Research Projects of Binaural Physiology Lab
As expected, the responses were consistent with the ILD sensitivity of the cells, with a minimal response when the AM signals were in phase (above). Characteristic delays of LSO cells were studied by varying the modulation frequency. As expected the characteristic phases of LSO cells were clustered around 0.5, reflecting the IE binaural interaction. By recording AM responses from each of the different cell types in this circuit (below), we could trace the gradual decline of phase synchronization at each synaptic relay. We found that the restricted modulation frequency range over which LSO cells show ITD-sensitivity does not result from loss of envelope information along the afferent pathway but is due to convergence or postsynaptic effects at the level of the LSO. Measurements of time delay through the two sides of the circuit to the LSO showed that on average the contralateral inhibitory input arrived only 200 microsecs later than the ipsilateral excitatory input, even though the contralateral pathway is longer and has an extra synapse. This is in accord with characteristic delay measurements and estimates derived from single cells.
Physiological studies of binaural interaction
- Anatomical studies of binaural circuits
- Behavioral studies
Links on this page will provide either abstracts or manuscripts of published work from the lab.
You may need the Adobe Acrobat Reader, which you can download from Adobe.
There are three primary cues that are important for sound localization: interaural time disparities (ITDs), interaural level disparities (ILDs), and spectral cues. The two binaural cues, ITDs and ILDs, are encoded in the auditory brainstem nuclei in two parallel circuits that start in the bushy cells of the anteroventral cochlear nucleus (AVCN). The spherical bushy cells provide input to the medial superior olive (MSO), which encodes ITDs using a Jeffress-like coincidence model. ILDs are encoded in the other two primary nuclei of the superior olive: the lateral superior olive (LSO) and medial nucleus of the trapezoid body (MNTB). LSO cells receive excitatory input from the ipsilateral AVCN and inhibitory input from the contralateral AVCN, relayed through the MNTB.
Spectral cues are provided by the filtering action of the external ears and head; they are believed to be important for localization in the vertical dimension though there is little direct evidence for this hypothesis. A major long-term goal of the lab is to understand the anatomical and physiological features of these circuits.
Return to top
- Responses to virtual space stimuli
We have studied the auditory receptive fields of cells in the LSO using the virtual space technique. LSO cells responded in a manner consonant with their IE binaural interaction: they responded best to stimuli delivered from the ipsilateral sound field and their response to stimulation with ipsilateral ear alone was nearly always greater than their response to binaural stimulation. Thus the majority of LSO cells fell into the binaural inhibited category. To determine which localization cue contributed to the spatial sensitivity, we manipulated the head-related transfer functions by independently varying (or holding constant) in azimuth each of the three cues while holding constant (or varying) the others. For most cells, ILD was the most effective cue.
Return to top
- Responses of low frequency LSO cells
The classical studies of the lateral limb of the LSO describe cells with low CF which are monaural, rather than the binaural IE cells found in the other limbs of the LSO. However, our earlier studies of the MNTB indicated the presence of inhibitory inputs to LSO cells of all CFs. Therefore we undertook to study the low frequency limb in more detail. We found that most cells were binaural, not monaural, and like their high frequency counterparts, showed IE binaural interaction. They were excited by the ipsilateral ear and inhibited by the contralateral ear. When stimulated with binaural beats, they responded best when the stimuli were out of phase, and had characteristic phases near 0.5. Their ITD functions showed troughs near 0 for both tones and noise, in agreement with their IE binaural interaction.
- Dorsal nucleus of the lateral lemniscus (DNLL)
The DNLL provides GABAergic inhibitory input to bilaterally to the inferior colliculus and to the contralateral DNLL, but the function of this inhibitory input is not known. We have been recording intra-axonally from DNLL cells in the lateral lemniscus and characterizing them physiologically and anatomically after injection of neurobiotin. Preliminary results show that low frequency DNLL cells are sensitive to interaural phase and have characteristic phases near 0.
Return to top
- Binaural interaction in the central nucleus of the inferior colliculus (ICC)
All auditory information ascending to the auditory cortex converges in the ICC. We have continued our long-standing studies of binaural interaction in the ICC.
