CHARACTERIZATION OF THE RESPONSES OF INFERIOR COLLICULUSNEURONS OF THE CHICKEN TO ELECTRICAL STIMULATION OFTHE COCHLEAR NERVEbyPETER RICHARD NEUFELDB.Sc. (Hon), Queen’s University at Kingston, 1991A THESIS SUBMiTTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESDepartment of PhysiologyWe accept this thesis as conformingto the required standardTHE UNWERSITY OF BRITISH COLUMBIAJuly 1994© Peter Richard Neufeld, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of Pkc.The University of British ColumbiaVancouver, CanadaDate AL 2’I J9L/DE-6 (2188)11ABSTRACTResponses of inferior colliculus neurons of the anaesthetized, cochlea-ectomizedchicken to electrical stimulation of the cochlear nerves were recorded extracellularly. At leastthree physiologically distinct cell types were found in the central nucleus of the inferiorcolliculus. Type 1 fired randomly and with a high spontaneous rate, exhibiting a Poissondistribution in the spike interval histogram. Stimulation of either cochlear nerve produced aninhibition lasting 3 to 46 ms (mean of 16.7, n=16). Type 2 exhibited little or no spontaneousactivity, and responded to a short stimulus with a single spike or burst of spikes (n=21). Type 3exhibited regular, spontaneous firing with preferred intrinsic frequencies in the audio range(n=26), usually resulting in multimodal spike interval histograms. Single pulse stimulation ofthe contralateral nerve reset the firing rhythm, resulting in periodic post-stimulus timehistograms (PSTH). The intermodal interval for a PSTH of a type 3 cell was identical to theintermodal interval for a spontaneous interval histogram. A reverse correlation of a randomsequence of stimulus pulses with the response spikes revealed preferred frequencies in the inputwhich were similar to the output frequencies seen in post-stimulus time and interval histograms.Thus, these type 3 neurons exhibited both an oscillatory spontaneous and evoked firing pattern,and an intrinsic frequency selectivity which is presumed to give rise to the observed oscillation.These cells were found to be grouped together in a relatively small portion of the inferiorcolliculus. The location of several type 3 cells was identified using WGA-HRP injected fromthe recording electrode. These cells were found to be in the core of the central nucleus of theinferior colliculus.These findings suggest the existence of an intrinsic mechanism for frequency filteringand time coding in the CNS.111TABLE OF CONTENTSABSTRACT.iiTABLE OF CONTENTS iiiLIST OF FIGURES vACKNOWLEDGMENTS viiNTRODUCTION 1I. Time Resolution of Hearing 1II. Auditory Signal Transduction and Temporal Coding 1A. Signal transduction and the auditory filters 1B. Limitations of spatial discrimination of frequency 2C. Phase locking and temporal discrimination 3D. Limitations of temporal discrimination 4E. Summary 5III. Signal Processing in the CNS 5A. Auditory neuroanatomy and the time pathway 5B. Directional hearing 7C. The time resolution of directional hearing 9IV. Summary and Hypothesis 9V. Neuronal oscillations and resonance 11VI. Experimental Rationale 13A. The choice of an in vivo approach to the ICC 13B. The role of the inferior colliculus 13C. Animal protocol 14METHODS 16I. Animals 16II. Surgical preparation 16III. Recording electrodes 17ivIV. Stimulus generation.17V. Cell isolation and spike discrimination 18VI. Data Acquisition 18VII. Estimation of inherent frequency 20VIII. Histological verification of recording sites 20RESULTS 22DISCUSSION 37I. Comparisons with Previous Studies 37II. Type 1 cells 38A. Spontaneous activity 38B. Response to stimuli 38III. Type 2 cells 39A. Response to stimuli 39B. Interaural time sensitivity 40C. Oscillatory activity 41IV. Type 3 cells 43A. Spontaneous oscillations 43B. Oscillations in the impulse response 44C. Response to a random pulse sequence 45D. Frequency preferences 45V. Conclusions 47REFERENCES 49VLIST OF FIGURESFigure 1: Spontaneous interspike interval histogram and temporal response histogramcharacterizing ‘type 1’ cells 23Figure 2: Temporal response histogram characterizing ‘type 2’ cells 25Figure 3: Graphs of interaural time delay sensitivity for three ‘type 2’ cells 26Figure 4: Spontaneous interspike interval histograms characterizing ‘type 3’ cells 27Figure 5: Distribution of intrinsic frequencies of ‘type 3’ cells 30Figure 6: Temporal response histogram characterizing ‘type 3’ cells 31Figure 7: Spontaneous interspike interval histogram, temporal response histogram andreverse correlation histogram of a single ‘type 3’ cell 33Figure 8: Photograph of sagittal section of optic tectum showing horseradishperoxidase staining at recording sites 35Figure 9: Photograph of sagittal section of optic tectum showing horseradishperoxidase staining at recording sites 36viACKNOWLEDGMENTSMany thanks are due to my supervisor, Dr. Schwarz, for making this work possible withhis support and encouragement, not only of my scientific efforts, but of my interests in musicand sailing. All graduate students should be so lucky...Thanks also to Eleanor To for providing assistance with the histology, and Dr. Finlaysonfor leniency during the final months of writing when I should have been working for him.To Dr. Mathers, thank-you for chairing my advisory and examination committees withan experienced hand. Thanks also to Drs. Puil, Church and Vincent for their advice andcriticisms during the evolution of this project into its fmal form.I would also like to thank the entire physiology department for making my time at UBCenjoyable and, some say, educational. From a memorable trip to the chocolate buffet in my firstweek to the pot lucks this summer, with all of the relays, longboat races and wall stormings inbetween, the graduate students and faculty have made a newcomer feel at home. Joe Tayworked miracles in the darkroom in turning my hastily prepared figures into beautifulslides.. .“No, I don’t need them yesterday. Anytime before noon today would be fine, Joe.”Thanks to Dr. Pederson for always being available with information or a calm word of advice,and for hosting the annual retreat.These pages record such a small part of my experiences in this department...1INTRODUCTIONI. Time Resolution of Hearing:The auditory system is capable of incredibly fine time resolution. Seemingly simpletasks such as listening to music or turning towards the source of a sound require discriminationof events in the cochlea and the CNS to within 10 J.ts or less. The ear itself has not been shownto be capable of this resolution, so it must lie with some as yet undiscovered property of theauditory pathway.II. Auditory Signal Transduction and Temporal Coding:A. Signal transduction and the auditory filters:Sounds which arrive at the ear cause displacements of the tympanic membrane(eardrum) which are translated via the middle ear and oval window into displacements of fluid(perilymph) in the cochlea. Movement of this fluid in turn causes displacement of the basilarmembrane, on which the sensory hair cells are fixed. Thus, a sound wave whose displacementmay be expressed as a complex of sinusoids causes a displacement of the basilar membranewith a very similar complex sinusoidal pattern. This movement of the basilar membrane movesthe hair cells relative to their sensory cilia, which are fixed at the tectorial membrane. Thismovement causes receptor potentials in the hair cells, leading to synaptic generation of one ormore action potentials in the auditory nerve afferents (cf. Moore, 1989).The basilar membrane does not simply mimic the vibrations of the tympanum. At thebase of the cochlea (closest to the oval window), the membrane is stiff, and vibrates more easilyat higher frequencies. At the apex of the cochlea (furthest from the oval window), it is lessrigid, and vibrates more easily at lower frequencies. As a result, a sinusoidal stimulationarriving at the oval window travels up the basilar membrane from the base towards the apex,causing a maximum displacement of the membrane at the region “tuned” to the frequency of thesine wave. Different points on the membrane along the path of the traveling wave will alsooscillate at the frequency of the stimulus, but with a smaller amplitude.2This “tuning” of the basilar membrane results in filtering of sound signals. A complexsignal, composed of many sine waves of different frequencies and phases (typical of mostnatural sounds), is separated into its component frequencies by virtue of the tuning properties ofthe basilar membrane. A given region of the membrane vibrates maximally only if there existsa frequency component of reasonable amplitude near the preferred frequency of that region(termed the characteristic frequency, or CF). Just what is meant by ‘near’ depends on severalfactors. First of all, as mentioned previously, any sine wave has the capacity to excite the entiremembrane if it is intense enough. Secondly, CF appears to be represented on the membrane asincreasing logarithmically with distance towards the base, so that the high frequency regionsnear the basal end appear compressed relative to the low frequency regions at the apex. Theterm ‘critical band’ has been developed to describe the frequency resolution around a givencentre frequency. The critical bandwidth increases linearly on a log-log plot with increasing CFand is roughly 160 Hz for a CF of 1000 Hz. Each point on the basilar membrane can be thoughtof as a filter with a centre frequency determined by its position and a bandwidth equal to thecritical band. These filters have been termed the ‘auditory filters’ (cf. Moore, 1989). Note thatthe critical bandwidth and CF both increase logarithmically with distance along the basilarmembrane, so that the critical band can be expressed as a fixed distance on the membrane. Thiswidth has been measured in humans and found to be about 0.9 mm (cf. Moore, 1989).B. Limitations of spatial discrimination of frequency:The auditory filters give rise to a spatial resolution along the basilar membrane.Different frequency components of a complex sound excite different regions of the membrane,which in turn lead to different hair cells passing their excitation to auditory nerve fibres. Theresolution is limited by the width of the auditory filters, which at high frequencies can be verylarge. Thus the high frequency overtones of a musical note can occupy the same critical bandand be difficult or impossible to resolve, and two simultaneous tones lying within a critical bandwill be heard with only one pitch. This has been supported psychophysically (cf. Moore, 1989).3The low frequency components of most musical pitches generally fall into separatebands and therefore should be easily discriminable. What has been found, though, is that forlow frequencies (around 1 kHz), humans are able to distinguish differences in the frequencies oftwo tones of as little as 2 Hz. Assuming that frequencies are perceived as different when theyexcite different critical bands, such small frequency changes should be imperceptible. Thedifference in the basilar membrane excitation patterns between a 1000 Hz tone and a 1002 Hztone cannot account for the observed resolution.An additional discrepancy arises when attempting to describe the perception of complextones in terms of spatial resolution. The perceived pitch of a complex tone does not necessarilycorrespond to the position of maximum displacement of the basilar membrane. As an example,a complex tone consisting of all of the harmonics of a musical note but lacking the fundamentalfrequency will be perceived as having a pitch equal to the missing fundamental.C. Phase locking and temporal discrimination:There exists another mechanism in the peripheral auditory