restraint post was a roundstainless steel tube with a square interior into which a bar could beinserted. The post was implanted at the top of the skull on the midline,along the axis of neck rotation. The head-restraint post was connectedvia a coupling bar to a computer-controlled motorized clutch (1.47ft/lb motor, MicroMo/Faulhaber) that itself was attached to the pri-mate chair. Activating the clutch mechanically coupled the head to themotor, allowing passive head on neck yaw rotation, whereas deacti-vating the clutch allowed for active, head-free, rotation of the head inthe horizontal plane. The animals could only move their head or havetheir head rotated in the horizontal plane � 60° right or left of thestraight ahead orientation (0°) although, for the tasks described here,that full range was not utilized, and the targets used were limited to �40°. Head movements were primarily restricted to rotation in thehorizontal plane; however, as a matter of comfort and safety, and toallow a more natural yaw during head movement, which includedsmall movements out of the yaw plane, there was a large springsituated between the bar coming from the head-restraint post and theshaft of the motorized clutch that allowed small amounts of transla-tion, pitch, and roll. The primate chair and motion control were setinside of a 3-ft three-coil magnetic field system (C-N-C Engineering).Visual targets were presented on a video screen (HD-P61R1U, JVC)in the horizontal plane at the level of the straight-ahead eye positionof the animals. Targets were moved across a range of viewing angles

up to � 24°. During experimentation, the room lights were turned off,and the animals sat in an apparatus that had been draped in lightblocking material on all sides around the outside of the 36-inch coilform. For nontargeted trials, the video screen was turned completelyoff to produce complete darkness. A cartoon diagram of the apparatusis shown in Fig. 1.

Alignment of the eye coil and target position was maintained bycoordinated calibrations. The eye coil was calibrated using a modifi-cation of the total field-calibration technique (http://sine.ni.com/cs/app/doc/p/id/cs-604). In short, the eye and head coils were calibratedpostimplantation. The animal was pitched and rotated to face a rangeof different directions in three dimensions. The different orientationsof the eye and head coils provided the coil signal across differentorientations, covering an even distribution along the three orthogonalplanes of three-dimensional space. The three-dimensional spherecreated by the recorded eye coil data at each orientation was then fitalong each of the coordinate axes, and the parameters defining thecircle or oval in each plane were used to calibrate input of the eye coilsto accurately predict the orientation of the coil for any three-dimen-sional orientation. This method makes no assumptions about the size,shape, or number of turns in the implanted coil. The head and chairmotor servo-system was then aligned with the coil calibrationby matching the head coil signal to the motor position feedback infor-mation. Target calibration was performed by aligning the targetmotion with a laser attached to the head-restraint coupling. The offsetsin the alignment of the target with the eye coil were then removed byhaving the animal fixate on a stationary target while head-restrainedand stationary.

The computer control system coordinated the visual display andmotors that produced head and body movement, and the systemmonitored and recorded physiological data from head and eye coilsand triggered rewards to the animal based on programmed criteria.The custom software was written in LabView (National Instruments)and was composed of individually executable protocol modules. Eachmodule contained the capacity to free up or passively rotate the head,to passively move the chair holding the monkey, or to create andmanipulate visual targets. Each module also monitored the head andeye coils to determine eye, head, and gaze information. The module

Fig. 1. Schematic representation of the experimental apparatus with indepen-dently controllable head rotation, trunk rotation, and visual target.

2446 SENSORY CONVERGENCE IN THE PIVC

J Neurophysiol • doi:10.1152/jn.00731.2013 • www.jn.org

on Novem

ber 3, 2014D

ownloaded from

http://sine.ni.com/cs/app/doc/p/id/cs-604

http://sine.ni.com/cs/app/doc/p/id/cs-604

then made a decision based on the available data as to which moduleto call next. By changing global variables specific to each animal (e.g.,target window size, time on target, number of rewards, rewardduration, etc.), the sequence of the execution of protocols could bedirected by the accurate performance of the animal. This also allowedfor individualized training on the performance of protocols.

Training. Initial training for each animal was head-controlledtraining of gaze to a 4-mm red target. Training was reinforced by anutritionally appropriate liquid food reward containing 1.08 calo-ries/ml developed in conjunction with the supervising veterinarianthat contained fruit, Pediasure (Abbott Laboratories), fish oil, anduncaffeinated Power Gel (Nestle). The size of the reward windowaround the target and time on target required to evoke a reward wereadjusted to make the task more difficult as it was learned. Once thattask was learned, the head was freed to move in the horizontal plane,and animals easily learned to perform the same gaze task with thehead free. The animal was then trained to follow a 7-mm greenhead-related target by using feedback to reinforce the desired headmovement. With the use of a second smaller green target, visualfeedback was provided about the current orientation of the head.When the head coil feedback target and head position target werealigned within the range of the target window, the animal wasrewarded. The red gaze-related and green head-related target taskswere learned quickly. The animals were able to accurately pursueusing either type of visually driven target in over 90% of the trialsafter 2 wk of postsurgical training.

Data collection. Once the animals were well trained on the tasks,we began recording neural activity from the posterior insular cortexusing �1 MOhm tungsten microelectrodes (Fredrick-Haer). The elec-trodes were introduced into the brain by means of a guide tubepositioned in a grid system that spanned the cortical region of interest.Guide tubes were lowered only 1–2 mm into the brain to enablecharacterization of neurons encountered both within and around thePIVC region so that a more complete survey could be performed. ThePIVC region was initially surveyed using stereotaxic coordinates,using vestibular sensitivity as a confirmatory characteristic as inprevious studies (Akbarian et al. 1988; Grusser et al. 1990a, 1990b).This functional region is not well defined by anatomical boundariesbut is defined by the presence of physiological responses to vestibularstimulation (Akbarian et al. 1988; Grusser et al. 1990b) and connec-tionally by projections to the vestibular nuclei (Guldin et al. 1993;Akbarian et al. 1994). The location of the electrode chamber alloweda brief survey of the surrounding cortical regions to better define thelocation of the PIVC during experimentation. The responses of theneighboring intraparietal cortex could be used to identify the insularcortex ventral to the intraparietal sulcus when surveying the anteriorportions of the PIVC. In the caudal portions of the PIVC, someelectrode tracts were continued through the white matter deep to thePIVC, and responses of the dorsal medial superior temporal cortex(MST) ventral to the PIVC could be noted using vestibular and visualstimulation (Shinder and Newlands 2005). The search stimulus duringexploration of the PIVC included a set of stimulus protocols thatincluded head-free, gaze-target-driven head-on-neck, trunk-under-head, and whole-body rotations (protocols 12, 13, and 14 in the activesection of Table 1). Data were collected from cells that produced aneural response to some form of head or neck motion. This includedresponding to whole-body rotation, head-on-neck rotation, or trunk-under-head rotation either individually or in combination. The stimulisubsequently presented during the characterization of each cell fo-cused on the stimulus modalities (head, neck, or target) that elicitedresponses. Given the mechanical disturbance of the various differentmotion stimuli, and the necessity of switching back and forth fromactive to passive movement, many neural recordings ended beforecompletion of the full complement of available stimulus conditions.On the other hand, neurons were frequently held long enough thatmultiple trials were performed for many of the protocols.

During recording of each neuron, several channels of data werecollected. The bandpass (100 Hz to 10,000 kHz)-filtered signal fromthe electrode was collected at 40 kHz. The horizontal and vertical eyeand head position signals were collected from the three-dimensionalfield-coil system at 1 kHz. The eye and head position signals from thefield-coil system were differentiated to produce eye and head velocityand subtracted to produce a signal of the eye movement relative to thehead (from eye in space to eye in head). Similarly, all of the motorcontrol signals, the head potentiometer signals, the position andvelocity feedback signals from both the head and rotator motors, andthe target position, size, and color signals were all collected at 1 kHz.The signal originating from the rotator motor represented the chair, trunk,or whole-body rotation. Target velocity was derived from the targetposition, and target type could be denoted by the target color.

Protocols. The target, chair (trunk), and head rotation were coor-dinated to produce the motion stimuli used in this study. All of theprotocols in these experiments were yaw sinusoids with the headpositioned with the stereotaxic plane horizontal, at one frequency (0.2Hz), and at one of two peak velocities (15 or 30°/s). All trunk rotationswere passive, but head-on-neck rotations could be passive or activedepending on the state of the clutch and the presence of the target.Several different types of motion stimuli combinations were used tocharacterize PIVC neurons. Head-controlled stimuli were performedby activating the clutch fixing the head to the head motor. Movementsof the lower body were limited but not completely restricted by thechair. During head-controlled testing, three basic protocol stimuliwere used, each stimulating only one of the sensory modalities studiedin these experiments: passive whole-body rotation of the head andtrunk in the dark at 15 or 30°/s (whole-body dark), trunk rotationunder a stationary head in the dark 15 or 30°/s (trunk-only dark), andmovement of a visual target at 15°/s (target only). The use of targetmotion for smooth pursuit was tested at 30°/s during training, but themonkeys did not consistently maintain accurate tracking of the targetthroughout the stimulus presentation on all trials. Therefore, to eval-uate consistently accurate behavior, the 30°/s stimulus was not usedduring recordings. These protocols are categorized by movement ofthe chair motor, head motor, and target in Table 1 (protocols 1, 2, and3, respectively).

A second set of protocols combined at least two stimuli (vestibularstimulation, neck stimulation, or a visual target) while the head wasstill restrained. These included four relatively simple protocols (num-bers corresponding to Table 1): 4) passive whole-body rotation of thehead and trunk at 15 or 30°/s with a gaze target that moved with theanimal (whole-body with target), 5) trunk rotation at 15 or 30°/s undera stationary head with a head- and space-fixed gaze target straightahead (trunk-only with target), 6) head-on-neck rotation above astationary trunk in the dark at 15 or 30°/s (head-on-neck dark), and 7)head-on-neck rotation at 15 or 30°/s with a target that moved with thehead (head-on-neck with target). There were two other conditions inwhich the gaze target was combined with chair rotations. The targetand chair motion were either synergistic or opposing and at differentspeeds. Synergistic combinations of chair and target motion wereperformed by rotating the head 0.2 Hz at 15°/s over a trunk moving at15°/s in synchrony while the target was moved across the screen in thesame direction as the head at 0.2 Hz but at 30°/s (Table 1, protocol 8).During opposing combinations of target movement with chair-rotationtasks, the target and chair rotated in opposite directions at 0.2 Hz at15°/s while the head moved on the trunk at 30°/s such that the headtracked the target (Table 1, protocol 9). Similar protocols wereadministered where the head was moved on the neck at 7.5°/s and ahead target was moved synergistically (in phase) at 15°/s or inopposition at 7.5°/s (out of phase) with the head movement (protocols10 and 11).