- Physiological correlates of the precedence effect
The precedence effect (PE) is an illusion produced when two similar sounds are delivered in quick succession (interclick delays of 2-8 msec) from sound sources at different locations so that only a single sound is perceived. The localization of the perceived sound is dominated by the location of the leading sound. If the delays are very short (< 1-2 msec), summing localization occurs and a phantom source is perceived whose location is toward the leading sound. Most cells, exhibited a form of the precedence effect in which the response to the lagging click was suppressed when ICDs were short. The mean half-maximal suppression varied considerably from cell to cell. With short ICDs in the summing localization range, cells also showed responses consonant with the human psychophysical result that the sound source is localized to a phantom image between the two speakers and toward the leading one. Stimuli delivered in free field along the median sagittal plane also exhibited a precedence effect. We also systematically explored the effect of varying the position of the leading speaker in free field and its duration and level.
Return to top
- Virtual space receptive fields in ICC
Using the technique of virtual auditory space, we have studied the receptive fields for virtual-space stimuli in the ICC and compared them with existing free field data in the literature. In addition by comparing the monaural and binaural virtual space receptive fields, we could derive the binaural interactions of the cell using two new measures: the binaural interaction type and binaural interaction strength.
- Intracellular studies
Intracellular recordings using sharp microelectrodes from ICC cells allowed direct examination of the synaptic inputs to ICC in response to binaural acoustic stimulation. In some ICC cells the binaural interaction reflected input from lower centers, while in other cells the interaction appeared to be in the ICC itself. Many complex interactions between excitatory and inhibitory sources so that different synaptic mechanisms could underlie similar response patterns. In some cases, intracellular injection of HRP enabled us to identify the anatomical cell type.
- A weakness in all of the experiments described above is that they are all done in anesthetized animals. The extent to which the anesthesia affects the neuronal responses is difficult to ascertain, but it is not logically tenable to assume that there is little effect. If that were true, then we would have to say that the anesthetized brain is similar to an unanesthetized one. For this reason, among others, we have spent considerable time and effort in developing a preparation that will allow us to record from unanesthetized animals. See Behavioral studies of sound localization below.
- Axonal projections of bushy cells of the AVCN
Using the difficult technique of intracellular recording and injection of HRP or neurobiotin, we have studied the relationship between physiological response type and axonal projections of bushy cells to the superior olivary complex.
Return to top
- Globular bushy cells
Responses of globular bushy cells are highly correlated with the primary-like-with-notch response to pure tones at CF and their axons travel in the most ventral division of the trapezoid body. A prominent projection is to the contralateral MNTB where there is a calyx of Held (right), the largest synapse in the nervous system. In addition there are collateral branches to the ipsilateral lateral nucleus of the trapezoid body, postior and dorsal periolivary nucleus, and lateral superior olive.
- Spherical bushy cells
Spherical bushy cells are associated with the primary-like response to pure tones at CF. Their axons exit the AVCN in ventral stria (or trapezoid body) and travel in the medial division, projecting bilaterally to the MSO. The contralaterally projecting axon has a distinct delay-line (right) form, like that proposed by the Jeffress model, but the ipsilaterally projecting axon does not. All of the contralaterally projecting axons that we injected had a more caudal than rostral projection while maintaining the tonotopic organization in the MSO. This caudally projecting delay line is in agreement with the topographic map of ITDs found in the MSO of the cat.
- Anatomical studies of HRP labeled MNTB cells
Cells in the medial nucleus of the trapezoid body usually have a primary-like-with-notch response, as expected from their calyceal input from globular bushy cells, and tended to have profusely-branching dendritic trees. Their axons projected to the ipsilateral LSO, as expected, from early classical studies, and to the ipsilateral MSO. Electron micrographs of the axonal terminal showed non-round vesicles on the soma and proximal dendrites of the LSO, confirming the inhibitory nature of its projection.
- Anatomical studies of HRP labeled inferior colliculus cells
The inferior colliculus (IC) serves as a critical integrator of inputs from ascending auditory pathways to the thalamus and auditory cortex. It receives input from a multitude of ascending pathways. Because it is more accessible than many of the other lower auditory centers, we also know more about the physiology of cells in the IC (see above). Intracellular labeling of physiologically-identified cells in the ICC revealed a variety of cell types. One cell (as seen in the figure to the right) with an onset response showed an extensive dendritic tree (red) with an even more impressive axonal collateral system (black). Other cells had the expected disk-shaped dendritic tree oriented parallel to the isofrequency contours in the IC.
