| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1Department of Anatomy, 2Department of Physiology, and 3Neuroscience Training Program, University of Wisconsin-Madison, Madison, Wisconsin 53706
Submitted 3 May 2004; accepted in final form 4 June 2004
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
35% of the target amplitude, was taken into account, the magnitude of the horizontal shifts in the SC auditory receptive fields matched the observed behavior. The modulation by eye position was not due to concomitant movements of the external ears, as confirmed by recordings carried out after immobilizing the pinnae of one cat. However, the pattern of modulation after pinnae immobilization was inconsistent with the observations in the intact cat, suggesting that, in the intact animal, information about the position of the pinnae may be taken into account. |
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; King and Palmer 1983; Meredith and Stein 1985; Middlebrooks and Knudsen 1984; Stein and Clamann 1981; Stein and Meredith 1993), meaning that acoustic and visual stimuli at the same spatial locations stimulate the same part of the SC. The alignment of auditory and visual topography is thought to provide a substrate for sensorimotor integration (Harris et al. 1980; Stein and Meredith 1993). However, as first pointed out by Pöppel (1973), since the coordinate systems for vision and audition are different, with the visual map oculocentric and the auditory map head-centered, this alignment holds only when the two coordinate systems are aligned, i.e., when the eyes and head point in the same direction.
Thus in behaving animals, whose eyes are free to move in the head, one would expect the SC maps to become misaligned whenever the animal is not looking straight ahead. If this alignment is critical for sound localization or for bimodal integration, the consequence of misalignment between these two maps might be mislocalization of the source of a sound whenever the eyes are away from the primary position. Furthermore, the magnitude of the expected error in sound localization should be proportional to the deviation of the eyes from the primary position (Harris et al. 1980). This hypothesis can be tested behaviorally.
Alternatively, the collicular representation of auditory space could shift to compensate for changes in eye position to maintain the alignment of the receptive fields (RFs) across the different modalities. Such a shift would require a modulating signal proportional to the position of the eyes within the orbit to yield a representation of the eye motor error, which is the difference between eye position and the target. In this scheme, also known as the "motor error hypothesis" (Sparks 1986), the shifted auditory RFs would not encode the location of auditory sources relative to the head, but rather their location relative to the position of the eyes within the orbit. This hypothesis can be tested physiologically.
Jay and Sparks (1987) found evidence supporting the motor error hypothesis in the SC of the monkey: they found that the auditory response was modulated by eye position, although it only partly compensated for the motor error. More recently, Groh et al. (2001) and Zwiers et al. (2004) both found evidence of modulation by eye position in auditory responses by a variable number of cells in the inferior colliculus (IC) even though their monkeys were not trained to orient to the acoustic stimuli. However, in the cat, conflicting evidence has been presented. Harris et al. (1980) claim that cats do not move their eyes enough to require compensation, while Hartline et al. (1995) and Peck et al. (1995) provide evidence for compensation. Furthermore, shifts of auditory receptive fields in the SC of cats could result from changes in the acoustic input to the eardrums brought about by the highly mobile pinnae (Middlebrooks and Knudsen 1987). However, the interpretation of the available data in all three studies (Harris et al. 1980; Hartline et al. 1995; Peck et al. 1995) is hindered by limited sample size, untrained and inconsistent subjects, and inadequate monitoring of eye or ear position. We have designed this study to overcome these problems using cats who consistently localized sound by looking at the acoustic sources and well-controlled experimental conditions.
Our results show that most cells driven by acoustic stimuli in the deep and intermediate layers of the SC of the cat are modulated by eye position in a manner consonant with the motor error hypothesis. Although the shift in auditory RFs did not compensate fully for the motor error, it did match the behavioral motor error, which was calculated on the basis of the propensity for the saccades to auditory targets to undershoot the target. Finally, by recording from cells in the SC in a cat whose pinnae had been immobilized, we showed for the first time that the modulation cannot be due to the changes in acoustics brought about by movements of the pinnae of the cat while fixating different positions. However, the modulation seen in the pinna immobilized cat was not consistent with the modulation seen in the intact cat, suggesting that eye position, pinna position, and a motor efferent copy command may all play a role in the acoustic modulation in the SC. A preliminary account of portions of this work has been presented (Populin and Yin 1998a).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
A detailed description of the materials, methods, and animal training procedures has been presented previously (Populin and Yin 1998b). Briefly, 10 cats were trained to localize sound by looking at the perceived location of the sources. Single neuronal recordings made in the SC of six of them are included in this report. Under sterile conditions, a recording cylinder was implanted on the midline of the skull to access the SC on both sides with vertically oriented electrode penetrations. The animals had already undergone a surgical procedure to implant eye (Judge et al. 1980) and pinna (Populin and Yin 1998c) coils and a post to restrain the head and had been trained on the behavioral tasks.
Anesthesia was first induced with an intramuscular injection of ketamine (20 mg/kg) and acepromazine (0.2 mg/kg), followed by intravenous sodium pentobarbital or gas (halothane or isoflurane). All surgical and experimental procedures employed were approved by the University of Wisconsin Animal Care Committee and were in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
The experiments were carried out in a dimly illuminated (2.2 x 2.5 x 2.5 m) single-walled sound-attenuated chamber with the inner walls and major pieces of equipment in it covered with 10.2-cm Sonex foam (Ilbruck, Minneapolis, MN) to attenuate acoustic reflections. All experiments were carried out under head-restrained conditions.