In addition to the unimodality and multimodality head-controlledprotocols listed above, head-free (in the yaw plane) protocols wereadministered that matched five of the multimodality head-controlledprotocols [the head-on-neck passively driven without a target, or with

2447SENSORY CONVERGENCE IN THE PIVC


on Novem

ber 3, 2014D

ownloaded from

an in- or out-of-phase gaze target at 7.5°/s (protocols 6, 10, and 11)having no head-free analog]. During the active head movement testingprotocols, two different targets were used. In one condition, the targetrequired the animal to maintain gaze on the target to receive a reward(a red target). In another condition, the target required the animal tomaintain their head position on the target to receive a reward (a greentarget was used; and a separate, smaller green target was used torepresent real-time head position to help guide behavior). Theseprotocols included (number corresponding to protocols listed in Table1): 12) whole-body rotation of the trunk with a gaze or head target thatmoved with the animal (head-free whole-body rotation with a target

referenced to the chair signal), 13) trunk (chair) rotation under anunrestrained head with a space-fixed gaze or head target straight ahead(head-free trunk rotation with spatially fixed target), 14) active head-on-neck rotation pursuing a gaze or head target (active head-on-neckwith target), 15) synergistic gaze or head target and chair motion(target at 30°/s and chair at 15°/s), and 16) opposing gaze or headtarget and chair motion (both at � 15°). For all five of these head-freeprotocols, the gaze target and eye target trials (each 6 cycles at 0.2 Hz)were always paired with a 3-s break between the gaze target (first) andthe head target (second). The first three of the head-free protocols(whole-body rotation with a visual target aligned to the chair signal or

Table 1. Protocols used for testing PIVC neurons combined trunk, head-on-neck, and target angular rotations in the horizontal plane

Condition Chair Motor Head MotorTarget in

SpacePhase: Chair

re HeadPhase: Chair

re TargetPhase: Head

re Target Rotation Protocol

Head controlled (unimodal)1a �15°/s 0°/s NA 0° NA NA Passive whole body �15°/s in dark1b �30°/s 0°/s NA 0° NA NA Passive whole body �30°/s in dark2a �15°/s �15°/s NA 180° NA NA Passive trunk under head �15°/s in dark2b �30°/s �30°/s NA 180° NA NA Passive trunk under head �30°/s in dark3 0°/s 0°/s �15°/s NA NA NA Gaze target only �15°/s

Head controlled (multimodal)4a �15°/s 0°/s �15°/s NA 0° 0° Passive whole body �15°/s with target4b �30°/s 0°/s �30°/s NA 0° 0° Passive whole body �30°/s with target5a �15°/s �15°/s 0°/s 180° NA NA Passive trunk under head �15°/s with target5b �30°/s �30°/s 0°/s 180° NA NA Passive trunk under head �30°/s with target6a 0°/s �15°/s NA NA NA NA Passive head on neck �15°/s in dark6b 0°/s �30°/s NA NA NA NA Passive head on neck �30°/s in dark7a 0°/s �15°/s �15°/s NA NA 0° Passive head on neck �15°/s with target7b 0°/s �30°/s �30°/s NA NA 0° Passive head on neck �30°/s with target8 �15°/s �15°/s �30°/s NA 0° 0° Passive head with target �30°/s, synergistic

trunk �15°/s9 �15°/s �30°/s �15°/s NA 180° 0° Passive head with target �15°/s, opposed

trunk �15°/s10 0°/s �7.5°/s �15°/s NA NA 0° Passive head on neck �15°/s with

synergistic target �30°/s11 0°/s �15°/s �7.5°/s NA NA 180° Passive head on neck �15°/s with opposed

target �7.5°/sHead free

12a �15°/s Gaze �15°/s 0° Trunk and gaze target aligned �15°/s12b �30°/s Gaze �30°/s 0° Trunk and gaze target aligned �30°/s12c �15°/s Head �15°/s 0° Trunk and head target aligned �15°/s12d �30°/s Head �30°/s 0° Trunk and head target aligned �30°/s13a �15°/s Gaze 0°/s NA Trunk �15°/s with gaze target stationary in

space13b �30°/s Gaze 0°/s NA Trunk �30°/s with gaze target stationary in

space13c �15°/s Head 0°/s NA Trunk �15°/s with head target stationary in

space13d �30°/s Head 0°/s NA Trunk �30°/s with head target stationary in

space14a 0°/s Gaze �15°/s NA Trunk stationary with gaze target moving in

space �15°/s14b 0°/s Gaze �30°/s NA Trunk stationary with gaze target moving in

space �30°/s14c 0°/s Head �15°/s NA Trunk stationary with head target moving in

space �15°/s14d 0°/s Head �30°/s NA Trunk stationary with head target moving in

space �30°/s15a �15°/s Gaze �30°/s 0° Synergistic trunk �15°/s with gaze target

�30°/s15b �15°/s Head �30°/s 0° Synergistic trunk �15°/s with head target

�30°/s16a �15°/s Gaze �15°/s 180° Opposed trunk �15°/s with gaze target

�15°/s16b �15°/s Head �15°/s 180° Opposed trunk �15°/s with gaze target

�15°/s

Rotations were performed with self-generated or motor-imposed head-on-neck rotations, but all trunk rotations were passive (driven by the chair motor). Twostimulus amplitudes were used, and stimuli could be presented in the dark, with a gaze (eye)-related target, or with head (head on neck)-related targets. PIVC,parieto-insular vestibular cortex.



on Novem

ber 3, 2014D

ownloaded from

fixed in space and pursuit of a gaze or head target with the trunkstationary) were performed at either 15 or 30°/s peak velocity. All ofthese protocols are summarized in Table 1. In most cases, particularlyfor the head-free protocols, the protocols were run on each neuronseveral times, often up to 30 min apart, to investigate the stability ofthese responses over time.

Data analysis. Many elements of the data analysis have beendescribed elsewhere (Newlands et al. 2009). In brief, the recordedneural signal was triggered offline using a combination of time-amplitude windows and spike-waveform-principle components tocreate a spike time stamp. Because many PIVC cells are slow firing,we used a Kaiser window filter (cutoff frequency 0.4 Hz, bandpass 5.0Hz) to convert the time stamp from the spikes into a continuousfunction (Cherif et al. 2008), which was then compared with thedigitized stimuli (eye position, head position, and chair-rotator posi-tion) to calculate the response sensitivity or gain (the peak-to-peakoutput of the neuron/peak-to-peak stimulus in spikes per second perdegree per second) and phase (the relative timing of the peak of theresponse to the peak of the stimulus). Accurate fitting of the neuralresponses that rectified during rotation in one direction required thatsensitivity and phase were computed using a nonlinear least-squaresalgorithm (Levenberg-Marquardt method, see Newlands et al. 2009).Gains were computed relative to horizontal head velocity duringwhole-body and head-on-neck rotations and to chair rotational veloc-ity during trunk-only rotations. As mentioned above, individual trialswere comprised of six cycles of continuous movement (30 s), whichwere repeated throughout the overall period during which the cell wasrecorded. The sensitivities and phases from each of the trials wereaveraged together to compute response sensitivity and phase of thecell for protocols that were administered more than once on a givencell. For all neurons, we determined whether the firing response wascorrelated with the stimulus by calculating the Rayleigh coefficient,a circular statistical measure of correlation between the stimulus andresponse that incorporates information about direction to eliminate theinfluence of repetitively moving through the same orientations duringrotation. We used only data where the calculated Rayleigh coefficientwas statistically significant at the P � 0.05 level (Mardia 1972). Whenactivity was not found to be statistically significantly correlated withthe stimulus, the neuron was considered unrelated to that stimulus.

Often, it was useful to compare the sum of the responses toindividual stimuli with the response to a combined stimulus usingthose stimuli. To do this, the firing-rate responses to the individualstimuli had to be added together to form a prediction of the responseto the combined stimulus. The gain and phase of the predictedstimulus either computed as the simple vector sum of the individual

responses or the averaged response over the six recorded cycles forindividual stimuli were combined to predict the resultant response.The actual response to combined stimulation was compared with thepredicted response using a mean-error measure. All the calculationsand statistics defined above were done using SPSS, Excel, orLabView.

Histology. Current data were confirmed by histological evaluationsof electrolytic lesions produced at the end of selected electrode tracts.The perfused brains were first dehydrated by soaking in 20% su-crose/4% paraformaldehyde solution for several weeks. The area ofthe insular cortex was cut in 50-�m sections. The electrolytic lesionsand scarring from electrode tracks were located in the area of theinsular cortex expected to contain PIVC.

RESULTS

We encountered 978 neurons in the PIVC region of the lefthemisphere in two rhesus monkeys trained to maintain theirgaze or head alignment on a visual target while their head andtrunk were independently manipulated in the horizontal plane(474 from monkey 1 and 504 from monkey 2). The region ofthe posterior insula was fully explored (Fig. 2), but head-movement-sensitive neurons were only found in subscribedsubregions similar to that found previously (Akbarian et al.1988; Grusser et al. 1990a, 1990b). Of these, 222 (23%)responded to either passive (head restrained) or active (headfree to move independently of the chair) horizontal head orneck rotation (117 from monkey 1 and 105 from monkey 2).Using head and trunk rotation, either passively or with the headunrestrained, for search and characterization methods, 74 PIVCneurons were recorded long enough for the presentation of asufficient number of stimulus conditions to be adequatelycharacterized (51 from monkey 1, and 23 from monkey 2).Thus 74 of the 222 cells responsive to some form of headmovement were fully characterized (33.3%).

Recording sites. The PIVC straddles the posterior border ofthe granular-insular cortex (physiology: Akbarian et al. 1988;Grusser et al. 1990a, Guldin et al. 1992; imaging: Bottini et al.2001; Petit and Beauchamp 2003; stimulation: Kahane et al.2003) but is best defined by physiological responses to vestib-ular stimulation (Akbarian et al. 1988; Grusser et al. 1990b).However, the dorso-caudal border of posterior insula cangenerally be used as an anatomical landmark to define the

MST

Dorsal

Ventral

CaudalRostral

LF

STS

IPS

Monkey 1 Monkey 2

GI

RIS2

LIPVIP

TPO

AC

TP

Fig. 2. Location of recorded neurons on surface diagram ofunfolded cortex. The recording locations in each of the 2monkeys are presented in separate maps. The posteriorinsular cortex is displayed as if the lateral fissure had beenopened up and flattened out. The gray regions representgyri normally on the outer surface of the brain. The sulci(LF, lateral fissure; IPS, intraparietal; STS, superior tem-poral) and cortical regions (GI, granular insular; RI, retro-insular; AC, auditory; LIP, lateral intraparietal; MST, me-dial superior temporal; S2, secondary somatosensory; TP,temporoparietal; TPO, temporoparietal occipital; VIP, ven-tral intraparietal) are labeled separately. The 74 neuronsrecorded in the insular cortex and presented in subsequentanalyses are denoted by circles. The neurons in the VIP andMST were presented previously (Shinder and Newlands2005). Open circles, 1 cell found at that site; shaded circles,2 to 5 cells found; solid circles, 6 or more cells. Squareshave a similar fill and color scheme and represent motion-responsive neurons found in neighboring brain regions.These neurons are not analyzed here.



on Novem

ber 3, 2014D

ownloaded from

location of the PIVC (Eickhoff et al. 2006). That part of theposterior insula is a multimodal area representing both vestib-ular and proprioceptive stimuli (Akbarian et al. 1988; Grusseret al. 1990b; Downar et al. 2000; Bottini et al. 2001; Petit andBeauchamp 2003; Fasold et al. 2008; Mazzola et al. 2012;Ferre et al. 2012; zu Eulenburg et al. 2013). In this study,electrolytic lesions were confirmed in the caudal retroinsularcortex about 2 mm posterior to the interaural axis, about 18mm lateral, and about 7 mm ventral from the surface of thebrain. Guide tube scarring could be seen from just above theretroinsular cortex at 3 mm posterior of the interaural axis to 11mm anterior to interaural above the granular insular cortex.Recorded cell locations overlapped the gray matter defined bythe histology from the granular insular cortex to the retroinsu-lar cortex. Exploration of the region produced neural record-ings from neurons encountered in the electrode tract on the wayto the insular cortex or in electrode tracts surrounding theinsular cortex. In this way, neurons from the ventral intrapari-etal and medial superior temporal cortices were found. Beforehistological confirmation, these areas were defined by theirstereotaxic location and confirmation of sulci locations bygray/white matter patterns elucidated by electrophysiologicalsurvey. The stereotaxic location of tracts that yielded theneurons presented in this study is shown in Fig. 2, which is anunfolded view of the insula (after Grusser et al. 1990a, Fig. 3).The following results describe the responses of those 74neurons recorded from the left posterior insula in two mon-keys. The left insula was selected for comparison to previousstudies (Akbarian et al. 1988; Grusser et al. 1990a, 1990b).Because motion signals are similar in the insula across thehemispheres (Chen et al. 2010, 2011, 2013a), comparableresults would be expected for neurons in the right posteriorinsula.