system which is thought tocontribute to frequency discrimination, particularly at the low frequency end. In contrast to thespatial discrimination arising from auditory filters, we will call this second mechanism temporaldiscrimination.As was stated in section II.A., each region of the basilar membrane has a particular CF.Similarly, nerve fibres possess CFs by virtue of their connections to individual hair cells. Agiven nerve fibre will exhibit an increase in average firing rate in response to any tone at afrequency within that fibre’s response area, and a maximal firing rate when stimulated at the CFof the unit. The rate of firing is not the only parameter of nerve firing which is dependent uponfrequency. Temporal discrimination rests on the ability of nerve fibres to encode frequencyinformation in the timing of spikes. In response to a pure tone, nerve firings tend to besynchronized to the phase of the stimulus waveform. This phenomenon, which is critical totemporal discrimination, is called phase locking. A phase locked fibre will not necessarily fire4at every cycle of the stimulus, but when it does fire, spikes occur at (roughly) the same phase ofthe stimulus waveform each time.Consider the effect of phase locking on the interspike intervals of a nerve fibre whenstimulated by a 1 kHz tone. The minimum interspike interval would be 1 ms, corresponding toone period of the stimulus waveform. In addition, there would be interspike intervals occurringat multiples of this period (2 ms, 3 ms etc.). Thus, the period of the stimulating waveform iscarried unambiguously in the firing pattern of one neuron over time, and a population ofneurons taken together would produce spikes at every cycle of the stimulus. Phase locking canalso explain the perception of a pitch at a frequency not present as a sinusoidal component in acomplex tone (e.g. the phenomenon of the missing fundamental). For example, an amplitudemodulated tone with a carrier frequency of 1000 Hz and a modulation frequency of 200 Hzconsists of three components: 800, 1000 and 1200 Hz. It is found that a portion of the excitedneurons phase lock to the overall repetition rate of the stimulus (200 Hz), which is equal to the‘missing fundamental’ of these three frequency components. This phenomenon is thought togive rise to the perception of a tone at that frequency.D. Limitations of temporal discrimination:The highest frequency to which a population of fibres can phase lock depends upon anumber of factors and varies considerably between species. Phase locking in owls has beenmeasured for tones up to 8-9 kHz (Sullivan and Konishi, 1984), whereas in some laboratorymammals the high-frequency limit is 3-4 kHz (Rose, et al., 1974).Temporal discrimination would seem to be able to provide the necessary resolution atthe low frequency end of the spectrum (i.e. where spatial resolution is insufficient). Onedifficulty which is commonly expressed is that there is no evidence for a physiologicalmechanism which is able to carry out the interspike interval measurements with sufficientaccuracy (cf. Moore, 1989). For the discrimination of a 1000 Hz tone from a 1002 Hz tone, theinterspike interval of 1 ms must be measured with an accuracy of 2 ts.5E. Summary:The auditory filters produce a spatial code of frequency and a means of separatingcomponents of a complex tone. This spatial code is insufficient to account for the observedresolution of the auditory system, particularly at low frequencies. Phase locking produces atemporal code of frequencies with high resolution. It extends up to 3-5 kHz for mammals, and 8-9 kHz for owls, which use auditory cues to locate prey. No physiological mechanism has yetbeen found which is able to make use of the phase locked information. Such a mechanism mustbe located in the CNS, and the following section reviews the anatomy and physiology of the CNSwith a focus towards temporal coding in avian species. The choice of birds as the focus of thisstudy is due largely to the anatomical simplicity of the avian nucleus laminaris as compared withits mammalian analogue, the medial superior olivary nucleus, and is explained below in detail.III. Signal Processing in the CNS:A. Auditory neuroanatomy and the time pathway:The auditory pathway has been mapped out in considerable detail in a number of species.One of the principal pathways from the periphery to the auditory cortex in birds is as follows:First order auditory nerve fibres synapse on the ipsilateral nucleus magnocellularis (NM); secondorder neurons send collaterals to both nuclei laminaris (NL); third order neurons synapseprincipally contralaterally (with a small ipsilateral projection) on the central nucleus of the inferiorcolliculus (ICC); fourth order neurons synapse on the ipsilateral nucleus ovoidalis of thethalamus; the thalamus sends projections to the auditory cortex. For reviews of these projections,see Yin and Chan (1988), Konishi, et al. (1988), and Conlee and Parks (1986). The ascendingconnections reported above are known in some detail. Less is known about descendingconnections, although some of the efferent projections have been mapped in the chicken: The ICprojects to the caudal pontine reticular formation, which in turn projects down to the cochlea(Schwarz, et al., 1992).The structures described above are each presumed to subserve a different function, andinvestigations at the cellular level may provide clues to those functions. The nucleus6magnocellularis is made up predominately of ovoid cells with very short dendrites (Jhaveri andMorest, 1982), receiving large calycine endings of cochlear nerve fibres on their somata (Konishi,et al., 1988). The nucleus laminaris is a monolayer of fusiform cells (expanded into severallayers in the owl), with dorsal and ventral dendritic arborizations, receiving excitatory inputs fromipsilateral and contralateral magnocellularis on the dorsal and ventral sides, respectively (Pallottaand Peres, 1989, Parks, et al., 1983, Smith and Rubel, 1979). As mentioned previously, it is thestructure of the NL which makes it ideal for study, particularly when compared with the structureof its mammalian counterpart, the medial superior olivary nucleus (MSO). The MSO is a complexof several cell types, one of which, the principal cell, extends dendritic arborizations medially andlaterally (Tsuchitani and Boudreau, 1964). The inferior colliculus is divided into several parts, acentral (ICC), external (ICX) and shell (ICS) nucleus (Knudsen, 1983). The ICC of chickenscontains, among other cell types, large GABAergic neurons (Granda and Crossland, 1989), someof which may project up to the nucleus ovoidalis (Muller, 1988). Glycinergic projections havebeen traced retrogradely from the ICC of chickens to three lower nuclei on the ipsilateral side(Schwarz, et al., 1994).The pathway outlined above is often called the time pathway because it preserves andprocesses temporal information as phase locked firing. This is to distinguish it from an alternativeanatomical route termed the intensity pathway. The nuclei and interconnections making up theintensity pathway will not be detailed in this review, except to say that neurons in these structuresgenerally show sensitivity to sound intensity, and phase locked firing is not observed. As aresult, nuclei of the intensity pathway are not suitable subjects for a study of the fine temporalresolution referred to in section II.The time pathway has typically been studied for its relationship to directional hearing, aphenomenon which, like frequency discrimination, requires very fine resolution. A review of thefunctioning of this system must therefore start with an overview of directional hearing, but it mustbe remembered that these two tasks, directional hearing and frequency discrimination, areintimately related by the cues upon which they operate (timing and distribution of actionpotentials) and the structures which carry out the processing (the nuclei of the time pathway).7B. Directional hearing:Directional hearing, or sound source localization, is a task which humans carry outreasonably well without effort, and which is developed exquisitely in specialized animals such asthe owl and bat. As a result of extensive psychacoustic studies, the directional cues which areavailable to the ear are well characterized. Most everyday sounds possess a very complexfrequency spectrum, and a number of directional cues are available. For pure tones there are onlytwo cues for the definition of the sound incidence angle in the horizontal plane (or ‘azimuthalangle’), but these are sufficient to allow localization (Blauert, 1983). They are the interauralintensity difference (11D) and interaural time or phase difference (lTD or IPD).An lID is the difference in the intensity (i.e. volume) of a sound between the two ears, andis caused by the acoustic shadowing of the head. A sound located to one side of the head will beless intense at the opposite ear providing its wavelength is less than ca. ten times the diameter ofthe head (Michelsen, 1992). Intensity differences are thought to be processed and compared bythe intensity pathway referred to in section II.A. (Sullivan and Konishi, 1984, Takahashi, et al.,1984), and contribute to localization in the azimuthal plane for high frequencies (i.e. shortwavelengths) and, in barn owls with asymmetrically located ears, the vertical plane (Volman andKonishi, 1989). It should be noted that, just as the auditory filters can discriminate frequenciesexceeding the phase locking range, liDs allow azimuthal localization of high frequency tones.An lTD usually refers to the ongoing difference in timing of a sound between the twoears, and is caused by a difference in path length from the source to each ear. It is expressed inunits of time (usually ,is). A difference in timing of a sinusoid produces a difference in phase,therefore the term IPD (interaural phase difference) may be used for steady state tones in place oflTD. which is sometimes reserved for differences in the time of arrival of a sound between thetwo ears. An IPD is expressed in units of radians, but can be converted to the equivalent units oftime if the frequency is known. For our purposes, lTD will be used to indicate the ongoing timedifference.The value of the lTD depends on the spacing between the two ears and the azimuthal angleof the sound source. For example, a sound source located at an azimuth angle of 90° to the right8of a human (head diameter of 23 cm) would reach the left ear about 700 ts later than the right (thetime it takes sound at 340 mIs to travel around a sphere with a diameter of 23 cm). Phase lockingof auditory nerve fibres will produce impulses in the auditory nerves with the same timediscrepancy. For sounds whose half-wavelength is equal to or less than the head diameter, thephase difference would be greater than 180°, resulting in ambiguity as to which ear was closer tothe sound. In practise, ilDs or head movements could resolve the ambiguity. For higherfrequency sounds, the wave length of the sound becomes less than the path difference betweenthe two ears, so that the same phase difference could be produced by a number of different sourcelocations. Using the human as an example, the IPD is an unambiguous cue at frequencies belowabout 1500 Hz. For animals with smaller heads, the maximum path difference is smaller, so thehighest frequency which provides unambiguous cues increases, as does the lowest frequencywhich provides useful lID cues. The largest lTD which a chicken experiences has been measuredto be about 300 ts, so that phase ambiguity should arise for frequencies above 3.4 kHz. In fact,lTD cues cease to become detectable