Eye and pinna movement recordings and stimuli presentation
Eye and pinna movements were recorded with the scleral search coil technique (Robinson 1963). The analog output of the CNC Engineering (Seattle, WA) system used to measure eye and pinna movements was digitally sampled at 500 Hz with an A/D converter and stored on disk for off-line analysis.
Acoustic stimuli were generated with a Digital Stimulus System (Rhode 1976) and presented via Radio Shack Super tweeters (model 40-1310A) that had been modified to transduce low frequencies. Visual stimuli were presented from small (2.0 mm diam) red (
max = 635 nm) light-emitting diodes (LEDs) that subtended a visual angle of 0.2° and were mounted at the center of each speaker. The transducers were placed in the frontal hemifield, 62 cm away from the cat's head behind a dark curtain that allowed sounds to be heard and LEDs to be seen when lit. The standard acoustic stimulus was a 163-ms noise burst (0.1–25 kHz) with 7-ms rise-fall windows and was repeated as needed to produce bursts of longer durations. All aspects of data acquisition and stimulus presentation were controlled by a MicroVAX II minicomputer (Digital Equipment, Maynard, MA) running custom software.
Physiological recordings
Recording sessions started 5–7 days after the implantation of the recording cylinder, took place five to six times per week, and lasted about 3 h each. The animals were continuously monitored through closed circuit TV for signs of distress. The end of a recording session was typically signaled by the animal falling asleep or stopping due to satiety.
Parylene-covered tungsten microelectrodes (MicroProbe, Potomac, MD) were driven through the intact dura mater with an eccentric Narishige (Narishige International, East Meadow, NY) microdrive. Extracellular signals were amplified and band-pass filtered (300–3,000 Hz). A window discriminator (Bak Electronics, Mount Airy, MD) was used to generate pulses every time an action potential met the spatial and temporal criteria set by the investigator. These were recorded with 1-µs resolution with a custom-made event timer. All data presented in this report are from well-isolated single neurons.
Behavioral tasks
We initially trained the cats on a variety of visual and auditory oculomotor tasks (visual and auditory fixations, standard and delayed saccades, and sensory probes, as described in Populin and Yin 1998b) delivered in random order so that on any given trial the cat could not predict what the task would be. This training period took an average of 7 mo, at the end of which we did a sterile surgery and mounted a recording cylinder on the cat to commence physiological recordings from the SC.
The behavioral experiments presented below used the auditory standard saccade task (Fig. 1A). This task used a visual fixation stimulus (LED), which allowed us to control the initial position of the eyes, and an acoustic stimulus (SPEAKER), which was the target for the saccade. The cat was required to first acquire and maintain fixation on the fixation LED until it was turned off at the same time that the speaker was turned on and then make a saccade to the newly presented target and maintain fixation until the sound was turned off. A reward was delivered if temporal and spatial criteria were met. The rationale for establishing the spatial criteria for success is discussed in Populin and Yin (1998b).
|
), which was used for training because it required the cats to treat acoustic stimuli as targets for saccadic eye movements. In the sensory probe task, the cat is required to fixate for a predetermined period of time on one of three to five fixation LEDs located along the horizontal axis. During this period of fixation, a broadband noise burst (probe SPEAKER), presented from another speaker, was used to probe the neuron's responses to the sound. To receive a food reward, the cat had to maintain its eyes within an electronic window, typically (
6 x 6°), centered about the fixation LED until the end of the trial when both the fixation LED and probe speaker were turned off. Usually we varied one or more parameters (e.g., the fixation LED or the acoustic speaker location) during any particular recording session so that a variety of tasks was always presented and the cat could never anticipate what the upcoming trial would be. For the auditory delayed saccade task (data not shown), the fixation led was not turned off until the auditory target had been on for 300–500 ms. The offset of the fixation LED signaled the cat to make a saccadic eye movement to the auditory target. It is important to note that the sensory probe and delayed saccade tasks are identical up to the time that the fixation LED is extinguished to signal the saccade. Ideally, we would have preferred to use the delayed saccade task for our recordings, but the best response area of most units was outside the oculomotor range of the cat, which is approximately ±20°. We are confident that the cat considered the probe speakers as potential saccade targets because when we switched to runs with auditory delayed saccades or auditory fixation tasks, they immediately made short-latency saccades to those targets.
Pinnae immobilization
Our behavioral studies showed that cats move their external ears in a systematic fashion in response to acoustic stimuli and when fixating at different spatial positions (Populin and Yin 1998c). Therefore the position of the ears varies as the eye position changes. To control for changes in the acoustic input to the eardrums brought about by the moving pinnae, we immobilized them for a final set of physiological recordings in one cat. We would have preferred to immobilize them reversibly by anesthetizing the branches of the facial nerve while holding a single neuron, but this proved too difficult. We also tried taping the pinnae, but this was unsuccessful in holding the pinnae steady. Finally, we resorted to acutely denervating the pinnae. Under sterile conditions, we cut the auriculoposterior and temporal branches of the facial nerve that innervate the caudal and rostral pinnae musculature, respectively. This procedure was successful in immobilizing the pinnae for about 3 or 4 wk, until the nerves regenerated, and presumably also deafferented the pinnae so that the proprioceptive signals were also severed.