Test-retest reliability. In many neurons, we recorded stimulithat were repeated, from two to five times, often 10–30 minapart using the same stimulus parameters and combinations onthe same neuron. Using shorter presentations and repeatingthem during the recording seemed to maintain the level ofinterest of the animals in the task, and both of the animalsmaintained high levels of on-task performance. Therefore, astimulus was often presented, followed by different, interven-ing stimuli before the first stimulus was repeated again. Thismethod of presenting multiple instances of the same stimulusallowed us to evaluate the stationarity of the response to stimuliof the PIVC neuron. As we discuss in presenting these results,the neurons we recorded were quite heterogeneous in theirresponses to our array of tasks. An important consideration iswhether the array of responses described here are due toinfluences on neural responses that were not controlled and notappreciated such that the same neuron might respond differ-ently at different times. Reliable test-retest responses indicatethat the neural response differences represent heterogeneity inthe neuron population.

To test the consistency of the responses of the neuron, weexamined the test-retest consistency of all cells when thestimuli were presented more than once. From trial to trial,the gain and phase of any given stimulus did not vary [univariateANOVA testing of the neural responses during all stimulipresentations that were repeated found no main effect for theorder of the stimulus presentation; gain: F(1,945) � 0.20, P �0.66; phase: F(1,945) � 0.03, P � 0.87; and no interaction

effect of order by the type of stimulus presented]. Furthermore,the gain and phase from trial to trial was reliable [correlation ofthe first and second presentation of a stimulus (n � 473stimulus pairs, 68 cells), gain: 0.72, phase: 0.95; Pearsoncorrelation of first to third presentation (n � 117 stimuluspairs, 33 cells), gain � 0.75, phase � 0.97]. Therefore, thevariability in the way in which stimuli combine was due tointrinsic sensory mechanisms and not poor reliability in theway in which stimuli were encoded.

Stimulus amplitude. This study focused on the PIVC re-sponse to different modes of sensory stimulation and not onvariations of identical stimuli. In the assessment of responsesto changes in the strength of the stimulus, lower-velocitystimuli tended to produce higher gains. Higher gains were seenat lower velocities regardless of the visual, volitional, orstimulus context in which the rotation occurred (Table 2). Theincreased sensitivity could also be seen in the percentages ofresponsive cells, with larger percentages found at lower veloc-ities. We have the largest volume of comparison data betweenthe 15°/s and 30°/s velocities for the head-free protocols.Vestibular sensitivity decreased on average 0.17 � 0.04 (stan-dard error, S.E.) spikes per second per degree per second as thevelocity increased from 15° (condition 12a) to 30°/s (condition12b) (47 cells for which both velocities were available). Neckrotational sensitivity decreased 0.14 � 0.03 spikes per secondper degree per second (32 cells, conditions 13a vs. 13b), andhead-on-neck sensitivity decreased by 0.17 � 0.07 (30 cells,conditions 14a vs. 14b). As can be seen in Table 2, in all cases,sensitivity is higher at the lower velocity. Across all of the cellswhere stimuli were tested at both velocity amplitudes, the onlysignificant effect on the sensitivity of the neuron was thestimulus amplitude [ANOVA of stimulus, rotation type, andtarget presence; the only significant effect was the main effectfor stimulus amplitude: F(1,457) � 20.67, P � 7E-6]. Therewere no significant effects for the type of rotation, whole-body,head-on-neck, or trunk-under-head rotation, or whether therotation was driven by the chair or a head or gaze-relatedtarget. Although there appeared to be increasing gains fordecreasing vestibular and somatic stimulus amplitudes, re-sponses to changes in the velocity of visual stimuli remainunknown, as target-only rotations were conducted at a singlestimulus velocity.

Monosensory sensitivity (vestibular, neck, or visual). Thesimplest stimuli used were those that tested just one elementindependently: whole-body rotation in the dark (vestibular),trunk-only rotation with the head fixed in space (neck), andhead-fixed stable visual target oscillation (visual). No cell wasever noted to respond to the static position of the head in thehorizontal plane. Static positioning of the head relative to thetrunk was not performed. Passive vestibular sensitivity wasfound in 49/61 (80%) of the cells tested. The neck-rotation testwas performed by rotating the head on the neck at the samevelocity in the opposite direction of simultaneous chair rota-tion. Because the head motor is mounted on the trunk motorplatform, the head remains motionless in space as long as thehead and trunk motor rotations are precisely matched in fre-quency and amplitude and 180° out of phase. The phase of theresponse is defined relative to chair velocity. Neck-rotationalsensitivity was found in 44/58 (76%) cells tested. We foundthat 51/63 PIVC neurons tested displayed sensitivity to thevisual target sinusoidally oscillated at 15°/s in the horizontal



on Novem

ber 3, 2014D

ownloaded from

Egaze

Hspace

FR

Htrunk

Trspace

Taspace

2 s

Egaze

Hspace

FR

Htrunk

Trspace

Taspace

Condition 9 (head + target – trunk)

Condition 7 (head + target)

Condition 8(head + target + trunk)

Condition 3 (target)

Condition 1(whole body)

Condition 2 (trunk)

Egaze

Hspace

FR

Htrunk

Trspace

Taspace

2 s


Condition 11 (head – target)

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

Condition 6 (head)

Condition 4 (whole body + target)

Condition 5 (trunk + target)

Egaze

Hspace

FR

Htrunk

Trspace

Taspace

Fig. 3. Traces from 1 neuron tested with the 11 conditions used with the head-controlled protocols. The condition numbers correspond to the conditions listedin Table 1. Egaze, eye in space velocity (with saccades removed); Hspace, head velocity in space. FR, spike raster and Kaiser Window filtered representation ofthe firing rate are both shown. Htrunk, head-on-neck rotator velocity. Trspace, trunk (chair rotator) velocity. Taspace, gaze target velocity in space.



on Novem

ber 3, 2014D

ownloaded from

plane at eye level while the animal remained passively re-strained and stationary. Although this represents 81% of thecharacterized cells, search criteria when screening neurons didnot include sensitivity for pursuit of the gaze target. Therefore,cells that have visual sensitivity but no vestibular or necksensitivity, if present in the PIVC, are unlikely to have beenfound. Thus all cells with visual target sensitivity that werecorded had a vestibular or neck response, whereas 82% ofcells with vestibular (40/49) and 77% with neck sensitivity(34/44) also had visual pursuit sensitivity. The average gain forvisual pursuit is higher than vestibular or neck sensitivity, but,because the velocity at which pursuit was conducted was 15°/sand most of the vestibular and neck sensitivity was tested at30°/s, the difference in sensitivity might be related to the peakvelocity. Neck sensitivity was not tested in the dark at thislower velocity, but vestibular stimuli at 15°/s were tested insome cases.

Figures 3 and 4 demonstrate the responses of two neuronsfor which complete head-controlled data (conditions 1–11)were collected. The neuron in Fig. 3 leads head velocity to theright and is in phase with trunk-under-head rotation to the rightand visual target motion to the right. The neuron in Fig. 4responds to head velocity to the left, trunk-under-head rotationto the right, and visual target motion to the right.

As in previous studies, in the current experiments, no PIVCneuron was found that responded to spontaneous eye move-ments (Akbarian et al. 1988; Grusser et al. 1990a, 1990b; Chenet al. 2010; Liu et al. 2011). Ocular rotations were notableduring whole-body rotation in the dark (conditions 1a, 1b), as

this stimulus elicits a compensatory vestibulo-ocular reflex(VOR). The VOR eye movements were suppressed duringwhole-body rotations with a target that moved along with thebody (conditions 4a, 4b). Six cells were tested at 15°/s in bothconditions, and 37 were tested in both at 30°/s. Despite themajority of these cells having visual target sensitivity, therewas no difference in the neural responses with or without atarget during whole-body rotation, suggesting that the ocularmovement does not significantly affect the firing rate of thispopulation of neurons [a gain difference of 0.08 � 0.09spikes/s per °/s at 15°/s, and �0.03 � 0.02 at 30°/s betweenconditions 1 and 4; ANOVA effect of rotation with and withouta target F(2,85) � 0.57, P � 0.45].

What has not been shown previously is the response of theneurons to target movement when that visual stimulus is notlarge enough to elicit self-motion. Figure 5 is an example of aPIVC cell response during pursuit of a target where themonkey did not perform the task all of the time. As can beseen, the response of the neuron is related to the targetmovement, not the eye velocity. Thus we consider this a visualrather than an oculomotor stimulus.