above 2 kHz, as this is the highest frequency for whichphase locking has been observed in the chicken (Warchol and Dallos, 1989).The manner in which the CNS makes use of lTD cues has been investigated in detail, andthe following conclusions can be drawn. The lowest point in the auditory pathway whichreceives binaural information (i.e. convergence of fibres from both ears) is the NL, whichreceives second order projections from the NM (see section III.A.). Information in the form ofphase locked spikes arrives on the dorsal surface of the NL from the ipsilateral NM, and on theventral surface from the contralateral NM (see section Ill.A.). The convergence of many phaselocked inputs results in a conspicuous field potential which has been termed the neurophonicpotential. This potential reproduces faithfully the waveform of the sound stimulus (Schwarz,1992). To what extent this neurophonic potential is generated by phase-locking remains unclear,as does the extent to which this field potential actually influences the output of the NL. Thesignificance of the neurophonic for the present study is that it interferes with single unit recordingin the NL, effectively excluding this nucleus from single unit extracellular studies.9Anatomical and physiological evidence has shown a delay line along the NL of chickensand owls, which is thought to compensate for an lTD at one ear by delaying the impulses from theother ear (Schwarz, 1992, Sullivan and Konishi, 1984, Takahashi, et al., 1984, Smith andRubel, 1979, Smith, 1981, Carr, 1988). The magnitude of the lTD. in conjunction with the delaylines, determines which cells in the NL receive coincident inputs. The physiological studies citedabove have demonstrated that the NL is capable of resolving ITDs in the physiologically relevantrange.The ability of these cells to detect coincidence with high resolution is central to theoperation of this mechanism.C. The time resolution of directional hearing:The localization of a 1000 Hz tone requires that the NL neurons be able to resolve adifference in the arrival time of spikes of much less than 1 ms (the period of the tone). This isremarkable given that the input to the coincidence detector is in the form of action potentials thathave durations of the order of hundreds of microseconds (Irvine, 1986), or EPSPs with durationsof several milliseconds. Expressed another way, given that the longest lTD that a humanexperiences is 700 ps, and an EPSP lasts more than 1 ms, the entire array of neurons willexperience simultaneous EPSPs for every possible lTD. Thus, it is unclear just how timeresolution on the order of a few microseconds is possible.IV. Summary and Hypothesis:The separation of a complex tone into frequency components which occurs in theperception of pitch, and the discrimination of two tones of very similar frequency, can beaccounted for by a central processor capable of accurately measuring interspike intervals of phaselocked neurons. The localization of a sound source in azimuth can be accounted for by acoincidence detection mechanism capable of detecting the relative arrival times of phase lockedspike trains from each ear. These two tasks are not independent of one another. In fact, theyappear to converge within the NL for the perception of a signal in noise. Consider a complex10sound such as speech, and a background noise, both containing many of the same frequencies.Providing that they do not originate from the same source, the various frequency components willhave different ITDs. Those arising from the speech sounds will all have the same ITD, and thosefrom the noise will share a different lTD. The discrimination of frequency components, and thesorting according to ITDs, enables the auditory system to perceive speech sounds which wouldotherwise be obscured by noise.The time resolution displayed by these two processes (on the order of jis) cannot beexplained in terms of traditional neuronal processes (i.e. action potentials and chemical synapseslasting ms) given what is known about the auditory pathway. We propose that an additionalmechanism, neuronal frequency selectivity or resonance, operates within this system, particularlyin the NL due to the key role this nucleus plays in temporal processing (see section III). Themembrane potential of neurons receiving phase locked information resonates electrically atauditory frequencies. This resonance would act as a filter, separating out a single frequency froman auditory nerve fibre phase locked to more than one frequency, and increasing the resolution ofthe coicidence detector (NL) to less than the period of resonance of the cell.As an example, consider a cell with an electrical membrane resonance at 1000 Hzreceiving phase locked input. A depolarizing deflection caused by synaptic input is more likely toexceed spike threshold if it occurs in phase with the cell’s membrane potential deflections arisingfrom resonant properties. A deflection occurring out of phase with the cell’s evoked orspontaneous oscillation is less likely to result in spike firing, and an optimal response wouldrequire a match between intrinsic and input frequencies. Convergence of many phase lockedinputs could further enhance the frequency resolution by averaging the signals tuned to the samefrequency in many fibres.One of the predictions of this hypothesis is that the intrinsic membrane properties of cellsof the NL are tuned to different frequencies, such that the nucleus as a whole spans the entirerelevant frequency- and lTD-sensitive range. Intrinsic frequency preferences could be evident asresonance in the neurons’ response to an input pulse, and possibly also in the spontaneousactivity of the neurons as oscillations in the firing pattern.11Evidence for the existence of resonant properties of neurons has been provided bynumerous studies, some of which are outlined below.V. Neuronal oscillations and resonance:The existence of intrinsic resonant properties has been established for several neuronpopulations. These resonances, when studied extracellularly, are expressed as inherentoscillations in the firing pattern. Some cells are said to be “pacemakers” when they spread theirrhythm to large areas of the CNS (Metherate, et al., 1992, Steriade, et al., 1985, Steriade andLlinás, 1988, Pinault and Deschênes, 1992, Nuñez, et al., 1992, Nuñez, et al., 1992). Inaddition to these widespread oscillations, visible in the EEG as, for example, delta, spindle or 40Hz waves, a number of other systems possess inherent oscillations due to underlying resonancesor particular synaptic organizations. Neurons in the vestibular system of mammals (Serafm, etal., 1992), cochlear hair cells of reptiles (Fuchs and Evans, 1988), avian auditory system(Langner, 1983, Fuchs, et al., 1988)(and see section VI.C.), mammalian auditory system(Eggermont, 1992) and non-excitable cells (Pallotta and Peres, 1989), display frequencypreferences in their firing patterns and/or input-output relationships.In recent years, much work has focused on these oscillations in the thalamus,thalamocortical connections and the cortex of mammals. Intracellular recordings from theseneurons have demonstrated the involvement of low-threshold calcium spikes (LTSs) and ahyperpolarization-activated cation current, termed ih , in the generation of delta oscillations(McCormick and Pape, 1990, Llinás, et al., 1991, Gutnick and Yarom, 1989). Together, thesecurrents result in an after-hyperpolarization followed by a rebound depolarization which, if strongenough to generate another LTS, leads to oscillations (Llinás, 1988). In the case of spindles, acalcium-dependent potassium conductance may interact with a persistent sodium conductance andIPSPs fed back to thalamic relay cells from the reticular nucleus (Llinás, 1988). Oscillations inthe auditory system are not well characterized. The present study represents the first detailedextracellular evidence for oscillations in the auditory system in the absence of cochlear effects.Some of these data were previously published (Schwarz, et al., 1993).12Neuronal oscillations may be independent of any inherent resonance of the cell inquestion. If a cell receives oscillatory synaptic input, it may exhibit oscillatory firing. Thefrequency in this case could depend upon a number of factors, including the number ofpresynaptic cells and their respective frequencies. In addition, intrinsic oscillations may or maynot exceed threshold, so that the absence of oscillations in extracellular records does not precludetheir existence.Similarly, inherent resonant properties may be present in a given cell with a stablemembrane potential at rest (i.e. no oscillations visible either intra- or extra-cellularly). Bydefinition though, if the cell resonates, the membrane potential will oscillate at the cell’s‘preferred’ frequency when perturbed in either direction by a stimulus. These oscillations may ormay not exceed threshold, so that an oscillatory firing pattern is not necessarily detected. Thepreferred frequency of a resonant cell can change if the relative ion concentrations across themembrane or the membrane potential change, or in response to any substance (such as aneuromodulator) which alters ion channel properties.The technique employed in any study of neuronal resonance must take account of theinformation sought. An in vitro approach provides a convenient framework for intracellular orwhole cell patch recordings, and a means of applying drugs both extra- and intra-cellularly. Suchan approach is essential for an understanding of the mechanisms involved in resonance (i.e. thecurrents, second messengers, etc.). Unfortunately, network connections are lost, and questionsof function cannot be easily addressed. Due to some unknown properties of the tissue, the NL ofthe chicken appears not suitable for an in vitro slice, whole cell patch approach (unpublishedobservations). An in vivo approach preserves network connections, and therefore permits thestudy of suprathreshold oscillations in resonating neurons or at the target(s) of their output.The frequency and lTD selectivity reviewed in section III.B. suggests that the mostreasonable location at which to find neuronal resonance is the NL. Unfortunately, for the reasonsoutlined above and in section Ill.B., this nucleus is not a suitable subject of study. Assuming thatthe oscillatory firing patterns of NL neurons are preserved at the next level in the pathway, thecore of the central nucleus of the inferior colliculus (ICCc), this nucleus would be the logical13choice for investigation. In addition, cells of the ICC may themselves possess resonantproperties. A brief review of previous studies of the avian ICC is presented below.VI. Experimental Rationale:A. The choice of an in vivo approach to the ICC:We have established that the auditory system exhibits a time resolution which is not easilyexplained by known neuronal mechanisms. Studies of the anatomy and physiology of theauditory pathway point to the NL and its target, the ICC, as being most likely instrumental in thefine temporal discrimination, but the NL itself is not an ideal structure for single unit recording.The hypothesis under investigation is that neurons in the central auditory pathway, particularly inthose regions involved in temporal processing, possess inherent resonances in the auditory range.This property would allow such cells to act as filters for frequency analysis of sound signals,thereby assisting in the combination of the filtered output from both cochleae. This might equipthese cells with a better time resolution for coincidence detection than they would possess in theabsence of resonance. Due to the need to study