Data analysis
In the psychophysical experiments, the dependent variable of interest was final eye position after saccadic eye movements to the acoustic targets. A criterion based on the variability of the velocity of the eyes during steady fixation was used to mark the start and end of an eye movement (Populin and Yin 1998b). Briefly, the end of and return to fixation were determined by the time at which the amplitude of the velocity function exceeded and returned to within 2 SD of the mean baseline velocity, respectively. The mean velocity baseline was computed from 100 ms before to 30 ms after target onset, a period during which the eyes were expected to be stationary. The eye position at the time of the return to fixation was considered to be the final eye position for a saccade. If corrective eye movements took place before the reward or within 500 ms after the end of fixation, the return to fixation measurement was carried out after the completion of the last movement.
Dot rasters and histograms were constructed to qualitatively evaluate neuronal responses. For quantitative evaluation, the mean number of action potentials per trial occurring within a predefined window was determined. For most neurons, a 500-ms window, starting 5 ms after the onset of the acoustic stimulus, was used. The auditory RF of a SC neuron is operationally defined as the area of space within which a stimulus evokes a response.
Because our primary concern was the role of eye position on the responses of SC neurons to acoustic stimuli, we usually began by studying the responses of the neurons to a probe from a single fixed position while varying the eye position using the sensory probe task. To determine if auditory responses were modulated by fixation position, 95% CIs were computed for the mean response from each fixation position. The auditory responses of a neuron were classified as modulated by fixation position if one pair of responses to different fixation positions was significantly different (P < 0.05). The exact shape of the resulting functions relating neural responses to fixation position is determined by the horizontal position of the speaker selected for studying the neuron, the shape of the RFs, and the way the RFs are modulated by fixation position.
When stable recording conditions and a cooperating animal coincided, a more complete representation of the neuron's RF was obtained by recording the responses evoked from sounds presented from different speakers along the horizontal axis for three to five different fixation positions. The data were plotted both as a function of the location of the speakers relative to the animal's head and as a function of the distance between the fixation point and the speakers (i.e., the motor error). The extent to which the RF functions obtained for the different fixation conditions shifted and became aligned or misaligned because of the change in coordinates constituted the test of the motor error hypothesis.
We computed the shift in azimuth of the RFs that resulted from two different eye positions relative to the eyes-centered condition as follows. For example, for each data point from the RFs measured when the cat fixated to the left, we measured how much that point was shifted in azimuth relative to the RF when the eyes were fixating straight ahead. This was done by computing at each data point the difference in azimuth at corresponding discharge rates, as shown by the dashed arrow in Fig. 2. This process was repeated using the data points from the RF obtained when the cat fixated at the center, as indicated by the solid horizontal arrow in Fig. 2. When the azimuth at which the line intersected the RF fell between data points, the azimuth of intersection was determined using linear interpolation. Shifts to the right of the center RF were positive, while shifts to the left were negative. These differences in azimuth were averaged. When measuring the differences, care was taken to respect the shapes of the RFs; for example, when RFs exhibited a "peak" response, the difference in azimuth took into account the medial and the lateral portions of the RFs to obtain a true estimate of the shift in the RFs. Neurons with RFs that were complex with many peaks and dips were not included. A motor error index (MEI) was computed by taking the ratio of the computed shift in the RFs and the size and direction of the shift in eye position. When the RFs exhibited a "peak" type profile, characterized by responses on either side of the peak that fell by 20% or more, an MEI was computed also for the medial edge of the RF only; this analysis is comparable to that performed by Jay and Sparks (1987). An MEI of 1.0 resulted if the RF shifted in the same direction and with the same magnitude as the shift in eye position. This process was repeated for RFs recorded with fixation to the right. For each neuron, we plotted the MEI for leftward fixations against the MEI for rightward fixations. Data points that fell in the upper right quadrant were therefore consistent with the motor error hypothesis, with perfect motor error for points at (1, 1).
|
After completion of the experiments, the cats were anesthetized with an intramuscular injection of ketamine and acepromazine, and two to three pins were placed in the brain with the recording microdrive at known positions within the recording chamber, rostrally and caudally to the recording area, and in parallel to the electrode penetrations. Figure 3 shows Cat6's midbrain with one pin inserted caudally to the left SC, and another pin inserted rostrally to the right SC. The hash patterns overlaying each SC represent the approximate areas in which electrode penetrations were made in relation to the location of the pins.
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
The first consideration was to determine whether Pöppel's (1973) paradox was behaviorally relevant, i.e., whether there was any effect of initial eye position on sound localization accuracy. To address this question we used the auditory standard-saccade task (Populin and Yin 1998b) with different initial fixation points (LED in Fig. 1A) and broadband noise stimuli as acoustic targets. Trials of this type were randomly intermixed among different types of visual and acoustic trials to numerous other targets (Populin and Yin 1998b). Figure 4 shows eye movements to an acoustic target at (0, 24°) recorded under those conditions. The cat was required to fixate LEDs at (±4, 0°) for 1,000 ms or an LED at (0, 0°) for 500 ms. The dots that form each trace represent digital samples of the analog output of the eye coil system. In 90% of acoustic trials on the horizontal plane, cats oriented with a single saccadic eye movement; for acoustic targets on the vertical plane, 95% of trials were completed with a single saccade. The slope of the main sequence computed from these eye movements was consistent with those measured in our previous work (Populin and Yin 1999). Data from six cats with different levels of experience were analyzed to address this question.