Figure 6A demonstrates the velocity sensitivity and phase ofresponses to monosensory stimuli for cells that were sensitive.These polar plots demonstrate the wide range of gains andphases noted for each of the stimuli in isolation. In many ofthese neurons, data were collected for two or three of theunimodal stimuli. Of the 56 cells tested for both head and necksensitivity, 40 cells responded to both stimuli, 5 were vestib-ular, 4 were neck proprioceptive, 6 were sensitive only to

Table 2. The cell counts and sensitivities for the neurons tested under the experimental conditions outlined in Table 1

Condition N Gain � SE Reference Stimulus Rotation Protocol

1a 7 0.44 � 0.10 Chair velocity Passive whole body �15°/s in dark1b 45 0.25 � 0.04 Chair velocity Passive whole body �30°/s in dark2a 4 0.78 � 0.25 Chair velocity Passive trunk under head �15°/s in dark2b 41 0.24 � 0.04 Chair velocity Passive trunk under head �30°/s in dark3 51 0.40 � 0.05 Target velocity Gaze target only �15°/s4a 22 0.52 � 0.14 Chair and target velocity Passive whole body �15°/s with target4b 51 0.26 � 0.04 Chair and target velocity Passive whole body �30°/s with target5a 5 0.38 � 0.17 Chair velocity Passive trunk under head �15°/s with target5b 48 0.24 � 0.04 Chair velocity Passive trunk under head �30°/s with target6a 4 0.52 � 0.19 Head velocity Passive head on neck �15°/s in dark6b 41 0.21 � 0.03 Head velocity Passive head on neck �30°/s in dark7a 9 0.36 � 0.06 Head and target velocity Passive head on neck �15°/s with target7b 50 0.25 � 0.03 Head and target velocity Passive head on neck �30°/s with target8 21 0.50 � 0.09 Chair velocity Passive head with target �30°/s, synergistic trunk �15°/s9 21 0.48 � 0.10 Chair velocity Passive head with target �15°/s, opposed trunk �15°/s10 11 0.66 � 0.12 Head velocity Passive head on neck �15°/s with synergistic target �30°/s11 13 0.86 � 0.16 Head velocity Passive head on neck �15°/s with opposed target �7.5°/s12a 47 0.45 � 0.06 Chair and target velocity Trunk and gaze target aligned �15°/s12b 65 0.28 � 0.03 Chair and target velocity Trunk and gaze target aligned �30°/s12c 44 0.44 � 0.07 Chair and target velocity Trunk and head target aligned �15°/s12d 59 0.27 � 0.04 Chair and target velocity Trunk and head target aligned �30°/s13a 41 0.44 � 0.07 Chair velocity Trunk �15°/s with gaze target stationary in space13b 53 0.25 � 0.04 Chair velocity Trunk �30°/s with gaze target stationary in space13c 42 0.40 � 0.06 Chair velocity Trunk �15°/s with head target stationary in space13d 56 0.22 � 0.03 Chair velocity Trunk �30°/s with head target stationary in space14a 39 0.50 � 0.07 Head and target velocity Trunk stationary with gaze target moving in space �15°/s14b 42 0.34 � 0.03 Head and target velocity Trunk stationary with gaze target moving in space �30°/s14c 51 0.42 � 0.05 Head and target velocity Trunk stationary with head target moving in space �15°/s14d 58 0.27 � 0.02 Head and target velocity Trunk stationary with head target moving in space �30°/s15a 68 0.51 � 0.04 Chair velocity Synergistic trunk �15°/s with gaze target �30°/s15b 65 0.51 � 0.05 Chair velocity Synergistic trunk �15°/s with head target �30°/s16a 67 0.46 � 0.05 Chair velocity Opposed trunk �15°/s with gaze target �15°/s16b 63 0.48 � 0.05 Chair velocity Opposed trunk �15°/s with head target �15°/s



on Novem

ber 3, 2014D

ownloaded from

Egaze

Hspace

FR

Egaze

Hspace

FR

Egaze

Hspace

FR

Egaze

Hspace

FR

2 s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

2 s

Condition 9 (head + target – trunk)


Condition 8(head + target + trunk)


Condition 1(whole body)

Condition 2 (trunk)


Condition 11 (head – target)

Condition 6 (head)

Condition 4 (whole body + target)

Condition 5 (trunk + target)

Htrunk

Trspace

Taspace

Htrunk

Trspace

Taspace

Htrunk

Trspace

Taspace

Htrunk

Trspace

Taspace

Fig. 4. Traces from another neuron tested with the 11 conditions used with the head-controlled protocols. The condition numbers correspond to the conditionslisted in Table 1. Labels as in Fig. 3.



on Novem

ber 3, 2014D

ownloaded from

visual target velocity, and 1 had weak vestibular and neckresponses that were statistically significant only during activemovement. There was no significant difference between ves-tibular and neck sensitivity [ANOVA of cells with both ves-tibular and neck responses; gain: F(1,79) � 0.17, P � 0.68].Figure 6B demonstrates the gain and phase relationships com-paring vestibular and neck (Fig. 6B, top), vestibular and visual(Fig. 6B, middle), and neck and visual (Fig. 6B, bottom)response for neurons that shared responses to the two stimuli.What this figure reveals is that there is no consistent relation-ship between vestibular, neck, and visual sensitivities at thesingle neuron level in the PIVC neurons studied. Although thestandard index of the gain between vestibular, neck, and visualconditions indicated that the differences in the gains clusteredaround a relatively stable value (squares and triangles), thephase differences between conditions did not follow the patternof the gain differences (bars). In some cases, the responses arein the same direction; in other cases, they are in oppositedirections or they have an in-between phase relationship.Although not easily shown in a two-dimensional figure, mostof these neurons were recorded under all three conditions, andno pattern of alignment of phases between the three sensorymodalities was observed.

Head-on-neck sensitivity in the dark. Head-on-neck rota-tions produce both vestibular and neck stimulation. To testsomatic-vestibular convergence, the response to the stimuluscombination (condition 6) was compared with that expectedfrom the combination of individually determined neck andvestibular sensitivities (determined in conditions 1 and 2).Thirty-three cells had response sensitivity to passive neck,vestibular, and head-on-neck stimuli in the dark. The observedmean vestibular and neck gains were 0.28 � 0.06 and 0.29 �0.06, respectively; the observed head-on-neck gain was 0.22 �0.03. Figure 7A demonstrates one neuron where vestibularsensitivity is rotation to the left. This is the same neuron as inFig. 4, but the figure is showing the averaged Kaiser window-filtered response. This cell also responds, under condition 2, torotation of the trunk to the left with the head held still. Theneuron displayed little sensitivity to head-on-neck rotation andappears to respond selectively to trunk rotation in space inde-pendent of the movement of the head. In contrast, the neuronin Fig. 3 responds to right whole-body motion (vestibularstimuli, condition 1) and rightward rotation of the trunk underhead (condition 2). For the combined head-on-neck rotation incondition 6 as shown in Fig. 7B, this neuron responds as if

dominantly coding the vestibular input. The other pattern seenwith combination of vestibular and neck proprioception signalsis shown in Fig. 7C, an example neuron that demonstrated aresponse dominated by neck input.

When we looked cell by cell and used the criteria of thelowest error between the observed response and the predictedresponse, 9/33 cells were best described as canceling of thehead and neck inputs, effectively best aligning with body inspace; 13/33 cells were the most closely related to necksensitivity, effectively coding the relative velocity of the bodyrelative to the head; and 9/33 neurons were best described ascoding vestibular sensitivity, effectively encoding head veloc-ity in space. Two neurons were not better described by any ofthese individual parameters.

Effect of visual target on vestibular, neck, and head-on-necksensitivity. Also demonstrated in Fig. 3B and Table 1, therewere three conditions with the target available that were equ-ivalent to the three in the dark conditions, except for thepresence of a target aligned with the head. We could compareresponses of neurons to vestibular stimulation with and withouta head-fixed target (condition 1 vs. condition 4), to neckstimulation with and without a head-fixed target (condition 2vs. condition 5), and head and neck stimulation with andwithout a head-fixed target (condition 6 to condition 7). Similarto when we combined head movement and neck twist to findthat these responses do not reliably combine across PIVCneurons to predict head-on-neck, when we add a visual targetto passive whole-body rotation, the responses are not simple orconsistent. In Fig. 3, the addition of the target to whole-bodyrotation (condition 4 compared with condition 1) has no effecton the response. In Fig. 4, the head and visual sensitivity are inopposite directions; when combined, the response is bimodal,one aspect of the response aligning with target velocity inspace, and the other aligning with head velocity. It is notablethat, for this neuron, the visual target movement sensitivityappears to be relative to space, not relative to the animal, as theanimal and target are not moving relative to one another.Figure 8 demonstrates three additional neurons that combinevisual and vestibular information differently. In all three ofthese cases; the vestibular and visual sensitivities are not inphase. In the first row, the visual target dominates the response(Fig. 8A). Again, the target is moving with the animal, so thisis target in space that is evoking response. In the middle row,it is the vestibular signal that is dominant (Fig. 8B). In the lastrow, both signals are represented in the response to rotationwith a target moving with the head (Fig. 8C). In total, 33neurons were tested for whole-body rotation sensitivity (con-dition 1), visual target-alone condition (condition 3), andwhole-body rotation with a target (condition 4). In 11 of theseneurons, the difference in the phase of the response betweenconditions 1 and 3 was more than 90°. Of these 11, theresponse in condition 4 was closer (based on the lowest errorbetween the observed response and the predicted response) tothe condition 3 response (visual dominance) for six neurons,three showed vestibular dominance, and two showed a sum-mation response.

Responses under condition 5 (trunk under head with target) didnot vary much from those under condition 2 (in dark). Lastly,head-on-neck (condition 6) rotation was compared with head-on-neck rotation with target (condition 7). For cells where theresponses to head-on-neck rotation in the dark and target in space

30o/s

30o/s

30o/s

2 s

30o

30o/s

Htrunk

Trspace

Taspace

Epos

Evel

FR 50sp/s

Fig. 5. Response of parieto-insular vestibular cortex (PIVC) neuron to a targetin an episode where behavior is poor. Epos, eye position. Evel, eye velocity(saccades not removed, but cutoff in the figure). Other labels are as in Fig. 3.



on Novem

ber 3, 2014D

ownloaded from

were in the same direction, the combined response was consis-tently in the same direction. Such a response is seen in Fig. 3(conditions 6 vs. 7). In Fig. 4, the dominant visual responseoverrides the weak head or neck response (conditions 6 vs. 7). For28 neurons that responded to visual targets and head-on-neckmovement and were tested in condition 7, eight of those neuronshad a disparity between the phase relationships between theresponses to head-on-neck rotation compared with target velocityof �90°. When looking at these examples carefully, in three cells,the response was closest to the visual response in space, in three,the response was closest to the head-on-neck velocity response,and, in two, there was a response that was not clearly aligned withor dominated by either the head or visual component.

Active or passive. The effects of motor signals were tested inthe PIVC by comparing the responses of neurons duringpassively applied and volitionally produced rotations of theneck. In all volitional head-movement protocols, the targetsneeded to be on to elicit the behavior. Thus there is novolitional protocol equivalent to a protocol in the dark or withtrunk and head both fixed while following the target with gaze(conditions 1, 2, 3, and 6). For the equivalent of whole-bodyrotation with a head-fixed target (condition 4), condition 12only requires the animal to hold his eyes and/or head alignedwith his trunk. Conditions 13–16 all involve voluntary move-ment of the head relative to the trunk (keeping the headstationary in space during trunk rotation in condition 13,

B A

-200

-150

-100

-50

0

50

100

150

200

-1.0

-0.5

0

0.5

1.0

Cell

Phas

e di

ffere

nce

(∆φ

)

Stan

dard

inde

x of

gai

n (IG

)

Condition 1/Condition 3

-200

-150

-100

-50

0

50

100

150

200

-1.0

-0.5

0

0.5

1.0

Cell

Phas

e di

ffere

nce

(∆φ

)

Stan

dard

inde

x of

gai

n (IG

)


-200

-150

-100

-50

0

50

100

150

200

-1.0

-0.5

0

0.5

1.0

Cell

Phas

e di

ffere

nce

(∆φ

)

Stan

dard

inde

x of

gai

n (IG

)


15o/s 30o/s

∆ φ IG

0 0 0 5 10 15 20 25 30 35 40 5 5

0 0 0 5 10 15 20 25 30 35 40 5 5

0 0 0 5 10 15 20 25 30 35 40 5 5

90°

270°

180° 0°

1.0 3.0

0.1

15o/s 30o/s

90°

270°

180° 0°

1.0 3.0

0.1

90°

270°

180° 0°

Condition 2 (trunk)


Condition 1 (whole body)