properties of the ICC which potentially arise fromthe NL, an in vivo approach, which preserves the interconnections, is necessary.B. The role of the inferior colliculus:Functionally and anatomically distinct regions of the avian IC (also called the nucleusmesencephalicus lateralis pars dorsalis or MLD) have been demonstrated (Knudsen and Konishi,1978, Dezsö, et al., 1993). Since then, these regions have been better characterized, and some oftheir connections mapped. The cells of the central nucleus (ICC) are arranged tonotopically in theowl and are sensitive to interaural time differences in sound stimuli (Wagner, et al., 1987). TheICC receives inputs primarily from the contralateral NL and from neurons in the intensity pathway(Conlee and Parks, 1986, for the chicken), and has been shown in owls to be sensitive to lTD(Wagner, et al., 1987, Takahashi and Konishi, 1988, Takahashi and Konishi, 1983).Commissural projections from the core of the ICC (ICCc) to the opposite lateral shell of the ICC(ICCs) are thought to assist in the representation of contralateral space by the lateral ICCs in owls14(Takahashi, et al., 1989). The remainder of the ICCs is not sensitive to interaural time and doesnot contain a representation of azimuth (Takahashi and Konishi, 1983). Chickens also possessconmuissural projections between the ICC on each side (Conlee and Parks, 1986), but studiescomplementary to those in the owl have not been carried out. The external nucleus (ICX) is aregion overlying the ICC laterally and anteriorly, and individual cells in the owl’s ICX receiveinputs from an array of cells in the ICC representing the entire frequency spectrum but only asingle lTD (Konisbi, et al., 1988). The result is a topographic arrangement of space-specificneurons in the ICX forming a map of auditory space (Knudsen and Konishi, 1978, Knudsen andKonisbi, 1978). It is important from the perspective of the present study to note that these space-specific neurons derive their lTD sensitivity from ongoing time disparities (IPDs), and areinsensitive to transient time disparities (ITDs) (Moiseff and Konishi, 1981).None of the studies outlined in the preceding paragraph preclude the existence ofoscillations in the firing patterns of IC neurons. In addition, these studies support the positionthat the ICC is the principle target of the NL and that it may preserve the fine time sensitivity seenin the NL.The present study represents the first characterization of the time patterns of chicken ICCneuronal responses to electrical pulse stimuli of the cochlear nerves.C. Animal protocol:The reason for the choice of an avian subject for this study relates to the differencebetween the NL and the MSO of mammals (see sections II.E. and III.A.), and the chicken is aneconomical choice. The in vivo approach to the ICC is rationalized in section VI.A.. It wasdecided as well that an anaesthetized preparation was more humane than an alert restrainedpreparation, particularly considering that the vestibular system was damaged during thepreparation, and a decerebrate preparation is not possible for recordings from the IC (located inthe midbrain). Studies carried out on anaesthetized and unanaesthetized animals are compared inthe discussion.15Electrical stimuli of the cochlear nerves (see methods) were used instead of acousticstimuli for several reasons. Firstly, removal of the cochleae completely eliminates unwantedacoustic background noise, including physiological noise (see Moore, 1989 for a discussion ofphysiological noise). Secondly, stimulating the cochlear nerves with brief voltage pulses allowsfor precise control of the timing of stimuli. Thirdly, spontaneously active, rhythmic firing hasbeen documented in avian cochlear nerve fibres and ascribed to rhythmical electrical potentials(oscillations) of the hair cells which they innervate (cf. Manley and Gleich, 1992). In order toeliminate oscillations of cochlear origin and limit the input to single, precisely timed spikes,experiments addressing intrinsic neuronal frequency selectivity in the auditory pathway must becarried out on animals without cochleae.16METHODSI. Animals:Female domestic chickens (Gallus domesticus) were used in all experiments. They wereobtained as adults from a local supplier and maintained in animal care facilities for up to 2 weeksuntil needed. In general, the animals were not fasted before experiments. Animals ranged in sizefrom 1.1 to 2.0 kg. A total of 23 animals were used for this study.II. Surgical preparation:Animals were anaesthetized with a single intra-muscular injection of a xylazinelketaminemixture; 7.3 mg/kg xylazine (Rompun) and 57.2 mg/kg ketamine (Ketalean). Anaesthesia wasmaintained throughout all experiments with intra-muscular injections of 3.5 mg xylazine/25 mgketamine approximately every two hours as required. Cloacal temperature was monitoredperiodically and maintained between 40 and 41.5°C using a heating pad. In a few cases, when theairway was obstructed by vomitus or saliva, the trachea was cannulated but the animals were notartificially ventilated.Once anaesthetized, the chicken was placed in a stereotaxic head holder. The head washeld between two ear bars and a bit was placed at the joint of the beak. The head was then tiltedforward such that the upper mandible was inclined downwards at a 45° angle (Kuenzel andMasson, 1988). The skin was removed from the top of the skull and the back of the neck over anarea bordered by the comb, the ears and the foramen magnum. All muscle was removed from thebone above the ears and on the neck. All of the trabecular bone posterior to the forebrain andlateral to the sagittal sinus was removed without exposing the dura. The bone overlying theampulla of the lateral semicircular canal was lifted off and the membranous labyrinth wasremoved, including the cochlea and the entire contents of the lateral and anterior semicircularcanals and vestibule. The resulting hole into the cochlea was enlarged sufficiently to provide aclear view of the auditory nerve. This procedure was carried out for both cochleae.A pair of stimulus electrodes was placed against the cochlear nerve near the internalacoustic canal. The stimulus electrodes were prepared in the following manner: silver wire17(99.99%, Medwire Corp.) of 76 jim diameter, coated in Teflon (total diameter 114 jim) wasstripped of insulation for the terminal 35 to 70 jim and coated in chloride by dipping in 2M NaClwhile applying 10 V anodic DC current (Hewlett-Packard Harrison 6200B DC power supply).The chloride coating was inspected under 40x magnification to insure an even and completecovering. Once in place on the nerve, the wires were cemented against the skull using dentalacrylic cement at a point approximately 0.5 to 1 cm from the nerve. The condensation wasremoved from the wires using a gauze wick, and the region overlying the cochlea filled in withpetroleum jelly from a syringe to stabilize the electrodes and prevent moisture build-up around thetips.Once both ears had been prepared in this way, the bone overlying the tectum of one sidewas removed, and the dura incised to expose a roughly rectangular region of tissue approximately2 mm square. The recording microelectrode was then inserted into the tectum using a three-dimensional mechanical micro-manipulator supporting a Trent Wells Mark ifi hydraulicmicrodrive.III. Recording electrodes:Recording electrodes were glass micropipettes pulled from 3 mm diameter glass capillarytubes (1.5 mm inside diameter) using a Narishige vertical one-stage electrode puller. The lengthof the shaft from shoulder to tip was about 10 mm. The electrode was then broken to an outsidediameter of between 3 and 6 jim, with an impedance of 0.8 to 1.2 M2 to a 1 kHz square wave.Electrodes were filled with the recording solution consisting of 2M NaC1. For horseradishperoxidase (HRP) labeling of recording sites (see below), the solution consisted of 0.5 mg ofprotein dissolved in 50-60 jiL of either 1.0 M NaC1 or normal saline.IV. Stimulus generation:A NeuroLog system was used for stimulus generation. TTL pulses were used to triggerthe pulse buffers, digital oscilloscope (Nicolet) and the recording scope (EGAA computerscope,RC electronics). Two pulse buffers allowed for independent control of stimulus delay and18amplitude for the right and left ear electrodes. These pulses were then fed into NeuroLogstimulus isolation units, which generated short (0.1 ms) electrical pulses of 1-200 pA. Pulseswere presented at a rate of 2/s or as a random sequence. The random sequence was generated bya noise generator (Coulbourn Instruments S81-02), the output of which was low-pass filtered at 1kHz (Krohn-Hite model 3343 filter) and fed into a NeuroLog spike trigger unit before passinginto the pulse buffers. The output pulses to the auditory nerves were presented with a Poissondistribution possessing a modal interval of 2 to 3 ms. For interaural time series experiments (seebelow), the stimulus to one or the other ear was delayed at the pulse buffer with a resolution of 10ps between 20 ps and 3 ms. Delay times were monitored with the aid of the oscilloscope.V. Cell isolation and spike discrimination:The electrode was oriented on the tectal surface according to anatomical cues. In general,the target was as far medial and rostral as possible while avoiding sinuses and blood vessels.Once on the surface of the tectum, the electrode was advanced approximately 1 mm beforebeginning recording. The ICC was identified physiologically as a region with a visible fieldpotential evoked by electrical stimulation of the auditory nerve. Individual cells were isolated byforward and backward movement using the microdrive. The spike discrimination was carried outby the EGAA software using an adjustable voltage threshold. A recording was considered to bedue to a single unit when all of the spikes crossing the threshold were of similar amplitude(±10%) and duration and the interval histogram showed no interspike intervals less than 2 ms.VI. Data Acquisition:IC neurons were recorded with glass pipette microelectrodes as described above. Theelectrical activity was amplified and filtered in one of 2 ways: preamplified lOOx (BAK electronicspreamplifier), band-pass filtered between 100 Hz and 10 kHz (NeuroLog filter) and amplified 50x(NeuroLog amplifier) or; preamplified lOOx and band-pass filtered between 300 Hz and 10 kHz(A-M Systems microelectrode AC amplifier model 1800) and amplified 50x (NeuroLogamplifier). The spikes in the resultant signal were transformed into TTL pulses and sampled with19a 20 jis resolution (EGAA computerscope software on a 386 clone). The data were collected andstored as one of the following types of histograms:1. Interval histogram (111): except where noted, this histogram is constructed without anystimulus to the cochlear nerves. The time interval between consecutive spikes is plotted on the Xaxis, and the number of such intervals recorded is plotted on the Y-axis. This histogram wasused to characterize the intrinsic periodicity (if present) in the spontaneous activity of the neuron.The sampling window was 160 ms long, allowing detection of interspike intervals of 160 ms(corresponding to a frequency of 6.25 Hz)2. Post-stimulus time histogram (PSTH): this histogram is constructed in the presence of regularstimuli at one or both nerves. The X-axis is time following the trigger. The stimulus falls atapproximately 2 ms on the X-axis, and every spike occurring over the following 350 ms is plottedon the time axis. The number of spikes occurring at a particular post-stimulus time is plotted onthe Y-axis. In general, PSTHs were generated from up to 100 stimuli. This histogram was