|
), subject C9 undershot the acoustic target. Although the trajectories of the eye movements were different in the three fixation conditions, the eye movements ended up in roughly the same spatial location. The eyes took the most direct path when the fixation LED was on the same side of the target (9, 0°), and the most curved when the fixation LED was on the opposite side (–9, 0°).
|
The data shown in Fig. 5C are from subject C5's 11th experimental session. The acoustic target was located at (8, 13.5°). This was the first session in which subject C5 was presented with this acoustic target. The trajectories of the eye movements were similar to those of subject C9 in Fig. 5A. Although they show more corrections than the other subjects, they consistently reached the target.
For the experiments in Fig. 5, D and E, the acoustic target was located at (–18, 0°), and the initial starting positions were at (±9, 0°) and (0, 0°). Subjects C6 and C14 were well practiced at the time of the recordings, which were the 81st–83rd and 159th–163rd behavioral sessions, respectively. Both subjects mislocalized this acoustic target, although with different types of errors: while subject C6 localized the target below the horizontal plane, subject C14 localized it above. Interestingly, the pattern of errors made by each subject was consistent across the three experimental conditions.
Last, the data shown in Fig. 5F are from subject G7 and were collected in the 55th–60th behavioral sessions. The acoustic target was placed at (0, 24°), and the starting points were at (±6, 0°) and (0, 0°). Unlike the previous subjects, G7 exhibited a clear bias in its final eye position, which is indicative of incomplete compensation for the initial position of the eyes along the horizontal plane.
Figure 6 summarizes localization performance from the subjects presented in Fig. 5; data from some additional targets are included. In each panel, the position of acoustic targets is plotted with an asterisk, the starting fixation positions are plotted with filled symbols, and the corresponding mean final eye positions are plotted with open symbols of the same shape. Error bars represent 95% CIs computed for the x- and y-axes separately. All trials, successful and unsuccessful, were included in the analysis.
|
Subjects C6's and C14's localization of an acoustic target at (–18, 0°) showed an effect of eye position, as revealed by the nonoverlapping 95% CIs computed for each of the three mean final eye positions. Subject C6's data (Fig. 6D) show a significant effect of eye position for the target at (–18, 0°) only. The magnitude and sign of the mean error for the eye movements initiated from the primary position (0, 0°) compared with those starting from fixation position to the right (9, 0°), but not to the left, of the midline are nearly proportional to their initial separation. There was no effect of initial eye position on the localization of other acoustic targets such as those located above (0, 9°) and to the right (18, 0°) of the primary position.
Subject C14 (Fig. 6E) made large, yet consistent, localization errors on both the horizontal and vertical components in all three conditions. However, the differences among the three mean final eye positions cannot be predicted based on the initial fixation position. Specifically, the final eye position for the (±9, 0°) starting fixation points was not significantly different. Furthermore, the localization error was smaller in the (9, 0°) than in the (0, 0°) fixation condition, which is the opposite of what one would expect if initial eye position was not taken into account to perform the eye movements to the acoustic target. Subject G7 (Fig. 6F) also showed a mixed pattern, i.e., an effect of initial eye position for some targets, those located along the vertical axis, but not an effect of initial eye position for others, those located along the horizontal axis.
Thus for the six cats tested, only one of them showed a consistent mislocalization that was related to initial eye position. The accuracy of saccades to auditory targets of three of the cats was unrelated to initial eye position, and for two cats, there was a small but inconsistent effect.
On average, the mean horizontal localization error computed from 267 trials from four cats, across three different initial eye positions, for targets with large horizontal components, was an undershoot of 35%.
Effects of fixation position on single SC neuron responses
GENERAL OBSERVATIONS.
As a microelectrode approached the SC from above, a predictable pattern of active and quiet areas was found in most penetrations. In cortex, we encountered visually driven activity followed by lack of activity in the superior cistern, which indicated that the SC was near. The signature of the superficial layers of the SC was intense neural activity that changed in rate as the cat moved its eyes or visualstimuli were swept across the screen in front of the cat. The onset of this pattern was abrupt and in marked contrast with the lack of activity just above. The area with this type of activity extended dorso-ventrally about 1.5 mm. The position of the tip of the electrode within the visual map of the SC (Feldon et al. 1970) was inferred from the responses evoked by stimulating discrete areas of the contralateral visual field with a visual stimulus while the cat fixated an LED at the primary position (0, 0°). We did not attempt to isolate neurons in the superficial layers of the SC. We searched for neurons that responded to acoustic stimuli as we lowered the electrode into the intermediate layers of the SC using a mixture of trials with visual and acoustic stimuli. In the caudal SC, auditory neurons were typically found between 1.5 and 3.5 mm ventral to the top of the superficial layers, but in approaching the rostral SC, the layer containing auditory neurons became progressively thinner (<0.5 mm), with fewer auditory neurons.