1.0 3.0

0.1

Fig. 6. A: polar plots representing the gain(radius) and phase [relative to right head,trunk (neck twist), or target velocity] of theresponses for responsive PIVC neurons tounisensory stimuli. B: histograms representthe relationship between head, neck, and vi-sual head-controlled stimuli for neurons hav-ing sensitivity to at least 2 of the stimuli. Thecells are aligned by standard index of gain.Symbols distinguish between conditions with15°/s velocity and 30°/s velocity. Shaded barsrepresent phase differences at 30°/s andhashed bars 15°/s. Triangles represent thestandard index of gain at 15°/s, and squaresrepresent 30°/s. For condition 3, only 15°/svelocity was tested, symbol used is a circle.Standard index of gain for the three graphsare (gain of condition 1 - gain of condition2)/(gain of condition 1 � gain of condition2), (gain of condition 1 - gain of condition 3)/(gain of condition 1 � gain of condition 3),and (gain of condition 2 - gain of condition3)/(gain of condition 2 � gain of condition3). Phase differences are condition 1 phasere:trunk - condition 2 phase re:trunk, con-dition 1 phase re:trunk - condition 3 phasere:target, and condition 2 phase re:trunk -condition 3 phase re:target, respectively.The data from the example neuron in Fig. 3are highlighted in red, and the data from theexample neuron in Fig. 4 are highlighted ingreen.



on Novem

ber 3, 2014D

ownloaded from

head-on-neck rotation in condition 14, and rotating the headwith or against the direction of trunk rotation in conditions 15and 16). For comparison, the responses of neurons during suchvolitional head-movement protocols would be readily com-pared with responses during similar passively produced con-

ditions 5, 7, 8, and 9, in which a similar visual target waspresent. Examples of equivalent passive and volitional condi-tions for a single neuron are shown in Fig. 9. There was nosignificant difference between active and passive responses(paired t-test; comparisons of gain and phase, both peak ve-

IFR

(sp/

s)

0

15

30

Cyc

les

1

6

C neck input dominance

Stim

ulus

(°/s

)

-40

-20

20

40

0

IFR

(sp/

s)

0

30

60

Cyc

les

1

6

B vestibular input dominance

Time (s)

5432154321 54321

A cancellation

Condition 2 (trunk)

Condition 1 (whole body)

Condition 6 (head)

Cyc

les

1

6

IFR

(sp/

s)

0

150

250

50

Trunk Head Head Trunk

Trunk

Head

Fig. 7. Examples of 3 neurons (A, B, and C) where the responses to vestibular, neck, and combined stimulation are compared. IFR, instantaneous firing rate. Top:averaged IFR over 5 cycles filtered with Kaiser Window. Bottom: spike raster, each dot represents 1 spike. The lowest panel in C applies to A and B as well;trunk velocity is black, and head velocity is gray. Left: trunk velocity and head velocity are the same. Middle: trunk is moving with the head held stationary.Right: for condition 6, there is no trunk movement; the head is moving, but the trunk is not.



on Novem

ber 3, 2014D

ownloaded from

locities, P � 0.30 for all comparisons between conditions 4b to12d, 5b to 13d, and 7b to 14d).

Velocity reference frames. The response of PIVC neurons tovisual-target tracking requires further examination to betterunderstand what these responses represent. During the task incondition 3, the animal maintains its gaze on a moving targetwhile its head and trunk are held still. As noted above, the

response is not an ocular rotation signal. Instead, the responsesof the PIVC neurons during visual-target tracking likely rep-resent the target motion. However, the target is moving inspace and relative to both the head and trunk. Our data allowus to investigate whether the target-velocity signal carried bythe PIVC neurons represents target motion in space, targetmotion relative to the head or trunk, or a combination of these.

A visual target dominance

B vestibular input dominance

C summation

0

25

50

IFR

(sp/

s)C

ycle

s

1

6

IFR

(sp/

s)C

ycle

s

1

6

0

25

50

Stim

ulus

(°/s

)IF

R (s

p/s)

0

50

100

- 40- 20

0

20

40

Condition 3(Target in space)

Condition 4(Whole body + Target in space)

Cyc

les

1

6

1 2 3 4 51 2 3 4 5 1 2 3 4 5Time (s)

Whole body Target in space Whole body + Target in space

Condition 1(Whole body)

Fig. 8. Three examples of cells with the visual, vestibular, and combined sensitivity. Panels as in Fig. 7. Black line is visual stimulus, and dashed line is rotationalstimulus.



on Novem

ber 3, 2014D

ownloaded from

If the response to the target is relative to the head or neck, theresponse should diminish when the target and head or bodymove together and reverse phase when the target movementrelative to either the head or trunk flip, regardless of the targetmotion relative to the external environment.

Tasks were developed to more directly test whether visualsensitivity was referenced to the body of the animal as opposedto an external reference frame (target in space). The head waspassively rotated and aligned with the target, but the trunk waseither moving in the same direction as the head and target, athalf the velocity (� 15°/s vs. � 30°/s, condition 8), or the trunkwas rotated in the opposite direction to the head and target (�15°/s), such that the target movement of the trunk was � 30°/s(condition 9). These protocols were compared with conditionswhere the target, head, and trunk all moved together (condition4) and where the head and target moved over a stationary trunk(condition 7). Similar to these four head-controlled conditions,the reference frame was also examined under head-free con-ditions 12d, 14d, 15d, and 16d. The difference between theseconditions (4, 7, 8, and 9, or 12d, 14d, 15d, and 16d) is only therelationship of the trunk to the head and target. If the responsesof these conditions were similar, then the response to the target

is not affected by the relationship of the body to the target andhead. We used the variability of phase relationships betweenthese conditions to inform us as to the likely reference framethat was coded by neuronal responses to target motions, be itrelative to the trunk, head, or environment (target movement inspace). For example, the neuron in Fig. 3 demonstrates thesame response relative to the target and head velocities regard-less of whether the trunk is moving with the head and target(condition 4), in the same direction as the head and target butat half the amplitude (condition 8), is not moving (condition 7),or is moving in the opposite direction (condition 9). Therefore,the example neuron in Fig. 3 does not represent target motionrelative to the trunk but has target-rotation sensitivity (condi-tion 3), indicating that the neural response represents targetmotion in space. Similarly, in Fig. 4, the majority of theresponse in all four conditions is in phase with head and targetvelocity (with target being matched by gaze velocity).

Figure 10 demonstrates how changing the velocity of thetrunk and head relative to the target affects the phase of theneural response relative to the target velocity and thus the ref-erence frame in which target velocity is coded. In Fig. 10A,passive conditions are considered. In Fig. 10A, the neurons are

Condition 14d Condition 12d Condition 13d

Active head movement

Passive head movement

Condition 7b Condition 4b Condition 5b

Egaze

Hspace

FR

Htrunk

Trspace

Taspace

2 s

30o/s

50sp/s

30o/s

30o/s

30o/s

30o/s

2 s

Egaze

Hspace

FR

Htrunk

Trspace

Taspace

50sp/s

30o/s

30o/s

30o/s

30o/s

30o/s

Head + target Whole body + target Trunk + target

Fig. 9. Example neuron recorded during active and passive rotations of the head. Labels as in Fig. 3. Conditions as in Table 1.



on Novem

ber 3, 2014D

ownloaded from

arranged in order of the variability of the phase relationship tothe target across conditions. We averaged the phase of theresponses in the four conditions and then calculated the vari-ation of each condition from the average of all conditions(phase variation � phase under 1 condition, the average phaseacross conditions). Twenty-three neurons, for which at leastthree of the four conditions listed above were run, werearranged based on the phase variation of the neuron responserelative to the target movement. For the first 10 of the 23neurons, under all of the conditions tested, the phase withreference to the target was within 50° of the average for all theconditions. For these neurons, the responses were clearly notrelated to a trunk-centered reference.

The expected pattern for a neuron that was responsive totrunk movement in space under these conditions would beone where the phase of response with reference to target for

conditions 4 and 8 were similar, the response to condition 7was weaker, and the response to condition 9 was out ofphase with conditions 4 and 8. There were five neurons thatfollow this pattern demonstrated in Fig. 10A: cells 14, 19,20, 21, and 23.

Figure 10B was constructed as is Fig. 10A but consideredinstead the 70 neurons where at least three of the four analo-gous head-free conditions were tested (conditions 12, 15, 14,and 16). With the use of the same criteria as was used tocategorize neurons for the head-controlled data to define whichresponses were best described as related to the target, the 51neurons to the left-hand side of Fig. 10B all had phase rela-tionships within 50° of one another and are considered targetrelated. Only five neurons were noted that have the patternexpected for neurons responding to the velocity of the trunk inspace (57, 62, 63, 67, and 69). Only one neuron that was

CACondition 4 Condition 8 Condition 7 Condition 9

Phas

e va

riatio

n (°

)

150

-200

100

50

0

-50

-100

-150 0.0

0.1

1.0

0.0 0.1 1.0 Sensitivity re: target velocity

body moving with target

Sens

itivi

ty re

: tar

get v

eloc

itybo

dy s

tatio

nary

Head restrained Head unrestrained

Cell B

Condition 12 Condition 15 Condition 14 Condition 16

Phas

e va

riatio

n (°

)

150

100

50

0

-50

-100

-150

Cell

Gaze target Head target

0.2 0.4 0.6 0.8 1.0 1.2

40

30

20

10

0

Head velocity/Target velocity

Num

ber o

f cel

ls

ED

0.0

0.1

1.0

0.0 0.1 1.0

Target-related Not Target-related

Sensitivity re: Gaze target(Condition 14b)

Sens

itivi

ty re

: Hea

d ta

rget

(Con

ditio

n 14

d)

0756065505540453035 25 20 15 10

2 4 6 8 10 12 14 16 18 20 22

Fig. 10. A: comparison of phase of neuronalresponse re:target (and head) for 4 of thehead-controlled stimuli. For the y-axis, 0° isthe mean phase of the responses for that par-ticular neuron across the 3 or 4 conditionstested. The neurons are arranged in order ofascending variance in the phase of the re-sponse. Those neurons to the left have veryconsistent response phases across the stimuli;those to the right have very inconsistent re-sponses across the stimuli. B: as in A, exceptresponses under head-free conditions are com-pared. C: for neurons that had consistent(within 50°) phase relationships across theconditions tested, the sensitivity of the neuronsto conditions 4 and 7 (head controlled) andconditions 12 and 14 (head free) are com-pared. D: histograms comparing the propor-tion of pursuit of the target that is done withthe head for the head-free task of target pursuit �30°/s while the trunk is stationary. Condition14b (gaze target) is compared with condition14d (head target). E: sensitivity of neuronsduring pursuit of a gaze target compared withduring pursuit of an eye target. Neurons thatare classified as target related based on theirposition in B are identified separately from theother cells.



on Novem

ber 3, 2014D

ownloaded from

described as a target-related head-controlled neuron was notalso target related with the head free. Six of the eight neuronsthat were related to neither target nor trunk in space with headcontrolled fell into the target-related category with the monkeyhead free.