usedto characterize ability of a stimulus pulse to entrain the activity of the neuron.3. Reverse correlation histogram (RCH): this histogram is constructed in the presence of arandom train of pulses delivered to one cochlear nerve. The Y-axis is the number of stimulioccurring at a particular time. The X-axis is the time before each spike. Every spike is assigned t= 0, and the timing of every stimulus in the random train preceding the spike is plotted on the Xaxis. Generation of a RCH required that the amplitude of the spike potentials be greater than thestimulus artifact. This was necessary due to manner in which this histogram was constructed.The computer scope was triggered by a voltage deflection exceeding the preset threshold(presumed to be a spike), and plotted the pre-trigger time of every TTL pulse. Stimulus artifactsexceeding the voltage threshold of the computer scope would trigger the scope and be treated asspikes. Thus, the computer scope voltage threshold had to be set above the maximumdisplacement produced by stimulus artifacts and below the minimum displacement produced byspikes. This histogram was used to measure the intrinsic frequency preference (if any) in theinput to the neuron.20For some cells, an interaural time series was performed. This consisted of a series ofbinaural PSTHs, constructed from 30 stimuli. The stimulus amplitudes were adjusted such that amonaural stimulus was just below threshold for the cell under study. Each PSTH wasconstructed with a different interaural delay, ranging from up to 3000 jis ipsilateral ear leading to3000 jis contralateral ear leading. The time delays used were based on several considerations:1. As the maximal interaural time discrepancy that a chicken experiences is approximately 300 Is(based upon head diameter), relatively more points were chosen in this range.2. The oscillating cells that were found had on average a period of approximately 10 ms (seeresults), suggesting that the oscillatory period of the type 2 cells described in the results mighthave a similar value (see discussion), so the range of points used should cover at least half of thatperiod (5 ms).3. The entire experiment should be done as quickly as possible because of the constant risk oflosing the cell. The resulting compromise consisted of steps of 40 jis around 0 ITD, with largersteps (100-1000 is) at larger ITDs.VII. Estimation of inherent frequency:For cells exhibiting regularly spaced peaks in either their IH, PSTH or RCH, theirinherent oscillatory period was taken to be the average distance (on the time axis) betweensuccessive peaks. This distance was measured with the aid of an analysis routine in thecomputerscope software. A vertical line was placed through the centre of each peak (as estimatedby eye) and the value on the time axis of each line was displayed with a resolution of 0.04 ms.Due to the width of the peaks and the difficulty in discerning the precise centre of each peak, thisprocedure generated a value for the oscillatory period with a precision of ± 0.5 ms, dependingupon the sharpness and number of peaks.VIII. Histological verification of recording sites:For selected experiments in which a region of oscillating neurons were found, wheat germagglutinin conjugated to horseradish peroxidase (WGA-HRP type VI, Sigma) was dissolved in21the recording solution as described above. The protein was ejected from the electrode tip using 2hA current pulses (tip positive) delivered by a NeuroLog system NL800 stimulus isolator. Thepattern of current pulses was generated by a 0.25 Hz square wave (Simpson model 420 functiongenerator) fed through a NeuroLog pulse buffer into the stimulus isolator, resulting in currentpulses of 2 seconds in duration separated by 2 second pauses. The ejection was maintained for10-15 minutes at the desired site, then the electrode was withdrawn at about 4 jim/s (whilemaintaining the current pulses) to mark the track. Following WGA-HRP ejection, the animal wasperfused under deep anaesthesia with 0.5% sodium nitrite in .9 N saline at ca 35°C followed byparaformaldehyde/gluteraldehyde. The brain was removed and sectioned sagittally on a freezingmicrotome into 50 jim thick sections. These were then incubated for horseradish peroxidase(HRP) and counterstained with the Nissl stain thionine. Black and white photographs were takenunder white light using a photomacroscope (Wild M400).22RESULTSA total of 81 neurons were isolated. Of these, 69 were classified into 1 of 3 types on thebasis of their spontaneous and evoked activity. The remaining 12 could not be classified.Type 1 cells exhibited spontaneous activity without an intrinsic frequency preference (seebelow) and were briefly suppressed by stimulation of one or both cochlear nerves. Type 2 cellsexhibited little or no spontaneous activity, and responded to stimulation of either cochlear nervewith a single spike or short burst of spikes. Type 3 cells exhibited a high degree of spontaneousactivity and displayed a frequency preference in their firing pattern (see below). This samefrequency preference was seen in response to single stimuli or random trains of stimuli of thecochlear nerve.Type 1: These cells fired spontaneously and randomly, producing a roughly Poissondistribution in the interval histogram (IH). Figure la shows the typical appearance of an IH of atype 1 cell. There is only one, broad peak, extending over a 20 ms period. The firing rate (notshown) was normally quite high, often approaching 100 spikes/s, but without a dominantfrequency in the firing pattern.Type 1 cells responded to inputs from both nerves with a period of suppression lasting 3to 46 ms (mean 16.7, S.D. 11.3, n=16) after a latency of 3 to 14 ms (mean 7.5, S.D. 3.3).During this period, these cells were completely silent. Figure lb shows a typical example of aPSTH from a type 1 cell. The thin peak occurring at 2 ms represents the stimulus artifact. A veryslight increase in firing probability is visible at approximately 5 ms (3 ms post-stimulus),suggestive of a type 2 response (see below). A profound suppression was often the onlyindication that the cell was responsive to stimulation of the cochlear nerve. In this example, thesilent period extends from approximately 5 to 15 ms on the time axis. Following this period, thecell resumes firing with random probability, visible here as an approximately flat line along thetime axis.Type 1 cells were found distributed throughout the region yielding field potentialsfollowing cochlear nerve stimulation (the presumed ICC) and were often the only cell type in the2315-CIz5-I I I I40 60 80 100I[IIii IIIIII liii liii IIIIIIII I III II ill IUhIIII1III NI III0 20 40 60 80 100 120 140 160Time (ms)70-60-50-40-I30-zA20-01 I0 20L..LL. I. 1.1Time (ms)20-10-B0Figure 1A: An interval histogram of a type 1 cell. The time between every 2 consecutive spontaneousaction potentials (spikes) is plotted on the X-axis. Histogram constructed from 150 sweeps. Thebin width is 320 jis. Cell 9.3.H., file EL91531H. B: Post stimulus time histogram of a type 1cell. The stimulus artifact is visible at time = 2 ms. A region of suppression is visible betweentime = 5 and 15 ms. Histogram constructed from 100 stimuli. The bin width is 20 is. Cell2.18.C., file ELi l/69PS.24area. A total of 16 such cells were recorded, although many more were isolated, suggesting thatthis is a common cell type in the ICC.Type 2: The second type exhibited very little or no spontaneous activity. Out of 21 type 2cells isolated, 14 exhibited no spontaneous activity and the remaining 7 fired less than 5 spikesper second. Their response to a single pulse stimulus was phasic, that is, a single AP or a shortburst of APs after a latency of approximately 6 ms. The response latency was similar forcontralateral and ipsilateral stimuli (contralateral mean 6.06 ms, S.D. 1.8, ipsilateral mean 6.71ms, S.D. 2.1, n = 13). Figure 2 shows a typical PSTH of a type 2 cell. A single very sharp peakoccurs at a relatively fixed interval (latency) after the stimulus. Following this onset response,this cell remained silent until the next stimulus.For some type 2 cells, an interaural time series was performed as described in themethods. When recording extracellular activity, oscillations are only visible when there is steadystate spike activity. A type 2 cell, without any spontaneous activity, might still oscillate atsubthreshold levels. Such oscillations could be revealed by repeated stimulation, such as occursduring an lTD series. Figure 3 shows the results from three such experiments. The series shownin figures 3a and 3c were extended to ±3000 p.s, but showed no change in response between±1000 and 3000 Jis, and so the plots were truncated to allow a reasonable resolution around zero.Although type 2 cells did respond differently at different 1TDs, no cyclic pattern was visible in thelTD series.Type 2 cells were found spaced throughout the presumed ICC and tended to have largeamplitude field potentials, suggesting very large cells. They were the most common cell typeisolated, suggesting that this may be the dominant cell type in the ICC under anaesthesia.Type 3: These cells exhibited a high spontaneous rate with a periodicity in their firingpattern, as shown in the IH. Figure 4 shows three examples of IHs from type 3 cells. The firstone (fig. 4a) shows only a single peak with a very narrow base (compare fig. la). This isrepresentative of a cell whose firing rate is equal to the inherent frequency. The interspike intervalfor this cell is given by the time of the single peak, which is centred around 15 ms. Thiscorresponds to an inherent frequency of 67 Hz. The second one (fig. 4b) exhibits several modes,258-6-— I I I I I0 20 40 60 80 100Time (ms)Figure 2Post stimulus time histogram of a type 2 cell. The stimulus occurs at time = 2 ms (not marked).All of the cell’s responses occur at approximately 8 ms (6 ms after the stimulus). Histogramconstructed from 50 stimuli. The bin width is 20 jis. Cell 12. 16.C., file ELi 1/32PS.26I)C11:‘IC4t:ABI • I1210•Figure 3Interaural time series for 3 cells. Each point represents the sum of responses to 30 stimuli at agiven interaural time discrepancy shown on the X-axis. A positive contralateral delay indicatesthat the stimulus to the cochlear nerve contralateral to the recording site was delayed relative to thestimulus to the ipsilateral nerve. A negative contralateral delay indicates that the contralateralstimulus led the ipsilateral one by the specified amount. A: Cell 11 .27.C. B: Cell 10.2.D. C:Cell 12. l0.E. The rationale for the interaural delays chosen is given in the methods.I • IC-800 -400 0 400 800Contralateral Delay (jis)Figure 4Interval histograms from three type 3 cells. The time between every 2 consecutive spontaneousaction potentials (spikes) is piotted on the X-axis. A: This cell exhibits a single mode atapproximately 15 ms, corresponding to its most common interspike interval. 