On isolating a neuron responsive to sound, we determined the stimulus position (from our available target locations) from which the strongest response could be evoked as well as the approximate threshold. The most sensitive speaker position was selected, and the level of the stimulus was set at 20 dB above threshold. Our first priority was to determine if the responses of the neuron using the sensory probe task, evoked from the most effective source location, were modulated by eye position. In cases in which we were able to maintain the recording for a prolonged period, we characterized other parts of the neuron's RF along the horizontal axis. As reported previously (Populin and Yin 2002), the responses of most SC auditory neurons to identical acoustic stimuli were smaller in the sensory probe task compared with the auditory fixation task. In some cases, the effect was so strong that there was virtually no response to the sensory probe.
With few exceptions, the 143 SC auditory neurons in our sample responded to broadband noise stimuli presented from sources in the contralateral field using the sensory probe task. Figure 7 shows the responses of a typical SC neuron to an acoustic sensory probe with the fixation point at (0, 0°). The cartoon at the center of Fig. 7 shows the position of the speakers relative to the fixation point. From top to bottom, Fig. 7A shows vertical and horizontal eye position plotted as a function of time, a dot raster showing the responses of the neuron in individual trials, and a histogram of the number of spikes/bin normalized by the number of trials, summarizing the responses of the neuron in the trials shown. The data are plotted synchronized to the onset of the acoustic stimulus at 0 ms. The eye position traces corresponding to the other targets were similar to those plotted in Fig. 7A, thus they were omitted for simplicity. This neuron, recorded from the right SC, responded most strongly to stimuli presented from the left hemifield (Fig. 7B). The response of this neuron to a long-duration (1,000 ms) broadband noise stimulus consisted of a burst of two to five action potentials with a mean first spike latency of 18.3 ms.
|
If the auditory responses of SC neurons are modulated by eye position, it would indicate that auditory RFs are dynamic and that they might not encode the position of sound sources in head-centered coordinates. To address this issue, we used the sensory probe task (Fig. 1B) with different fixation positions and a single source of acoustic stimulation. Figure 8 shows responses of a neuron that was modulated by eye position. The three columns show responses of a neuron in the right SC to the same speaker at (–18, 0°), the area of space where the strongest responses could be evoked, while the cat was fixating three different positions: (±9, 0°) and (0, 0°). The largest responses (Fig. 8C) were recorded when the cat fixated an LED at (9, 0°), the farthest away from the sound source, and the smallest responses (Fig. 8A) were recorded when the cat fixated the LED closest (–9, 0°) to the sound source. Trials from each of the three fixation conditions were presented in random order, thus long-term changes in variability should be averaged out. The results are summarized in Fig. 8D, where mean spike count, normalized by the number of trials, is plotted as a function of horizontal fixation position. The error bars, representing 95% CIs, indicate that the mean responses recorded at the three fixation positions were significantly different. We studied 106 neurons in the SC using the test shown in Fig. 8 with a single speaker and three to five fixation positions, and three-fourths of them (81/106) were significantly modulated by eye position. Most of the neurons that were modulated by eye position were similar to that shown in Fig. 8, although we also found neurons whose largest responses occurred when the cat fixated the LED nearest the target.
|
Figure 8 shows the responses of a neuron to acoustic stimuli presented from a single source recorded at three different eye positions. Ideally, one would like to extend these observations to a larger portion of a neuron's RF to determine if such changes in neural response were due to a shift in the RF as a function of eye position. Recordings of this type were possible in 28 neurons, 3 of which are shown in Fig. 9.
|
On the other hand, the RFs shown in Fig. 9, B and C, are more complete because we captured the peak and both flanks. Neuron C6-59 (Fig. 9B) had a narrow horizontal RF, which unlike most neurons, extended into ipsilateral space. The three functions, corresponding to the RFs measured in each of the three fixation conditions, are clearly separated at both the medial and lateral aspects of the field. Plotting the same data as a function of horizontal motor error aligns the three curves at both the medial and lateral edges, yielding a consistent measure of RF regardless of fixation position. RFs with similar profiles, although broader, were obtained from other neurons (e.g., Fig. 9C). The RFs of this neuron's medial edge were well aligned when plotted as a function of horizontal speaker position. However, unlike the neuron in Fig. 9B, plotting the same data as a function of horizontal motor error increased the separation between the curves.
To determine the extent to which RFs shifted as a function of fixation position, we computed a MEI as described in METHODS. The results of this analysis, taking into account all RF data available from each of the 28 neurons, are shown in Fig. 10A. For each neuron, the MEI due to leftward fixation is plotted against the MEI due to rightward fixation. Thus RF shifts consistent with the motor error hypothesis will fall in the upper right quadrant, with perfect motor error expected at an MEI of (1, 1). No shift would be indicated by an MEI of (0, 0). The mean MEI from all 28 neurons (0.65, 0.66), although smaller than (1, 1), is significantly different from (0, 0) because the 95% CIs do not encompass (0, 0). Thus the horizontal RFs of the neurons in our sample shifted in a direction consistent with the motor error hypothesis.
|
, who measured the magnitude of auditory RF shift in the monkey by taking into account only the medial edge, we computed the MEI for only the medial edge of the RFs of our neurons as described in METHODS. The filled symbols in Fig. 10B show the MEI for the medial edge of neurons that had a peak type of RF (n = 16), as defined in METHODS, and the open symbols show the MEI for the remaining neurons (n = 13). The mean MEI from this analysis (0.81, 0.48) is also consistent with the motor error hypothesis, in agreement with the 54% compensation reported by Jay and Sparks (1987). Potential contributions of the mobile pinnae
The modulation of auditory responses observed in the behaving cat appears to be similar to the modulation observed in the macaque monkey by Jay and Sparks (1987). Casual observation of the movements of the pinnae of both species indicates that cat's pinna movements are more prominent and thus more likely to change the acoustic input to the eardrum (Young et al. 1996), which can result in modulation of neuronal responses (Middlebrooks and Knudsen 1987).