The distinction between whether the response to the targetmotion for the 10 neurons under head-controlled and the 51under head-free conditions were trunk referenced or spacereferenced can be further examined by looking at the amplitudeof the responses. In the evaluation of the possible trunkreference of neural responses, if the sensitivity to the targetvelocity was referenced to the trunk, one might expect a muchweaker response to conditions 4 or 12 (where the trunk andtarget move together) than to conditions 7 or 14 (where thetarget is moving �30° relative to the trunk and the trunk isstationary). Figure 10C plots the relationship for these 10target-controlled neurons with the head fixed between condi-tions 4 and 7, and 51 target-based neurons were tested with thehead free between conditions 12 and 14. For these neurons,which compose the majority of the cells we tested, the ampli-tude of the responses are indistinguishable between the condi-tions of the trunk being stationary or moving with the target,which is most consistent with coding of the velocity of thetarget in space, not relative to the trunk.

Manipulation of the target relationship to the head was notdone as completely. Because our animals were not trained tomove their head and eyes independently, the most direct datawe have to investigate whether the target velocity-relatedsignal is best coded relative to the head or in space are thehead-controlled data. There are two lines of data that stronglysupport that target velocity is being coded as target in space,rather than relative to the head. The first is that for all of theneurons in Fig. 10, A and B, that appear to code target velocity;all of these responses occur with the target and head aligned. Ifhead is referenced, there should be no target-related response.For the 10 neurons from Fig. 10A that appear to respond totarget velocity in space, eight were tested with the head fixedwhile visually tracking a moving gaze target (condition 3). Inall eight, the phase relationship was in this condition (3) within31° of the average of the other four conditions (4, 7, 8, and 9).Another line of evidence is from eight neurons, all of which arein the group responding to target velocity either with the headcontrolled or head free, where we collected data with the headrotated on the neck either in phase with the target that movedfaster than the head (condition 10), at the same speed as thehead (condition 7), or in the opposite direction to the head(condition 11). As can be seen in Figs. 3 and 4, and is true forthe other six neurons tested, the responses in all three of theseconditions follow the target in space, regardless of the rela-tionship of the target to the head.

Lastly, we compared the gain of responses of neurons inconditions 14b and 14d. In condition 14, the animal pursuedthe target that is sinusoidally oscillating in front of the animal.When following the gaze (eye) target, the head often movedslower than the target. When following the head target, thehead moved with the target at the same speed when the animalwas behaving properly. The head velocity as a percentage ofthe target velocity was compared between the two conditions inthe histogram in Fig. 10D. There was a significant difference inhead velocity between these conditions with gaze vs. headtargets (paired t-test, P � 0.000001). If the neuron response

was primarily target related, then the responses relative to thetarget velocity should have been similar regardless of howmuch of the pursuit was accomplished with the head or withthe eyes moving relative to the head. If the response wasprimarily related to the relationship between head and targetvelocity, then the responses would differ between these con-ditions. Figure 10E compared the gains under these two con-ditions for 62 neurons with responses recorded under bothconditions 14b and 14d. Of these, 43 were a subset of the 51neurons that appeared related to target in space based on theanalysis represented in Fig. 10B. The other 19 neurons wereeither in the body-in-space-related group based on the analysisin Fig. 10B (3 neurons), did not fit into either the body- ortarget-related groups (10 neurons), or did not have enoughprotocols run to be included in Fig. 10B (6 neurons). For thetarget-related and nontarget-related neurons, the response rel-ative to the target velocity was not significantly impacted bywhether a head target or gaze target was used. This findingsupports the conclusion that the response of the PIVC neuronsin this study did not respond differently based on the task,whether following the target with the head or with the eyes.These data are consistent with the suggestion that the majorityof responses, especially those related to the visual target, arespatially referenced, rather than referenced to head or trunkvelocity.

DISCUSSION

The data presented here show that single neurons in theposterior insula display sensitivity to both self-motion andobject motion. These neurons are activated by combinations ofhead and trunk movements in the presence of a moving object.The patterns of signal combination ranged from equivalentsummation to suppression of one signal type by another (Figs.7 and 8). Heterogeneity in the responses suggests that theactual best stimuli for activation of these neurons were notincluded in our protocols. However, the combined activationfound on PIVC neurons, along with the absence of motor orocular information, is consistent with a perceptual function inwhich the posterior insula represents coordinated self-motionrelative to the motion of objects in space.

Coding of volitional vs. imposed movement. One importantnovel finding presented above is the equivalence of volitionallyactive and passively imposed motion in the posterior insula.The posterior insula was known to respond to the movementsof the body, head, neck, trunk, limbs, hands, and feet (Robin-son and Burton 1980; Friedman et al. 1986) before it wasknown to contain vestibular sensitivity (Akbarian et al. 1988).A role in active movement has also been suggested by theataxia resulting from lesions of the posterior insula (Duffau etal. 2006; Liou et al. 2010). In the vestibular nuclei, manyneurons have been reported that lack vestibular responsesduring active head movement (Cullen and Roy 2004). Theproposed active/passive gating of the vestibular signal is effi-cacious in suppressing unintended vestibular reflexes such ascervical or oculomotor reflexes that would work against thegoal of realignment of the head and eyes to fixate on an objectof interest while actively tracking that object with head move-ments. Therefore, because the posterior insula projects directlyto the medial, superior, and y-group vestibular nuclei (Akbar-ian et al. 1994), it was not unreasonable to speculate that the



on Novem

ber 3, 2014D

ownloaded from

posterior insula could be the source of motor signals thatcontextually modulate vestibular responses during active vs.passive head rotations (Guldin and Grusser 1998).

Recent findings from brain imaging argue against suchspeculation and show that the posterior insula does not requireactive or even actual movement to be responsive but can beactivated by the perception of movement (Kahane et al. 2003;Nowak et al. 2005; Karnath et al. 2005) and by imaginedmovements of the body (Jahn et al. 2004; Fasold et al. 2008;Deutschländer et al. 2009). Furthermore, self-motion percep-tion can be induced without involving the posterior insula(physiology: Chen et al. 2010; imaging: Becker-Bense et al.2012). Moreover, corticospinal projections from the posteriorinsula are to sensory, not motor, portions of the spinal cord(Galea and Darian-Smith 1994), and the posterior insula is notinvolved in motor planning or timing (Hoffstaedter et al. 2013).

As fundamental a principle as efference copy mechanismsare to monitoring reafference in the vestibular nuclei, there isalso a perceptual role for efference in the vestibular system tokeep accurate track of the relative velocities of the head, body,and objects in the environment, regardless of whether theself-motion is externally applied or internally generated. Theequivalence of active and passive responding in the posteriorinsula supports a role in sensory perception, consistent withmonitoring subjective states including movement. The de-scending corticofugal projections may therefore be related tosensory feedback for the sense of self-motion similar to thatseen in other sensory modalities (Alitto and Usrey 2003). Theloss of feedback and coordinated monitoring of self-motionand object motion also explains ataxia following lesions and itstransience as the brain adjusts to rely more on direct sensoryfeedback rather than ongoing monitoring of the alignmentbetween oneself and objects (Medendorp et al. 2008).

Visual signals in the posterior insula. The other importantand novel finding of this study was that discrete, moving,visual targets elicited a response in a majority of posteriorinsular neurons that also displayed some form of self-motionsensitivity. Given the collection of visual stimuli that haveelicited activation of the posterior insula in previous studies,combining the current findings of visual sensitivity in theposterior insula is complicated. Information about movingvisual targets may come from the interaction of the posteriorinsula with the smooth pursuit region of the frontal eye fields(Stanton et al. 2005) or visually related regions of the posteriormesial cortex (Cavada and Goldman-Rakic 1989; Morecraftand Van Hoesen 1998; Leichnetz 2001; Morecraft et al. 2004).However, visual information about self-motion does not cometo the insula directly from visual motion sensitive temporalcortex, which may explain differences in posterior insularneuron responses to visual and somatic motion information, asthe visual cortices providing object motion information are alsothe likely source of visual self-motion information, and theinteraction of visual information sources may occur upstreamof the posterior insula. In this study, we found the absence ofneural responses to eye movements in the dark, which corrob-orates previous studies that found no ocular sensitivity in thePIVC. Our study is also consistent with previous studiesindicating that fixating on a stationary target does not activateposterior insular neurons, as evidenced by the lack of effect ofa stationary target on the response to trunk-under-head rotation(Akbarian et al. 1988; Grusser et al. 1990a; Chen et al. 2010).

Therefore, posterior insular signals are unlikely to representgaze-control signals. Visually, the movement of large, full-field displays that induce a perception of self-motion produceposterior insular activation (physiology: Akbarian et al. 1988;Grusser et al. 1990b; imaging: Kleinschmidt et al. 2002) unlessa stationary fixation target is presented along with the large-field visual-motion stimulus (Chen et al. 2010). Interestingly,smaller visual stimuli that are generally larger than 10° but arenot large enough to produce the subjective experience ofself-motion have not been shown to activate the posteriorinsula in imaging studies (Brandt et al. 2002; Deutschländer etal. 2002; Bense et al. 2005; Dieterich et al. 2007; Cardin andSmith 2011). However, imaging studies have shown that singlevisual objects (Indovina et al. 2005; Nagel et al. 2006; Lindneret al. 2006) and coherent visual-field stimuli smaller than 10°(Dupont et al. 1994; Sunaert et al. 1999; Antal et al. 2008)produce activation in the posterior insula. If a visual stimuluscan be defined as an object, or it induces the perception ofself-motion, then that stimulus has the potential to activate theposterior insula. The current findings of neural responses tovisual-target motion help connect the responses seen duringthe imaging studies above to the physiological responses of theposterior insula. Significantly, the mechanism behind how theposterior insula deals with the combination of visual stimulican be seen in the unexpected finding of Chen et al. (2010).They presented a stimulus that combined a visual object alongwith a moving visual field that induced the perception ofself-motion. However, the expected robust response to self-motion in the posterior insula was not found. Presuming, asabove, that the posterior insula combines various forms ofmovement information about coordinated head and trunk mo-tion relative to objects in the environment, then the presence ofan attended, stationary object negates the utility of the visualself-motion signal to monitor movements of the object andoneself. This corroborates the human studies that show that theperception of self-motion could be induced without activatingthe posterior insula (Becker-Bense et al. 2012). Thus the visualsignals important to the function of the posterior insula may bepredicted to be the ones that are most significant to themonitoring of ongoing behavior and not necessarily the detec-tion of self-related motion.

Vestibular signals in the posterior insula. Vestibular infor-mation in the posterior insula originates from the vestibular-related thalamus (see Shinder and Taube 2010 for review). Thesubcortical origins of vestibular input can be used to explainthe primacy of vestibular input in the posterior insula and alsoprovide insight into its vestibular properties. The vestibularsensitivity, in particular the increasing gain with decreasingstimulus amplitude, is reflected in the ascending input to theinsula (Marlinski and McCrea 2008a; Newlands et al. 2009).