50 sweeps, binwidth 20 pUs. Cell 4.20.D., file EL12/811H. B: This cell exhibits 4 modes with an intermodalinterval of 8.5 ms and a most common interspike interval of 8.5 ms. 100 sweeps, bin width 320ps. Cell 2.24.B., file ELi l/721H. C: This cell exhibits 10 modes with an intermodal interval of8.5 ms and a most common interspike interval of 25.5 ms. 100 sweeps, bin width 320 its. Cell4.20.C., file EL12/801H.272825-20 A-4C10-z5-II____________________________________________________________________0-• I I I •0 20 40 60 80 100Time (ms)120-B100-•-480-C60-40-z20-0 -•• - • i ‘ ‘ - I0 20 40 60 80 100Time (ms)2520C,)15IiiIJMIUiii E11E 10z500 20 40 60 80 100Time (ms)29with the first one larger than the others, and all of them equally spaced along the time axis. Thisindicates that this cell does skip beats, but that the intrinsic frequency (given by the intermodalinterval or the time to the first peak) is still the most common firing frequency. Thus this type 3cell differs from a type 1 cell in that all spikes fall into one of the equally spaced modes of the JR.The third one (fig. 4c) exhibits several modes, with the third one being the highest. The mostcommon interspike interval for this cell is approximately 25 ms (the interval of the third peak), butthe intrinsic period, which is represented by the intermodal interval, is 8.5 ms, corresponding to afrequency of 118 Hz. This pattern would occur if an oscillating membrane potential reachesthreshold more frequently after missing one or two periods.A total of 26 type 3 cells were isolated. Their intrinsic frequencies ranged from 67 Hz to250 Hz. The distribution of frequencies among the cells is shown in figure 5. The majority ofcells possessed an intrinsic frequency near 100 Hz, with none exhibiting frequencies below 60 orabove 260 Hz.These oscillatory neurons typically responded to stimulation of the contralateral cochlearnerve (and occasionally the ipsilateral as well) and were entrained to the stimulus while retainingtheir inherent frequency. As a result, the PSTH of a type 3 cell was multimodal, with the sameintermodal interval as the 1}I. Figure 6 illustrates this. This PSTH is taken from the same cellwhose JH is shown in fig. 4a, and displays the same intrinsic period (15 ms). Every stimuluspulse appeared to reset the rhythm of the cell, after which it resumed its regular firing until thenext stimulus. An onset response is visible after a short latency. This may be a response of theoscillating cell, or the response of a nearby type 2 cell, or a combination of both. The possiblecontamination of the recording by a type 2 cell is suggested by the strong onset response withoutany subsequent activity between stimuli. Increasing the intensity of the stimulus had no effect onthe intrinsic firing frequency of the cell (results not shown).In order to determine whether these cells can serve as filters by extracting a specificfrequency from a random input, a RCH was constructed from individual type 3 cells. Thus, arandom train of stimuli was applied to the cochlear nerve, and the timing of these stimuli wasplotted with reference to every spike. As expected, the same intrinsic oscillatory frequency found3076Figure 5Distribution of intrinsic oscillatory frequencies of the 26 type 3 cells isolated. The frequencieswere measured as described in the methods.C C00 C CNI I IC C C00 C(“1Frequency (Hz)315040IZ20I‘.4... 1A, ‘.4...’ 40 20 40 60 80 100 120 140 160Time (ms)Figure 6Post-stimulus time histogram of the same cell pictured in fig. 4a. The stimulus artifact is visibleas a vertical line at time 2 ms. An onset response is visible at 4-5 ms. The intennodal intervalis 15 ms, corresponding to a frequency of 67 Hz. Histogram constructed from 200 stimuli. Binwidth 160 Ls, cell 4.20.D., file EL12/82PS.32in the IH (fig. 7a) and PSTH (fig. 7b) was seen in the response to a random train of stimuli(RCH), demonstrating that these cells respond preferentially to stimuli which occur at theirpreferred interval (fig. 7c).Type 3 cells were found exclusively in a small region of the ICC and relatively few werefound in any given animal. They tended to have small amplitude spike potentials, making themdifficult to isolate from background activity, and were more difficult to maintain close to theelectrode tip than the other cell types. In order to determine histologically the location of thisregion, WGA-HRP injections were made as described in the methods. Figures 8 and 9 showsections through WGA-HRP labeled sites from two animals. The section shown in figure 8 wastaken from an animal which received 2 separate injections approximately 100 jim apart rostrocaudally. Both injections were made at the deepest (i.e. most ventral) point of oscillatory activity.The two spots shown by arrows are located in the core of the ICC (ICCc). They are roughly 100jim apart on the rostro-caudal axis and could therefore represent the separate tracts. The sectionshown in figure 9 was taken from an animal which received 2 injections from 2 separateelectrodes. Due to differences between electrodes, and the manipulations required to changeelectrodes during an experiment, the distance between tracts produced by two different electrodescould not be reliably measured. The spots marked “A” and “B” are both located in the ICCc, andmay represent 2 separate tracts or uneven ejection and/or diffusion from a single tract. EveryHRP labeled tract was located either within the ICCc or along a line directed towards the ICCc.Figure 7Inherent frequency of a type 3 neuron expressed in the interval histogram, post-stimulus timehistogram and reverse correlation histogram. A: IH constructed from 100 sweeps, in thepresence of a random train of stimuli, showing a large peak at 4.5 ms and a smaller one at 9 ms.The intervening noise could be caused by a poor isolation of the oscillating neuron or doubletfiring by this cell, file EL9/68111, bin width 80 ps. B: PSTH from the same cell exhibiting thesame intrinsic frequency in response to 100 individual pulse stimuli applied to the contralateralcochlear nerve. File EL9/64PS, bin width 320 us. C: RCH constructed simultaneously with theN shown in A. It exhibits the same periodicity in its responses. File EL9/67RC, bin width4Ojis. Cell 9.3.M.3334I II I I II I I II30 40 50I • I II I I40 50 6080—60—40—20—A0 -r1’ I I I I10CIDIz05040cdD1)C,)00302510‘III20 60Time (ms)BATime (ms)I.IC20—5-20 -15 -10 -5Time (ms)35Sagittal section through the right optic tectum of a chicken which was injected with HRP fromtwo separate electrode penetrations as outlined in the methods. The electrode tracts were madeapproximately 100 .Lm apart rostro-caudally. The 2 spots labeled by arrows indicate regions ofHRP activity located in the core of the central nucleus of the inferior colliculus (ICCc). Scale:each axis arrow is 280 j.tm long. ICCs: shell of the central nucleus of the inferior colliculus; D:dorsal; R: rostral. Experiment 6.11.Figure 836%_ I- •.‘i•._ -.:a ‘- —iccett:e ‘i.r; -• -:e •• o: -:/ --‘1 tSagittal section through the right optic tectum of a chicken which was injected with HRP fromtwo separate electrodes as outlined in the methods. The two spots labeled A and B indicateregions of HRP reactivity located in the core of the central nucleus of the inferior colliculus(ICCc). Scale: each axis arrow is 245 jam long. ICCs: shell of the central nucleus of the inferiorcolliculus; D: dorsal; R: rostral. Experiment 5.19.*t.%1k. .%%.. %.--aFigure 937DISCUSSIONI. Comparisons with Previous Studies:Most studies of auditory neurophysiology used sound stimuli, delivered either in the free-field or from closed systems within the ear canal to permit dichotic investigations. Stimuli havetaken the form of noise, clicks, or pure tones and had durations much longer than 1 ms, allowingthe coding for ongoing sound properties (as opposed to purely the presence of the signal) to bestudied. This included phase locking where possible. The interaural time sensitivitydemonstrated in the owl IC (see introduction) was elucidated under such conditions.The approach of the present study differs significantly from most of the literature in thisfield because of the experimental protocol. The removal of the cochlea and use of electrical pulsestimuli is “unphysiological” in that the signals traveling to the CNS are not likely to resembleanything the animal could experience naturally. The impulsive stimulus is shorter than theduration of an action potential, and not repetitive, excluding phase-locking in the CNS. Theexcitation of all auditory nerve fibres is assumed to be simultaneous. It is presumed that thesignal in the cochlear nerve is a single spike traveling simultaneously in most or all intact fibres.This paradigm was designed to reveal inherent frequency selectivity, including oscillations, ofcentral auditory neurons, and exclude contamination due to basilar membrane ringing, hair cellresonance, and periodic responses due to phase locking to a periodic stimulus.One important aspect of this paradigm, which does lend itself to comparison with anumber of previous studies, is the use of anaesthetic during recording. While it is reasonable toassume that anaesthetics will alter certain neuronal properties, it was felt that their use wasjustified on several grounds. Firstly, limited studies comparing anaesthetized preparations withothers (decerebrate or awake) have found most results to be similar. In particular, the mainresponse patterns characterizing auditory brainstem neurons have been found in all of these threedifferent paradigms without qualitative difference (Rhode and Kettner, 1986, Ritz and Brownell,1982, Webster, 1977). Secondly, as stated in the introduction, a decerebrate preparation is notcompatible with recordings in the midbrain, and an unanaesthetized preparation following38extensive surgery was not possible (see introduction). Finally, given the extensive use ofanaesthetics in the literature, comparisons with previous studies are easier to make using datafrom an anaesthetized animal.II. Type 1 cells:A. Spontaneous activity:Cells described in this study as exhibiting a high rate of irregular spontaneous activitywere classified as type 1. The pattern of activity of a typical type 1 cell is illustrated in theinterspike interval histogram shown in figure 1 a. The distribution of interspike intervals can beapproximated by a Poisson distribution, thus this cell type fires in a random pattern.The average spontaneous firing rate of these cells was usually between 20 and 120spikes/s. While this activity could arise from spontaneous fluctuations of the membrane potentialof these cells independently of synaptic inputs, the interval histograms are reminiscent of manycentral neurons receiving a continuous synaptic barrage on their dendritic trees (eg.motoneurons). A high rate of stochastic spontaneous activity at lower levels is prevalent in birdswith intact cochleae (Sachs, et al., 1974, Sachs and Sinnott, 1978), suggesting that this activitymay be inherited from below. In order to determine the functional significance of this pattern ofactivity, recordings must be made from these cells in the presence of acoustic stimuli. The originof the observed activity is uncertain. In the context of the processing of the auditory codes it is ofinterest that, in contrast to the type 2 cells described below, this cell type is not tonicallyinhibited.B. Response to stimuli:The only physiological evidence that type 1 cells receive auditory input is the suppressionof firing seen in the response to pulse stimuli. As shown in the PSTH (fig. ib), there was aperiod of post-stimulus suppression of firing. This suppression may be due to inhibition fromGABAergic cells within the IC reported by Granda and Crossland (1989), or projections fromlower centres. The avian nucleus intermedius of the lateral lemniscus contains primarily GABA39immuno-reactive cells, and projects to the contralateral ICC (MUller, 1987, Muller, 1988), andtherefore is a likely candidate for inhibitory input. A glycine-transporting projection system to theICC from several lower centres was recently described for the chicken (Schwarz, et al., 1994),suggesting that suppression may also be glycinergic.There was considerable variability in the latency to the onset of suppression recorded forthese cells; latencies varied between 3 and 14 ms, with a mean of 7.5 ms. The significance of thisvariability cannot be determined from the