Previously, we have documented that orienting behavior of cats with restrained heads to acoustic and visual targets involves stereotyped and coordinated movements of both the eyes and the pinnae (Populin and Yin 1998c). Because we required the cats to fixate on LEDs at different spatial locations to control eye position at the time of presentation of the acoustic stimuli, this would have systematically changed the position of the pinnae, which could have in turn altered the acoustic signals reaching the cat's eardrums. Such changes in acoustic input could account for the modulation of auditory responses in the SC documented above.
Figures 11 and 12 show the responses of two neurons, one which showed modulation by eye position and one which did not. The neuron in Fig. 11 displayed atypical responses because it was best driven by acoustic stimuli presented from an ipsilateral source (–18, 0°). We selected this neuron for illustration because the position of the left pinna was also recorded in the same trials. Five fixation positions were used: (±18, 0°), (±9, 0°), and (0, 0°). From top to bottom in Fig. 11A, each row of panels shows horizontal eye position, left horizontal pinna position, the responses of the neuron in individual trials in raster format, and a histogram normalized by the number of trials summarizing the neuron responses. As expected from our previous observations (Populin and Yin 1998c), the position of the left pinna at the time of stimulus presentation was a function of fixation position; a vertical arrow in each panel points to this portion of the pinna position records. The responses of neuron C6-75 to broadband noise stimuli presented from the same source at (–18, 0°) were significantly modulated by fixation position, and possibly, pinna position (Fig. 11B).
|
|
The ear movement traces in Figs. 11 and 12 confirm that there were systematic changes in ear position related to the fixation of the LED at different spatial locations. Therefore there were also undoubtedly systematic changes in pinnae position during the collection of the data depicted in Figs. 8 and 9, each of which showed modulation with eye position. Thus the question arises as to whether the modulation with eye position was due to the change in acoustic input resulting from the movements of the pinnae or to an eye position signal that modulated the responsiveness of the neurons. On the other hand, the lack of modulation in Fig. 12 shows that changes in pinna position do not necessarily change the responsiveness.
Effects of pinna immobilization
The data presented in the previous section suggest the possibility that pinna movement associated with changes in fixation could account for the modulation observed in auditory responses of SC neurons. Changes in pinna position in the anesthetized cat have been shown to affect the spatial tuning of single SC neurons (Middlebrooks and Knudsen 1987). To see if eye position, independent of ear position, is a source of modulation of auditory responses in the SC of the behaving cat, we immobilized the pinnae of one cat by cutting the auriculoposterior and temporozygomatic branches of the facial nerve bilaterally and recorded the responses of single SC neurons to broadband acoustic stimuli.
Typical eye and pinna movement data recorded simultaneously with the auditory sensory probe task from the intact cat are shown in Figs. 11 and 12 before the immobilization surgery. At the time of these recordings, just before the pinnae immobilization procedure, subject C6’s right pinna coil was broken, so baseline data only from the left pinna were acquired at the time. We chose not to submit subject C6 to an additional surgical procedure to repair the broken coil because baseline data from this structure, showing its movement, had already been recorded (Fig. 7 in Populin and Yin 1998c). The pinna adopted different positions during the period preceding the onset of the acoustic probe, corresponding to the different positions of the fixation LEDs. The largest pinna movement was recorded when the cat fixated the LED ipsilateral to the pinna, at (–18, 0°), and the smallest change was recorded when the cat fixated on the contralateral LEDs at (18, 0°) (Figs. 11 and 12A). Also consistent with our previous findings (Populin and Yin 1998c) are the short latency movement of subject C6's left pinna after the onset of the acoustic target.
Figure 13 shows that cutting the branches of the facial nerve was effective. After the pinnae immobilization procedure, C6's eyes continued to move normally (Fig. 13D), but there were no movements of either pinna when the cat fixated to the left or to the right (Fig. 13, C and E). Both the left and right pinnae remained motionless for 3 wk, after which time they began to recover their ability to move, presumably due to nerve regeneration. The single neuron recordings presented below were carried out while subject C6's pinnae were immobilized.
|
With the pinnae and head immobilized and the same acoustic stimuli presented while the cat fixated LEDs at different positions, any modulation of auditory responses in single SC neurons could not be attributed to changes in the acoustic input to the eardrum. Figure 14 shows the responses of a single neuron recorded from the left SC under those conditions. Broadband noise stimuli were presented from a speaker located at (63, 0°) at about 10–15 dB above threshold. The responses were transient, with little or no activity following the initial burst. Most importantly, they were significantly modulated by eye position (Fig. 14B). Most of the modulation was achieved within the initial burst. This neuron, like many in the deep and intermediate layers of the SC (Meredith and Stein 1983; Populin and Yin 2002), was bimodal and also responded to the visual presentation of the fixation LEDs when they were on the midline (Fig. 14A, middle) and in the contralateral hemifield (Fig. 14A, right). Arrowheads point to this transient activity in the histograms (Fig. 14A), which is synchronized to the presentation of the fixation LEDs, but because of the variable reaction time to make the saccade to the LED, they are not synchronized to the onset of the acoustic stimulus that was used to make the plots in Fig. 14.