We set out to understand more about the vestibular signal inthe posterior insula by studying those cells at the neuronal levelmost responsive to yaw rotation. In doing so, we only inves-tigated a subset of the cells that we isolated (about 1/4 hadsensitivity to yaw head and/or neck movement). Previousstudies found that 61% of PIVC neurons were vestibular(62/107 cells, Akbarian et al. 1988), and, of the vestibular-sensitive neurons, 59% were sensitive to yaw rotations in thehorizontal plane as were used here (96/162, Grusser et al.1990a, 1990b). Thus 36% of PIVC neurons would be predictedto be responsive in the current experiments based on those



on Novem

ber 3, 2014D

ownloaded from

studies. However, the criteria required to designate a neuron asresponsive was made more stringent in this study comparedwith the studies of Grusser et al. (1990a, 1990b). As opposedto using a 0.04 spikes/sec per °/sec lower limit to defineneuronal sensitivity to a stimulus, the current study uses 0.1spikes/s per °/s and statistically significant Rayleigh statistics.By comparison, Chen et al. (2010) found that, among vestib-ular-sensitive cells, which were the majority in the PIVC, yawrotation sensitivity was relatively uncommon, seen in only20% of PIVC cells. We believe that the current sampling ofresponsive neurons (23%) is in line with previous measures,adjusted for current methodologies.

In this study, vestibular signals differed in many wayscompared with the vestibular nuclei and other subcorticalareas. Yaw rotation signals in the vestibular nuclei lead veloc-ity (Dickman and Angelaki 2004; Newlands et al. 2009). Ourdata show a distribution of response phases without clearpreference for velocity or acceleration (Fig. 6A, top). Vestib-ular signals in the vestibular nuclei combine with oculomotorsignals rather than visual signals (Baker and Berthoz 1974;King et al. 1976). Many vestibular responses in the vestibularnuclei, cerebellum, and ventro-posterior thalamus combinedwith neck proprioceptive signals in precise ways to code eitherhead velocity in space or body velocity in space (Gdowski andMcCrea 2000; Marlinski and McCrea 2008b, 2009; Brooks andCullen 2009; Luan et al. 2013). Although some of the neuronsreported here have signals consistent with coding body velocityin space, they are a small portion of the responses we found.Based on these differences, we speculate that the vestibularsignals in the PIVC serve a different purpose than in otherareas. Rather than being involved in response to movement,these signals may primarily be used as an index of movementin space to transform the perception for body and objectmovement from being self-referenced to being referenced tothe environment. Evidence consistent with this hypothesiscomes from human imaging studies that noted a role for theretroinsular cortex in monitoring the direction of gravity rela-tive to object motion to anticipate gravitational acceleration ofmoving objects (Indovina et al. 2005). To monitor objects inspace, signals that inform object position relative to the organ-ism (visual signals) must combine with signals informing theorganisms of the relationship of the environment (vestibularsignals). Additionally, if the role of the region is to monitor theobjects in the environment and/or relative to the head andbody, then whether the head and body movements are self-directed or imposed is not relevant. Thus the characteristics ofPIVC neurons are consistent with a role of the PIVC inmonitoring objects in space and relative to the animal.

Multisensory integration. We found that most of the “ves-tibular” neurons in the posterior insula also responded to neckand visual stimuli. Across a population of cells, the visual,vestibular, and neck proprioceptive signals did not combinewith one another in consistent ways. Instead, we found exam-ples where the responses to the various sensory inputs addedtogether and cancelled one another out and other exampleswhere one of the signals dominated the others. Our data reveala complexity of signal processing in the posterior insula. Theheterogeneity of the neuron responses contain the elements ofa brain area that is involved in keeping track of self-movementand the relative movement of objects of interest.

The types of signals represented in the posterior insularelated to self-motion and object motion are also found in othercortical regions (Shinder and Taube 2010). However, signifi-cant differences exist. The visual responses seen in the poste-rior insula are not characteristic of self-motion detection (Chenet al. 2010), as is found in the medial superior temporal cortex(Chen et al. 2013a), nor does the posterior insular responseencode motor- and ocular-related information as is found in theventral intraparietal cortex (Morris et al. 2012; Chen et al.2013a, 2013b). Interestingly, the vestibular self-motion in theposterior insula is represented as a combination of motion inspace (world referenced) and motion relative to the body (bodyreferenced; Chen et al. 2013a). This corroborates current find-ings and suggests that the posterior insula bridges world- andbody-related movement information. In contrast, the MSTneurons show reference frames that are head or eye centeredbut not world referenced (Chen et al. 2013a).

The proposed role of the posterior insula in monitoringcoordinated self-motion and object motion has implicationsthat reach beyond that of an interaction of vestibular sensationwith body movement (Lopez et al. 2012b; Ferre et al. 2012).Imaging studies have suggested that the posterior insula isrelated to function of monitoring subjective states, includingpain perception (Mazzola et al. 2012; Longo et al. 2012;Aziz-Zadeh et al. 2012). The monitoring of movement as asubjective state indicates that movement is not referenced to asingular metric of motion (Chen et al. 2010; Becker-Bense etal. 2012) but enables self-motion and object motion to beafforded multiple qualities. The posterior insula appears able tobridge information it receives from vestibular, visual, andproprioceptive systems and create joint representations of bodyand body-effector motion with object motion. The data pre-sented here demonstrate properties of PIVC neurons heretounknown. Understanding mechanisms of how this sensoryconvergence in the posterior insula informs perception ofself-motion and object motion awaits additional studies.

ACKNOWLEDGMENTS

The authors acknowledge Dr. Min Wei’s creation of the figures in thismanuscript, for which permission has been received.

Current affiliations for the authors are as follows: M. Shinder: Dept. ofNeuroscience, Baylor College of Medicine, One Baylor Plaza, BCM 295,Houston, TX 77030. S. Newlands, Dept. of Otolaryngology, University ofRochester Medical Center, 601 Elmwood Ave., Rochester, NY 14642.

GRANTS

This work was supported by the National Institute on Deafness and OtherCommunication Disorders (R01DC0006429 and F31DC 007583).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

Author contributions: M.E.S. conception and design of research; M.E.S.performed experiments; M.E.S. and S.D.N. analyzed data; M.E.S. and S.D.N.interpreted results of experiments; M.E.S. and S.D.N. prepared figures; M.E.S.and S.D.N. drafted manuscript; M.E.S. and S.D.N. edited and revised manu-script; M.E.S. and S.D.N. approved final version of manuscript.

REFERENCES

Abreu AM, Macaluso E, Azevedo RT, Cesari P, Urgesi C, Aglioti SM.Action anticipation beyond the action observation network: a functional



on Novem

ber 3, 2014D

ownloaded from

magnetic resonance imaging study in expert basketball players. Eur JNeurosci 35: 1646–1654, 2012.

Akbarian S, Berndl K, Grusser OJ, Guldin W, Pause M, Schreiter U.Responses of single neurons in the parietoinsular vestibular cortex ofprimates. Ann NY Acad Sci 545: 187–202, 1988.

Akbarian S, Grusser OJ, Guldin WO. Corticofugal connections between thecerebral cortex and brainstem vestibular nuclei in the Macaque monkey. JComp Neurol 339: 421–437, 1994.

Alitto HJ, Usrey WM. Corticothalamic feedback and sensory processing.Curr Opin Neurobiol 13: 440–445, 2003.

Antal A, Baudewig J, Paulus W, Dechent P. The posterior cingulate cortexand planum temporale/parietal operculum are activated by coherent visualmotion. Vis Neurosci 25: 17–26, 2008.

Arzy S, Thut G, Mohr C, Michel CM, Blanke O. Neural basis of embodi-ment: distinct contributions of temporoparietal junction and extrastriatebody area. J Neurosci 26: 8074–8081, 2006.

Asahi T, Uwano T, Eifuku S, Tamura R, Endo S, Ono T, Nishijo H.Neuronal responses to a delayed-response delayed-reward go/nogo task inthe monkey posterior insular cortex. Neuroscience 143: 627–639, 2006.

Aziz-Zadeh L, Sheng T, Liew SL, Damasio H. Understanding otherness: theneural bases of action comprehension and pain empathy in a congenitalamputee. Cereb Cortex 22: 811–819, 2012.

Baier B, Conrad J, Zu Eulenburg P, Best C, Müller-Forell W, Birklein F,Dieterich M. Insular strokes cause no vestibular deficits. Stroke 44: 2604–2606, 2013.

Baker R, Berthoz A. Organization of vestibular nystagmus in oblique ocul-omotor system. J Neurophysiol 37: 195–217, 1974.

Becker-Bense S, Buchholz HG, zu Eulenburg P, Best C, Bartenstein P,Schreckenberger M, Dieterich M. Ventral and dorsal streams processingvisual motion perception (FDG-PET study). BMC Neurosci 13: 81, 2012.

Bense S, Stephan T, Bartenstein P, Schwaiger M, Brandt T, Dieterich M.Fixation suppression of optokinetic nystagmus modulates cortical visual-vestibular interaction. Neuroreport 16: 887–890, 2005.

Bottini G, Karnath HO, Vallar G, Sterzi R, Frith CD, Frackowiak RS,Paulesu E. Cerebral representations for egocentric space: Functional-ana-tomical evidence from caloric vestibular stimulation and neck vibration.Brain 124: 1182–1196, 2001.

Brandt T, Glasauer S, Stephan T, Bense S, Yousry TA, Deutschlander A,Dieterich M. Visual-vestibular and visuovisual cortical interaction. Ann NYAcad Sci 956: 230–241, 2002.

Brooks JX, Cullen KE. Multimodal integration in rostral fastigial nucleusprovides an estimate of body movement. J Neurosci 29: 10499–10511,2009.

Cardin V, Smith AT. Sensitivity of human visual cortical area V6 tostereoscopic depth gradients associated with self-motion. J Neurophysiol106: 1240–1249, 2011.

Cavada C, Goldman-Rakic PS. Posterior parietal cortex in rhesus monkey. I.Parcellation of areas based on distinctive limbic and sensory corticocorticalconnections. J Comp Neurol 287: 393–421, 1989.

Cereda C, Ghika J, Maeder P, Bogousslavsky J. Strokes restricted to theinsular cortex. Neurology 59: 1950–1955, 2002.

Chen A, DeAngelis GC, Angelaki DE. Macaque parieto-insular vestibularcortex: responses to self-motion and optic flow. J Neurosci 30: 3022–3042,2010.

Chen A, DeAngelis GC, Angelaki DE. A comparison of vestibular spatio-temporal tuning in macaque parietoinsular vestibular cortex, ventral intra-parietal area, and medial superior temporal area. J Neurosci 31: 3082–3094,2011.

Chen X, Deangelis GC, Angelaki DE. Diverse spatial reference frames ofvestibular signals in parietal cortex. Neuron 80: 1310–1321, 2013a.

Chen X, Deangelis GC, Angelaki DE. Eye-centered representation of opticflow tuning in the ventral intraparietal area. J Neurosci 33: 18574–18582,2013b.

Cherif S, Cullen KE, Galiana HL. An improved method for the estimation offiring rate dynamics using an optimal digital filter. J Neurosci Methods 173:165–181, 2008.

Cullen KE, Roy JE. Signal processing in the vestibular system during activeversus passive head movements. J Neurophysiol 91: 1919–1933, 2004.

Deutschländer A, Bense S, Stephan T, Schwaiger M, Brandt T, DieterichM. Sensory system interactions during simultaneous vestibular and visualstimulation in PET. Hum Brain Mapp 16: 92–103, 2002.

Deutschländer A, Stephan T, Hüfner K, Wagner J, Wiesmann M, StruppM, Brandt T, Jahn K. Imagined locomotion in the blind: an fMRI study.Neuroimage 45: 122–128, 2009.