results presented here, but several possible sources ofvariability can be suggested from the literature. IPSPs at various regions of a cell’s dendritic treecan certainly give rise to spike train suppressions at different latencies. The suppression mayfurthermore be caused by synaptic inhibition arising from one or both of the transmitter specificpathways listed above (or other transmitter systems as yet undocumented in the ICC may beinvolved), and may originate at one of several possible sites in the auditory pathway. These sitesin turn vary in their interconnections, introducing another source of variability in transmissiondelay. Alternatively, if the spontaneous activity of these cells arises at lower centres, there may besynaptic inhibition occurring at those centres. A characterization of the underlying mechanismsrequires intracellular recording from these cells with application of GABA and glycine agonistsand antagonists.Data were obtained from 16 type 1 cells, which were found throughout the region of theICC (judged to be that region of the IC yielding field potentials following cochlear nervestimulation). This small number is not indicative of the actual proportion of type 1 cells in theICC. Many more were isolated but not recorded since they were not a focus of this study ofneuronal frequency selectivity. It is estimated that this cell type is more numerous than type 3cells described below.III. Type 2 cells:A. Response to stimuli:Cells in the present study which exhibited little or no spontaneous activity and whichresponded to an impulse stimulus with a single spike or short burst of spikes were classified as40type 2. This simple response pattern is illustrated by the PSTH shown in figure 2. The responselatency was roughly 6 ms and relatively constant, and the response was invariably followed by aperiod devoid of spike firing. Previous studies in the owl IC have documented the existence of‘silent’ neurons at this level: Wagner (1990, 1992) found that IC neurons generally displayed nospontaneous activity. No comparable studies have been carried out on the chicken.Although this study did not address whether the response (fig. 2) was followed byinhibition, or was merely brief due to the duration of the stimulus, preliminary studies haveinvestigated this (Schwarz, et al., 1993). It was found that the response is followed by a periodof suppression lasting between 40 and 200 ms during which a stimulus of similar intensity willnot evoke a response. As with the suppression of type 1 cells discussed above, this may be dueto GABA or glycine release from local cells or from lower centres, or to inhibition of cells inlower nuclei providing inputs to the ICC.The PSTH of a type 2 cell is very similar to the PSTH of a class of cells termed “on” cellswhich were originally recorded in the mammalian cochlear nucleus under the influence of soundstimuli (Pfeiffer, 1966). The term ‘on’ type response has since been applied to any responsewhich exhibits a firing pattern similar to the PSTH in figure 2 when stimulated acoustically.Typically, a longer stimulus (—200 ms) is used, and an “on” cell responds only to the initialportion. The type of stimulus used in the present study, and the relative locations of the twoclasses of cells (mammalian cochlear nucleus and avian midbrain), precludes any directcomparison of the two cell types. To prevent confusion of these two cell types, the term ‘on’response has been avoided when referring to type 2 cells.B. Interaural time sensitivity:Cells in the inferior colliculus of the owl create a map of auditory space by firingmaximally to a particular frequency (defined as the cell’s characteristic frequency, CF) at aparticular IPD. As the IPD for a tonal stimulus is varied, the response cycles from maximal tominimal to maximal again with the period of the stimulus tone. This is a simple consequence ofthe repetitive nature of a cyclic stimulus. An IPD of it is often indistinguishable from an 1PD of413it, therefore one would expect a similar response in the IC to both IPDs. The spike probabilityis usually highest at the onset of the response, particularly for owl IC cells, with the result that afavourable IPD produces an “on” type response, whereas an unfavourable IPD does not evokeany response (Wagner, 1990). Cells of the owl’s ICC produce a more sustained response tofavourable IPDs, similar to the “on-type L” units described in the PVCN (Godfrey, et aL, 1975).These units maintain a low level of activity for the duration of the stimulus.In order to pursue the possibility that the type 2 cells described here possess an intrinsicinteraural time sensitivity independently of the periodic nature of acoustic stimuli, interaural timeseries were performed on several type 2 cells. We predicted a cyclic response to varying ITDs aswas reported above for IPDs, with the period of the cycle dependent upon the intrinsic filterfunction of the unit (the CF could not, of course, be determined in this study). At least oneresponse maximum should occur within the physiological range experienced by chickens (i.e. 300jis). Twice that range (between -600 ts and +600 ps) was studied using small lTD increments(see methods). The negative results shown in figure 3 must be considered preliminary since weabandoned a search for intrinsic oscillations in these cells. No clear response maximum orperiodicity was found at less than ±600 jis of lTD. There is, therefore, no evidence forsubthreshold oscillatory responses which could have served to produce an intrinsic interaural timeselectivity, or for network mechanisms accounting for it. The negative results do not rule out arole for these cells in interaural time delay processing. lTD selectivity could well depend uponoscillatory responses available only when using sound stimuli. It is therefore possible that thesecells require an ongoing binaural stimulus from which to extract the IPD. This latter conclusion issupported by Wagner (1990), who found that the IPD sensitivity of owl ICC “on” neuronsimproved with time during a 120 ms long stimulus.C. Oscillatory activity:The principal purpose of the lTD series trials was to test for the existence of subthresholdoscillations of type 2 cells. The possibility exists that the membrane potential of this cell typeoscillates below threshold, such that stimuli which are timed in phase with these intrinsic42oscillations will be more likely to elicit spikes. This increased firing probability should be visiblein the summed responses to 30 stimulus presentations making up a PSTH.To investigate this hypothesis, it is not necessary that the stimuli come from separate ears,or that they be presented in pairs (rather than triplets or quartets etc.), or that the amplitude of eachone be subthreshold to the generation of spikes when presented alone. Nevertheless, it wasdecided that the same paradigm used for the lTD sensitivity could be used for a preliminary study(see methods for a description of the lTD series paradigm). The time between stimuli (ITD) wasvaried over a larger range in order to test for oscillations as low as 100 Hz. This value of 100 Hz(corresponding to a period of 10 ms) was chosen because it was the most common frequencyexpressed by type 3 cells (described below, and see figure 5). Thus a range of ±5 ms lTD wasneeded.The results shown in figure 3 suggest that no periodicity was present in the probability ofspike firing over the time intervals studied. These results do not rule out the possibility that thesecells possess oscillating potentials. The parameters listed above , in particular the stimulusamplitude, may have been inappropriate for demonstrating the desired response. In order toproperly characterize the response properties, these experiments should be repeated with differentparameters. In particular, the stimulus amplitude should be varied, and include pairs of stimuliwhich are suprathreshold when presented alone, to test the hypothesis that the stimuli used in thepresent study were not sufficiently intense to evoke a response. Monaural stimuli should also beapplied, because it cannot be assumed that the signals originating at the two cochlear nerves exertthe same effect on this cell type. The number of stimulus presentations at each lTD should beincreased above 30, and smaller lTD steps used to increase the resolution. Unfortunately, it wasgenerally not possible to maintain the cell isolation long enough to carry out all of thesemeasurements.The large number of these cells suggests that they play a prominent role in the auditorypathway. A total of 21 were examined in this study, but many more were isolated. It is estimatedthat this may be the most common cell type in the ICC. The role that these cells play in auditorysignal processing could not be determined in the present study.43IV. Type 3 cells:A. Spontaneous oscillations:These cells exhibited a robust oscillatory activity at higher frequencies than mostoscillating neurons reported in the literature. The spontaneous activity of type 3 cells exhibitedpreferred interspike intervals, as illustrated in the interval histogram (fig. 4). For some type 3cells, only a single peak was visible in the IH (fig. 4a). In contrast to the IH of a type 1 cell (fig.1), the peak shown in figure 4a is much taller relative to its width at the base, and does notapproximate a Poisson distribution. The firing pattern of this cell is similar to that of ametronome, which never skips a beat. Expressed another way, the ‘rate’ of this cell, given by thetotal number of spikes averaged over the total time (67 spikes/s), was equal to its preferred‘frequency’, given by the inverse of the interspike interval (67 Hz).The firing rate of type 3 cells was not always equal to the preferred frequency. Morecommonly, these cells tended to ‘skip beats’, implying that they did not fire with every cycle ofoscillation. This resulted in multimodal interval histograms in which the peaks were equallyspaced along the time axis (fig. 4b and c). The cell illustrated in figure 4b expressed an intrinsicperiod of 8.5 ms, which corresponds to a frequency of 118 Hz. The tallest peak of an intervalhistogram indicates the most common interspike interval, which for this cell was also 8.5 ms.The existence of a peak at 17 ms indicates that this cell sometimes skipped one cycle, and thesmall peak at 25.5 ms indicates that it rarely skipped two cycles. The firing pattern of the cellillustrated in figure 4c was much more complex. The large number of peaks indicates a widevariety of interspike intervals. In this case, the tallest peak was located at 25.5 ms, correspondingto the most common interspike interval. The inherent frequency is given by the time to the firstpeak or the intermodal interval, which for this cell was 8.5 ms, corresponding to 118 Hz.Although each type 3 cell exhibited a single robust frequency in its spontaneous activity,the population of type 3 cells isolated (26 in total) covered a wide range of frequencies. Figure 5is an histogram illustrating the distribution of frequencies. The equipment was set up to detectfrequencies down to 6.25 Hz, but this was not thought to represent a serious limitation, as thelowest frequency observed was 67 Hz.44We predicted the existence of inherently oscillating cells to assist the fine time resolutionobserved in psychophysical studies of directional hearing, and to assist in spectral processing byproviding a multiple-band frequency filter. Membrane potential oscillations at auditoryfrequencies in cells of the NL, in concert with convergence of many phase-locked inputs, couldimprove the resolution provided by traditional neural mechanisms such as spike trains