|
We were able to characterize the horizontal component of the RFs of four neurons more extensively. Two examples of those recordings are shown in Fig. 15. In both cases, the RFs became more misaligned when plotted as a function of motor error. Figure 16A shows the MEIs based on all available RF data from these neurons, and Fig. 16B shows the MEIs computed only from the medial edge of the RFs. Data from one of the neurons were obtained at (±9, 0°) and (±18, 0°), which are plotted separately. Regardless of the type of analysis, none of the MEIs measured fell within the upper right quadrant, indicating that the neurons were modulated in a manner inconsistent with the motor error hypothesis. The mean and 95% CIs from the intact cat data (Fig. 10) are shown for comparison.
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Behavioral results
The behavioral data show that sound localization accuracy is generally not affected by the deviation of the eyes in the orbit at the time of acoustic stimulus presentation. This is not a surprising finding for grossly mislocalizing sounds simply because the eyes were not at the primary position would constitute a severe handicap for an animal. This means that the saccadic command to move the eyes must compensate for current eye position when a saccade is made to the same target from two different initial positions. For three of the six cats, there was no significant effect of initial eye position on the accuracy of saccades to targets, and for two other cats, there was a small but inconsistent effect. In only one cat (Figs. 5F and 6F) was there a consistent but small bias such that saccades starting from the left toward a vertical, but not a horizontal, target had a consistent error to the left, while those starting from the right had a rightward error. These results contradict Harris et al. (1980) in that they show that cats do compensate for deviations of the eyes from the primary position, but the results are in accord with both Hartline et al. (1995), who reported little effect of eye position on sound localization for deviations of about ±15°, and Peck et al. (1995), who reported partial compensation only. These data are also consistent with those from human (Yao and Peck 1997; Zahn et al. 1979) and monkey (Whittington et al. 1981) studies that found no effect of initial eye position on the accuracy of eye movements to acoustic targets.
Modulation of auditory responses in the deep and intermediate SC by eye position
The physiological data from the intact cat show that the responses of most SC auditory neurons (81/106) are modulated by eye position. In 28 neurons, we were able to sample the RFs more extensively along the horizontal axis using different points of fixation (Fig. 9). These data allowed us to examine the extent to which the horizontal component of the RFs of auditory neurons shifted as a function of fixation position. Based on the MEIs computed from the RFs, 68% of the neurons were modulated by fixation position in a direction consistent with the motor error hypothesis. The percentage increased to 76% when the MEI was based on the medial edge.
Since many neurons in the deep and intermediate layers of the SC are multimodal, responding to both visual and auditory stimuli (Meredith and Stein 1983), we should consider whether there is a visuomotor interaction that affects the eye position sensitivity shown in these experiments. Sixty-three percent of the neurons we studied were indeed bimodal (see example in Fig. 14 and Populin and Yin 2002). The example in Fig. 14 is typical of the short, transient visual response, which is over for almost 400 ms by the time the acoustic stimulus is turned on. While it is possible that there is some lingering inhibitory effect of the visual response on the auditory response many hundreds of milliseconds later, the parsimonious explanation would suggest that the effect is due to eye position and not to visio-motor interaction. Note that at the time the acoustic target is turned on, the eye is always fixating the LED, so there should be a constant visual input on every trial.
Thus our data show that 1) the responses of 76% (81/106) of auditory neurons are modulated by eye position, 2) the modulation of 68% (19/28) of neurons from which a more complete sampling of the RF is available is consistent with the motor error hypothesis, and 3) the mean MEIs, (0.65, 0.66) computed from all available RF data and (0.81, 0.48) for the MEIs computed from the medial edge of the horizontal RFs, account for a compensation of approximately two-thirds of the shift required for perfect motor error.
Because the cats undershot the targets significantly (Figs. 5 and 6), a more appropriate measure of the motor error would be the distance between the initial position of the eyes and the apparent location of the targets as indicated by the endpoint of the saccade. We refer to this as behavioral motor error (BME). Ideally, we would have preferred to compute the MEI for each neuron using BME and not the physical location of the targets relative to the head. Such MEIs would have accounted for a larger proportion of the deviation of the eyes from the primary position. Although we could estimate the BME from our psychophysical results (Figs. 5 and 6), we were unable to compute BME for the neurons we studied for the following two reasons. The oculomotor range (approximately ±20°) of the cat limits this analysis to neurons with RFs in a very small portion of the auditory field of most SC neurons, which extend out to 90° or more. Releasing the head to allow the cat to orient to targets anywhere in the frontal hemifield would have prevented us from systematically manipulating the position of the eyes in the head at the time of stimulus presentation of the acoustic stimuli. Despite these limitations, our results, computed as a function of target position, agree with those of Jay and Sparks (1987), who reported a compensation of 54% in the SC of the monkey and thus suggest that audio-visual interactions in monkey, a species with a larger oculomotor range, and cat SC are governed by similar rules. However, if we take into account the magnitude of the horizontal error made by our cats in localizing acoustic targets on the horizontal plane (Figs. 5 and 6), which on average was an underestimate of the targets by 35%, the mean MEIs computed from all available RF data (0.65, 0.66) actually account for the observed behavioral responses.