Return to top
- Sound localization with the head fixed
We studied the sound localization behavior of cats by training them using standard operant techniques to look at light and sound sources with their heads restrained. Their eye movements were monitored by using the scleral search coil technique. Saccades to auditory targets were less accurate and more variable than saccades to visual targets at the same spatial positions. We also studied the effect of using narrow bands of noise and stimuli that mimicked the precedence effect.
- Pinna movements during sound localization
By implanting a search coil subcutaneously on each ear, we monitored the movements of the external ear while the cat was working in our sound localization behavioral paradigm. Pinna movements (right) were prominent, systematic, and did not habituate while the animal oriented to either visual or acoustic stimuli. Pinna movements to acoustic stimuli had shorter latency than those to visual stimuli.
- Kinematics of eye movements to acoustic targets
The eye movements to acoustic targets exhibited an unusual slow velocity ramp which preceded saccades to acoustic, but not visual, targets. The slow ramp was seen before saccades to long duration noise bursts but not before saccades to clicks and during delayed saccades.
- Sound localization with the head free When we let the cat localize with the head unrestrained, the accuracy and precision of localization performance for acoustic stimuli improved dramatically as compared to the head fixed condition. The figure to the right shows the saccades to four different targets on the horizontal plane (target positions are shown by the small horizontal arrow on the right ordinate of each graph) with the head restrained in the top row and head free in the bottom row for visual (left column) and auditory (right column) targets. Note that the prominent undershooting of the acoustic (upper right), but not the visual (upper left), target, which was present with the head fixed, was eliminated when the head was free to turn (lower right).
- Dynamic localization of sound To test the cat's ability to localize sounds in a more natural situation, we used two-step saccade trials to deliver the sounds while the head was moving either toward or away from the sound. A short 25 msec sound stimulus was delivered in the middle of a head/gaze movement of the cat to a previously delivered visual stimulus. The cat was required to make a second saccade to the sound, which occurred while the head was moving at 100-200 deg/sec. The moving head had not statistically significant effect on the ability of the cat to accurately localize the sounds, showing that the cats' auditory system is able to compensate for the head movement made during the presentation of the sound.
- Vestibulo-auricular reflex (VAR) Since we had coils on the pinnae of the cats while they worked with their heads free, we were able to study the movements of the pinnae on the head while the cat oriented its head to sound. Since the pinnae moved in a rapid, consistent and goal-directed manner to both visual and acoustic stimuli when the head was stable and since they are situated on the head, we hypothesized that there should be a reflex, analogous to the well-known vestibulo-ocular reflex, which would move the pinna to compensate for the head movement. The figure below shows movements of the head (black), left ear (blue), eyes in space (gaze, green) and ear-on-head (head-ear, red) as the cat localized an acoustic target at 6 different points in space. The locations of the targets are shown by the horizontal left-pointing arrow on the right ordinate. All of the traces are drawn relative to the time of head onset at time 0. Panel i) shows that the the pinna moves first with a quick saccadic-like movement to the left (to about -20 deg) followed by the movement to the left of both the eyes and the head. Note that during the leftward head movement the pinna remains approximately stable in space even though the head is turning to the left. This can only happen if the pinna is counter-rotating to the right (ear-on-head trace) to compensate for the head movement. This is the vestibulo-auricular reflex. Note that for stimuli in the right side, (panels iv to vi) the left ear movement is approximately like the head movement, in other words, it is simply riding on the head.
- Effects of eye and ear position on auditory receptive fields in the cat's superior colliculus The deep and intermediate layers of the SC receive multimodal input: visual, auditory and somatosensory. These inputs are arranged in a topographic fashion acrosss the surface of the SC and these maps are believed to be in alignment. Howeveer, since the coordinate system for the maps are different (retinotopic for vision and head centered for audition), they can be in alignment only when tthe eyes are pointing straight ahead. We have studied the effect of varying eye and ear position on responses in the SC. Preliminary results are in accord with the motor error hypothesis of Jay and Sparks: the auditory receptive fields are modulated by eye position and this modulation persists even when the ears are paralyzed.