Dickman JD, Angelaki DE. Dynamics of vestibular neurons during rotationalmotion in alert rhesus monkeys. Exp Brain Res 155: 91–101, 2004.

Dieterich M, Bauermann T, Best C, Stoeter P, Schlindwein P. Evidence forcortical visual substitution of chronic bilateral vestibular failure (an fMRIstudy). Brain 130: 2108–2116, 2007.

Downar J, Crawley AP, Mikulis DJ, Davis KD. A multimodal corticalnetwork for the detection of changes in the sensory environment. NatNeurosci 3: 277–283, 2000.

Duffau H, Taillandier L, Gatignol P, Capelle L. The insular lobe and brainplasticity: lessons from tumor surgery. Clin Neurol Neurosurg 108: 543–548, 2006.

Dupont P, Orban GA, De Bruyn B, Verbruggen A, Mortelmans L. Manyareas in the human brain respond to visual motion. J Neurophysiol 72:1420–1424, 1994.

Eickhoff SB, Weiss PH, Amunts K, Fink GR, Zilles K. Identifying humanparieto-insular vestibular cortex using fMRI and cytoarchitectonic mapping.Hum Brain Mapp 27: 611–621, 2006.

Fasold O, Heinau J, Trenner MU, Villringer A, Wenzel R. Proprioceptivehead posture-related processing in human polysensory cortical areas. Neu-roimage 40: 1232–1242, 2008.

Ferre ER, Bottini G, Haggard P. Vestibular inputs modulate somatosensorycortical processing. Brain Struct Funct 217: 859–864, 2012.

Friedman DP, Jones EG, Burton H. Representation pattern in the secondsomatic sensory area of the monkey cerebral cortex. J Comp Neurol 192:21–41, 1980.

Friedman DP, Murray EA, O’Neill JB, Mishkin M. Cortical connections ofthe somatosensory fields of the lateral sulcus of Macaques: evidence for acorticolimbic pathway for touch. J Comp Neurol 252: 323–347, 1986.

Galea MP, Darian-Smith I. Multiple corticospinal neuron populations in theMacaque monkey are specified by their unique cortical origins, spinalterminations, and connections. Cereb Cortex 4: 166–194, 1994.

Gdowski GT, McCrea RA. Neck proprioceptive inputs to primate vestibularnucleus neurons. Exp Brain Res 135: 511–526, 2000.

Grusser OJ, Pause M, Schreiter U. Localization and responses of neurons inthe parieto-insular vestibular cortex of awake monkeys (Macaca fascicu-laris). J Physiol 430: 537–557, 1990a.

Grusser OJ, Pause M, Schreiter U. Vestibular neurones in the parieto-insularcortex of monkeys (Macaca fascicularis): visual and neck receptor re-sponses. J Physiol 430: 559–583, 1990b.

Guldin WO, Akbarian S, Grusser OJ. Cortico-cortical connections andcytoarchitectonics of the primate vestibular cortex: a study in squirrelmonkeys (Saimiri sciureus). J Comp Neurol 326: 375–401, 1992.

Guldin WO, Grusser OJ. Is there a vestibular cortex? Trends Neurosci 21:254–259, 1998.

Guldin WO, Mirring S, Grusser OJ. (1993) Connections from the neocortexto the vestibular brain stem nuclei in the common marmoset. Neuroreport 5:113–116, 1993.

Hoffstaedter F, Grefkes C, Zilles K, Eickhoff SB. The “What” and “When”of self-initiated movements. Cereb Cortex 23: 520–530, 2013.

Indovina I, Maffei V, Bosco G, Zago M, Macaluso E, Lacquaniti F.Representation of visual gravitational motion in the human vestibular cortex.Science 308: 416–419, 2005.

Jahn K, Deutschlander A, Stephan T, Strupp M, Wiesmann M, Brandt T.Brain activation patterns during imagined stance and locomotion in func-tional magnetic resonance imaging. Neuroimage 22: 1722–1731, 2004.

Kahane P, Hoffmann D, Minotti L, Berthoz A. Reappraisal of the humanvestibular cortex by cortical electrical stimulation study. Ann Neurol 54:615–624, 2003.

Karnath HO, Baier B. Right insula for our sense of limb ownership andself-awareness of actions. Brain Struct Funct 214: 411–417, 2010.

Karnath HO, Baier B, Nagele T. Awareness of the functioning of one’s ownlimbs mediated by the insular cortex? J Neurosci 25: 7124–7138, 2005.

King WM, Lisberger SG, Fuchs AF. Responses of fibers in medial longitu-dinal fasciculus (MLF) of alert monkeys during horizontal and verticalconjugate eye movements evoked by vestibular or visual stimuli. J Neuro-physiol 39: 1135–1149, 1976.

Kleinschmidt A, Thilo KV, Büchel C, Gresty MA, Bronstein AM, Frack-owiak RS. Neural correlates of visual-motion perception as object- orself-motion. Neuroimage 16: 873–882, 2002.

Leichnetz GR. Connections of the medial posterior parietal cortex (area 7m)in the monkey. Anat Rec 263: 215–236, 2001.

Lindner A, Haarmeier T, Erb M, Grodd W, Thier P. Cerebrocerebellarcircuits for the perceptual cancellation of eye-movement-induced retinalimage motion. J Cogn Neurosci 18: 1899–1912, 2006.



on Novem

ber 3, 2014D

ownloaded from

Liou LM, Guo YC, Lai CL, Tsai CL, Khor GT. Isolated ataxia after pure leftinsular cortex infarction. Neurol Sci 31: 89–91, 2010.

Liu S, Dickman JD, Angelaki DE. Response dynamics and tilt versustranslation discrimination in parietoinsular vestibular cortex. Cereb Cortex21: 563–573, 2011.

Longo MR, Iannetti GD, Mancini F, Driver J, Haggard P. Linking pain andthe body: Neural correlates of visually induced analgesia. J Neurosci 32:2601–2607, 2012.

Lopez C, Blanke O, Mast FW. The human vestibular cortex revealed bycoordinate-based activation likelihood estimation meta-analysis. Neurosci-ence 212: 159–179, 2012a.

Lopez C, Schreyer HM, Preuss N, Mast FM. Vestibular stimulation modifiesthe body schema. Neuropsychologia 50: 1830–1837, 2012b.

Lotsch J, Walter C, Felden L, Noth U, Deichmann R, Oertel BG. Thehuman operculo-insular cortex is pain-preferentially but not pain-exclu-sively activated by trigeminal and olfactory stimuli. PLoS One 7: e34798,2012.

Luan H, Gdowski MJ, Newlands SD, Gdowski GT. Convergence of ves-tibular and neck proprioceptive sensory signals in the cerebellar interpositus.J Neurosci 33: 1198–1210, 2013.

Luks TL, Simpson GV. Preparatory deployment of attention to motionactivates higher-order motion-processing brain regions. Neuroimage 22:1515–1522, 2004.

Macey PM, Wu P, Kumar R, Ogren JA, Richardson HL, Woo MA,Harper RM. Differential responses of the insular cortex gyri to autonomicchallenges. Auton Neurosci 168: 72–81, 2012.

Mardia KV. Statistics of Directional Data. New York: New York Academic,1972, pp. 133–135.

Marlinski V, McCrea RA. Activity of ventroposterior thalamus neuronsduring rotation and translation in the horizontal plane in the alert squirrelmonkey. J Neurophysiol 99: 2533–2545, 2008a.

Marlinski V, McCrea RA. Coding of self-motion signals in ventro-posteriorthalamus neurons in the alert squirrel monkey. Exp Brain Res 189: 463–472,2008b.

Marlinski V, McCrea RA. Self-motion signals in vestibular nuclei neuronsprojecting to the thalamus in the alert squirrel monkey. J Neurophysiol 101:1730–1741, 2009.

Mazzola L, Faillenot I, Barral FG, Mauguière F, Peyron R. Spatialsegregation of somato-sensory and pain activations in the human operculo-insular cortex. Neuroimage 60: 409–418, 2012.

Medendorp WP, Beurze SM, Van Pelt S, Van Der Werf J. Behavioral andcortical mechanisms for spatial coding and action planning. Cortex 44:587–597, 2008.

Morecraft RJ, Cipolloni PB, Stilwell-Morecraft KS, Gedney MT, PandyaDN. Cytoarchitecture and cortical connections of the posterior cingulatedand adjacent somatosensory fields in the rhesus monkey. J Comp Neurol469: 37–69, 2004.

Morecraft RJ, Van Hoesen GW. Convergence of limbic input to the cingu-lated motor cortex in the rhesus monkey. Brain Res Bull 45: 209–232, 1998.

Morris AP, Kubischik M, Hoffmann KP, Krekelberg B, Bremmer F.Dynamics of eye-position signals in the dorsal visual system. Curr Biol 22:173–179, 2012.

Nagel M, Sprenger A, Zapf S, Erdmann C, Kompf D, Heide W, BinkofskiF, Lencer R. Parametric modulation of cortical activation during smoothpursuit with and without target blanking. An fMRI study. Neuroimage 29:1319–1325, 2006.

Newlands SD, Lin N, Wei M. Response linearity of alert monkey non-eyemovement vestibular nucleus neurons during sinusoidal yaw rotation. JNeurophysiol 102: 1388–1397, 2009.

Nowak M, Holm S, Biering-Sorensen F, Secher NH, Friberg L. “Centralcommand” and insular activation during attempted foot lifting in paraplegichumans. Hum Brain Mapp 25: 259–265, 2005.

Penfield W, Boldrey E. Somatic motor and sensory representation in thecerebral cortex of man as studied by electrical stimulation. Brain 60:389–443, 1937.

Petit L, Beauchamp MS. Neural basis of visually guided head movementsstudied with fMRI. J Neurophysiol 89: 2516–2527, 2003.

Robinson CJ, Burton H. Organization of somatosensory receptive fields incortical areas 7b, retroinsula, postauditory and granular insula of M. fas-cicularis. J Comp Neurol 192: 69–92, 1980.

Shinder ME, Newlands SD. Vestibular responses of neurons in retroinsularand ventral intraparietal cortices of the rhesus monkey during active andpassive head movements. Soc Neurosci Abstr 392: 8, 2005.

Shinder ME, Taube JS. Differentiating ascending vestibular pathways to thecortex involved in spatial cognition. J Vestib Res 20: 3–23, 2010.

Simmons WK, Avery JA, Barcalow JC, Bodurka J, Drevets WC, Bell-gowan P. Keeping the body in mind: Insula functional organization andfunctional connectivity integrate interoceptive, exteroceptive, and emotionalawareness. Hum Brain Mapp 34: 2944–2958, 2013.

Stanton GB, Friedman HR, Dias EC, Bruce CJ. Cortical afferents to thesmooth-pursuit region of the macaque monkey’s frontal eye field. Exp BrainRes 165: 179–192, 2005.

Sunaert S, Van Hecke P, Marchal G, Orban GA. Motion-responsive regionsof the human brain. Exp Brain Res 127: 355–370, 1999.

zu Eulenburg P, Baumgärtner U, Treede RD, Dieterich M. Interoceptiveand multimodal functions of the operculo-insular cortex: Tactile, nocicep-tive and vestibular representations. Neuroimage 83C: 75–86, 2013.



on Novem

ber 3, 2014D

ownloaded from

Documents

Sensory convergence in the parieto-insular vestibular cortex