Our results suggest that the modulation with eye position seen in SC neurons correlates better with the actual motor response of the cat (undershoot of about 35%) than to the motor error. In the monkey, though there are data to suggest that the SC neurons are correlated better to motor error than to actual motor response (Frens and Van Opstal 1998; Stanford and Sparks 1994).
Effects of pinna immobilization on auditory responses in the deep and intermediate SC
Cats, however, move their pinnae in coordination with the eyes during the oculomotor tasks used in this study (Populin and Yin 1998c). Macaque monkeys make large pinna movements in response to unexpected sounds, but do not seem to hold the position for an extended time (Jay and Sparks 1987). The systematic changes in pinna position associated with fixating LEDs at different horizontal positions observed in our cats (Figs. 11 and 12) led us to consider that the modulation of responses to acoustic stimuli of SC neurons could be, at least in part, due to changes in the acoustic input to the eardrum.
To determine directly if the modulation of auditory responses in single SC neurons was due to the changes in the acoustics, we surgically immobilized the pinnae of one cat. Failure to observe modulation would have supported the hypothesis that acoustic factors were solely responsible for the modulation seen in the intact cat. Because all 24 neurons studied in the immobilized pinnae condition showed modulation by eye position, we reject the hypothesis that the modulation is caused exclusively by acoustic factors.
The patterns of the modulation of the spatial RFs documented in four neurons after immobilizing the pinnae, however, were inconsistent with our observations in the intact cat as assessed by the MEI (Figs. 10 and 16). It is possible that this discrepancy is due to the method by which we computed the MEI. As pointed out above, an MEI computed using the perceived target location rather than the actual location is likely a more appropriate measure of the motor error. Several factors precluded us from computing such a measure. First, as was the case in the experiments with intact cats, it was not possible to document this cat's localization accuracy at the range of eccentricities required to match those used to measure the auditory neurons' RFs. Second, due to the limited recording time under pinnae-immobilized conditions afforded by the regenerating facial nerves (3 wk), we were not able to conduct extensive psychophysical sound localization studies, focusing instead on the physiological recordings.
The data suggest that both an eye position signal and a pinna position signal contribute to shape motor error (Figs. 9 and 10). In the intact cat, copies of motor commands sent to the pinna may be used directly as a position signal or may be first evaluated against proprioceptive signals from the pinna. In either case, when the pinnae were immobilized, a mismatch would be expected between the desired position of the pinnae, the expected sensory consequences (i.e., acoustic filtering), and the eye position signal. While we are confident to have ruled out acoustical factors as the only source of the modulation, the small number of neurons recorded in the pinnae immobilized cat temper any conclusions drawn from the results following pinna paralysis.
Previous studies in the cat aimed at determining whether auditory RFs in SC neurons shift as a function of eye position produced conflicting reports. Harris et al. (1980) found no modulation in a sample of three neurons in the awake, untrained cat. They concluded that complex interactions between the visual and auditory representations in the SC are not needed because cats keep their eyes near the primary position most of the time, thus maintaining the alignment between the various sensory representations in the SC. Our results do not agree with this conclusion. Aside from their small sample of neurons, other methodological problems compromise their conclusions, including cats that were untrained, stimuli delivered without acoustic shielding, and eye positions that were averaged over large ranges.
The conclusions of more recent studies (Hartline et al. 1995; Peck et al. 1995) are consistent with our findings, but are difficult to evaluate. Hartline et al. (1995), for example 1) used essentially untrained cats, which made controlling eye position and distinguishing between sensory and motor responses problematic, 2) taped the pinna coils to the ears, which made them subject to slippage, 3) presented only a small amount of data, which made it difficult to evaluate their reliability, and 4) collapsed responses over relatively large ranges of eye and pinna positions. Peck et al. (1995) reported that their cats did not move their pinnae while performing their experimental tasks, although no records of pinna movements were presented. Our results (Figs. 11 and 12; Populin and Yin 1998c) show that cats with immobilized heads move their ears extensively and systematically in response to both visual and acoustic stimuli, even after thousands of trials. In agreement with our findings, Hartline et al. (1995) also reported that eyes and pinna tend to move in the same direction in a correlated fashion.
In summary, these results show the presence of an eye position signal that modulates the responses of SC neurons to acoustic stimuli. Eye positions signals in the SC are thought to originate from corollary disc
) and 15° to the left (
). For each data point on both curves, we measured the difference in azimuth between the 2 curves at the corresponding discharge rate, keeping track of the shapes of the RF and the sign of the shift and interpolating linearly between data points when necessary. For example, the horizontal separation between the data point at 10° on the Left Fixation RF to the Center Fixation RF is –7.5°, while the separation between the data point at 20° on the Center RF to the Left Fixation RF was –7°. This shift was measured for all data points on both RFs where possible and averaged. For example, the shift for the leftmost point on the center RF and the rightmost point on the left fixation RF cannot be measured. The mean RF shift of this unit, computed by taking into account all data points, was –7.6, which yields an MEI of 0.51 for a fixation point located 15° to the left.