- Bimodal interaction in the deep layers of the superior colliculus We have studied the bimodal interaction of visual and auditory stimuli in the deep SC of awake behaving cats. Surprisingly, we found that most SC cells showed occlusion when both visual and acoustic stimuli were delivered rather than the facilitation described in anesthetized cats.
- Psychophysics of sound localization in the cat
- Behavioral and neuronal correlates of the precedence effect in awake cats Studies of illusions have long been used since they can be instructive in informing us about the limitations of the sensory systems. We have done extensive behavioral and neural studies of the precedence effect. Cats show all of the properties of the PE that have been described in human subjects: they exhibit summing localization for delays between the stimuli that are short (in cats <~400 usec while in humans <~800 usec), and at very long delays the PE breaks down and the cats appear to hear both the leading and lagging sound as they often exhibit two saccades, while at intermediate delays the PE appears to be experienced. Using the awake, behaving preparation we have looked for neural correlates of the precedence effect in the inferior colliculus. Our results show that anesthesia does indeed bias the recovery time of most cells in the inferior colliculus and results in awake animals are closer to psychophysical results in human subjects.
- Most studies of the PE have examined stimuli in which the two speakers are separated along the horizontal meridian, or simulating such speakers with ITDs or ILDs. We have recently examined the presence of the PE for speakers in a variety of horizontal, vertical or diagonal placements in the head restrained, awake cat. We found little evidence for PE-like behavior for two speakers on the mid-sagittal plane but when the speakers were diagonal then evidence for both summing localization and localization dominance were seen to some degree.
- Behavioral studies of the Franssen effect We have also begun behavioral studies of the Franssen effect, another localization illusion that is effective for tones, which are difficult to localize. In this illusion, the subject hears two tones, one with an abrupt onset and slow decay while the other has a slow onset and stays on for several hundred msecs. Subjects will localize the sound at the location of the one with an abrupt onset even though it is no longer on when the judgement is made. The effect does not hold for noise stimuli, which are more easily localized.
- Effects of spectra Using broadband and narrowband noises as well as tones, we studied the effect of stimulus spectra on localization. In the horizontal plane localization was best for broadband sounds and gradually deteriorated for high pass, low pass, narrow band noise and tones. Localization in the vertical plane was generally worse than in the horizontal for all sounds and the cats were unable to localize tones in elevation. Results from localization of broad band noise and tones of 1, 6, 10, and 14 kHz are plotted. In A are the final gaze positions for the 8 targets varying in elevation (top row) and azimuth (second row). Different colors are gaze shifts to the different targets. Each data point is a different trial with all trials beginning with the cat fixating at the center. Targets are indicated by the open symbols, responses by closed symbols. Quantitative analysis of the data in A are shown in B in which the resulting gaze shift is plotted against the motor error, by which we mean the difference between the target and the cat's gaze position at the time the target was turned on. Shifts in elevation (B, top row) are plotted separately from shifts in azimuth (B, bottom row). The data points are fit by first order linear regression (red line). Gain is the slope of the regression line and represents localization accuracy (gain = 1 corresponds to perfect localization accuracy); d is the residual error after regression and is an indication of response precision or consistency; n is the number of trials.
- Effects of stimulus level and duration We found little effect of stimulus duration on localization performance with broadband noise stimuli. Except at near threshold levels localization in the azimuthal plane was also not affected by stimulus level but in the vertical plane we found a "negative level effect" for short, but not long duration sounds. Simulations using a peripheral nerve model suggest that the decrease in performance at high levels in the vertical plane are due to central mechanisms of multiple looks at spectral estimation.
- The negative level effect We studied the negative level effect (the deterioration of localization performance in elevation at high stimulus levels) using an array of stimuli with different temporal properties. In agreement with our previous study and those of others, we only found the negative level effect for click-like stimuli and only in elevation.
- Effects of forward masking on localization Most studies of forward masking have examined detectability. We studied the effect of a forward masking sound on the localization of the target sound with time gaps of 0 and 15 msec between the end of the broadband masker and onset of the broadband target. The figure shows the systematic improvement in localization as measured by the gain of the gaze shift vs motor error curve as the gap varies from 0 msec (FM0: stronger effect, poorer localization), to 15 msec (FM15) to the control with no masker for targets varying in the horizontal (blue) or vertical (red) plane.
This research was supported by NIH grants DC-02840, DC-00116, and DC-007177.