andEPSPs. A requirement arising from this hypothesis is that the entire frequency spectrum forwhich timing information is desired be represented by inherent oscillations. For the chicken,phase-locking in auditory nerve fibres is found up to at least 2 kHz (Warchol and Dallos, 1989).With a wavelength of 17 cm, sounds of this frequency are unlikely to produce much of an I]D,and therefore require fine time information to permit accurate localization. Unfortunately, out of26 oscillating neurons isolated, none exhibited inherent frequencies higher than 250 Hz.Preliminary studies found higher inherent frequencies of up to 840 Hz (Schwarz, et al., 1993).B. Oscillations in the impulse response:Oscillations of type 3 cells were entrained by impulse stimuli. As a result, the poststimulus time histograms of type 3 cells were multimodal, with the intermodal interval equal to theinherent frequency demonstrated in the cell’s IH. Figure 6 shows a PSTH from a type 3 cell.Spikes occurred at 15 ms intervals following each stimulus, corresponding to a frequency of67Hz. Entrainment of this cell to the stimulus is indicated by the relatively constant latency to thefirst response following 200 stimuli. The interval histogram for the same cell is shown in figure4a (see section IV.A.). The intermodal interval is identical in the two histograms, demonstratingthat this cell expressed the same inherent frequency both spontaneously and in response toimpulse stimuli. Increasing the amplitude of the stimulus had no effect on the intermodal interval(results not shown).These findings demonstrate that the oscillatory activity of type 3 cells is robust. The sameinherent frequency found in the Ill (see section IV.A.) is also seen in the PSTH and isindependent of stimulus amplitude. Support for the hypothesis that these cells are able to extract a45specific frequency from a pulse train requires the demonstration of a preferred frequency in arandom pulse sequence input.C. Response to a random pulse sequence:Two type 3 cells were isolated which had spike amplitudes greater than the stimulusartifact and were maintained at the electrode tip long enough to perform a reverse correlationhistogram. The RCH for these cells was multimodal, and displayed the same intermodal intervalas was shown in the cells’ N and PSTH (fig. 7). In contrast to the PSTH and N, in which theintermodal interval represents a commonly occurring interspike interval, the intermodal interval ofa RCH represents a preferred inter-stimulus interval, and the absolute value of the position of eachmode on the x-axis represents a commonly occurring response latency. Type 3 cells preferred afrequency in the stimulus input that was equal to the spontaneous and entrained inherentfrequency of the cell. This property could enable these cells to signal the existence of specificfrequencies in a complex input. With a population of such neurons having different inherentfrequencies, a complex wave input could be filtered into its component frequencies by a running“Fourier transform” similar to that resulting from the auditory filters (see introduction). A use forsuch a filter has previously been proposed in a theoretical model put forth by Srulovicz andGoldstein (1983), although the range of the results presented here is not sufficient to support themodel proposed.D. Frequency preferences: mechanisms and implicationsThe mechanism underlying the inherent frequency preference cannot be determined fromthe results of this extracellular recording study, but there are only a limited number of likelypossibilities. These cells may possess inherently resonant membrane potentials. Given theproper combination of ion channels and external conditions (i.e. presence or absence oftransmitters or modulators), the membrane potential of these cells could resonate within a verynarrow frequency range, and the constant barrage of inputs generated spontaneously from lowercentres could disturb the potential enough to maintain oscillations at the resonant frequency. The46extent of spontaneous activity in the auditory pathway would certainly be sufficient.Alternatively, the oscillations may persist in the absence of synaptic input, once again due to theparticular properties of these neurons. Intracellular recordings from these cells could solve manyof these mysteries. Another possibility altogether is that these cells are merely reproducing spiketrain activity received from lower centres, notably the nucleus laminaris. This possibility issupported by the finding that the NL generates an oscillating neurophonic potential in response totonal stimuli (Schwarz, 1992), and that this nucleus sends projections directly to the ICCc(Takahashi and Konishi, 1988, Wagner, et aL, 1987, Conlee and Parks, 1986), the location ofthe cells reported here (figs. 8 and 9, and see below).While it seems clear that these cells could serve as a frequency filter within the CNS, wehave not demonstrated that they filter coded sound frequencies. Their PSTHs (fig. 6) resemblethose of chopper cells reported in the manmialian cochlear nucleus (CN, Pfeiffer, 1966, Banksand Sachs, 1991), with two important differences. Firstly, the chopping frequency of cells in theCN is independent of stimulus frequency, and may change with stimulus intensity (Oertel, et al.,1988, Young, et al., 1988). In contrast to choppers, the ICC cells reported here oscillate atrelatively fixed frequencies which are stable with changing stimulus intensity. Secondly, thechopper response persists only as long as the stimulus, whereas ICC cells exhibit a steady-stateoscillating background activity which is reset by an impulse (figs. 4 and 6). In addition to thedifferences in the response to stimuli, there is another important difference between type 3 cells ofthe ICC and chopper cells of the cochlear nucleus: chopper cells show no or little irregularspontaneous activity, in contrast to the regular spontaneous firing of type 3 cells. For thesereasons, we have avoided applying the term “chopper” when describing the responses of thesecells.It would be useful to establish whether this cell type possesses any sensitivity to ITDs. Acombination of a simple spike count over the entire histogram period and a measure of sharpnessof the 111 or PSTH peaks (vector strength) could provide a parameter for comparison at differentITDs. It should be noted that almost all of the type 3 cells isolated were responsive only to47stimulation of the contralateral electrode, discouraging any lTD studies. The significance of thiscontralateral selectivity, if any, is not known.The histological determination of the location of type 3 cells shows clearly that theirdistribution is restricted to the ICCc. Figures 8 and 9, taken from two different animals, andadditional sections from two other animals (not shown), all point to a relatively small regionrostro-ventrally in the medial ICC. An interesting follow-up to this would be to combine HRPlabeling of oscillating neurons with anterograde tracing from the NL, particularly from thecontralateral side. The NL receives excitatory inputs from both ears, and has been shown toproject primarily to the medial aspect of the ICC (Conlee and Parks, 1986). Given that the type 3cells characterized in the present study were responsive solely to contralateral stimulation, it islikely that type 3 cells do not mediate the output of the NL.V. Conclusions:The ICC contains at least 3 types of neurons based upon their responses to electrical pulsestimuli of the auditory nerve. One of these exhibits spontaneous oscillations at auditoryfrequencies that are robust. These oscillations persist at the same frequency in response tostimuli, and as a result are able to extract frequency-specific information from a random input(noise). This supports our hypothesis that neurons in the central auditory pathway possessinherent resonances at auditory frequencies. Type 3 cells differ from NL neurons in theirconnectivity with the ears, however. Their frequency selectivity can, therefore, not explain theinteraural time selectivity and its fine resolution in directional hearing.A large degree of inhibition is evident in the firing patterns of the other cell types. Thisphenomenon deserves further study to establish the cause and behaviourailfunctional significance.The findings presented here (and published previously by Schwarz, et al., 1993)demonstrate the first extracellular evidence for inherent oscillations in the central auditorypathway. The oscillatory frequencies reported for the type 3 cells are one to two orders ofmagnitude higher than those reported in the mammalian thalamus, suggesting that the cellularmechanisms underlying the oscillations in these two systems are different. The results of thisstudy also represent the first physiological characterization of the chicken ICC.4849REFERENCESBanks, M.I. and M.B. Sachs (1991) Regularity analysis in a compartmental model of chopperneurons in the anteroventral cochlear nucleus. J Neurophys 65: 606-629.Blauert, J. (1983) Spatial hearing: The psychophysics of human sound localization, Cambridge,MA: MiT Press.Carr, C.E. and M. Konishi (1988) Axonal delay lines for time measurement in the owl’sbrainstem. Proc Natl Acad Sci 85:8311-8315.Conlee, J.W. and T.N. Parks (1986) Origin of ascending auditory projections to the nucleusmesencephalicus lateralis pars dorsalis in the chicken. Brain Res 367: 96-113.Dezsö, A., D.W.F. Schwarz and I.E. Schwarz (1993) A survey of the auditory midbrain,thalamus and forebrain in the chicken (Gallus domesticus) with cytochrome oxidasehistochemistry. J Otolaryng 22(5): 39 1-396.Eggermont, J.J. (1992) Stimulus induced and spontaneous rhythmic firing of single units in catprimary auditory cortex. Hear Res 61: 1-11.Fuchs, P.A. and G. Evans (1988) Voltage oscillations and ionic conductances in hair cellsisolated from the alligator cochlea. J Comp Physiol A 164: 151-163.Fuchs, P.A., T. Nagai and M.G. Evans (1988) Electrical tuning in hair cells isolated from thechick cochlea. J Neurosci 8: 2460-2467.Godfrey, D.A., N.Y.S. Kiang and B.E. Norris (1975) Single unit activity in the posteroventralcochlear nucleus of the cat. J Comp Neurol 162: 247-268.Granda, R.H. and W.J. Crossland (1989) GABA-like immunoreactivity of neurons in thechicken diencephalon and mesencephalon. J Comp Neurol 287: 455-469.Gutmck, M.J. and Y. Yarom (1989) Low threshold calcium spikes, intrinsic neuronal oscillationand rhythm generation in the CNS. J Neurosci Meth 28: 93-99.Irvine, D.R.F. (1986) “The performance of the auditory central processor.” In The auditorybrainstem, ed. D. Ottoson, H. Autrum, E.R. Perl, R.F. Schmidt, H. Shimazu and W.D.Willis. pp. 6-2 1. Vol 7 of “Progress in Sensory Physiology”. Berlin: Springer-Verlag.Jhaveri, S. and D.K. Morest (1982) Neuronal architecture in nucleus magnocellularis of thechicken auditory system with observations on nucleus laminaris: a light and electronmicroscope study. Neurosci 7(4): 809-836.Knudsen, E.I. (1983) Subdivisions of the inferior colliculus in the barn owl (Tyto alba). J CompNeurol 218: 174-186.Knudsen, E.I. and M. Konishi (1978) A neural map of auditory space in the owl. Science 200:795-797.Knudsen, E.I. and M. Konishi (1978) Space and frequency are represented separately in theauditory midbrain of the owl. J Neurophys 4 1(4): 870-884.50Konishi, M., T.T. Takahashi, H. Wagner, W.E. Sullivan and C.E. Carr (1988)“Neurophysiological and anatomical substrates of sound localization in the owl.” InAuditory Function: Neurobiological Bases of Hearing, ed. G.M. Edelman, W.E. Gall andW.M. Cowan. pp. 721-745. New York: John Wiley and Sons, Inc.Kuenzel, W.J. and M. Masson (1988) A stereotaxic atlas of the brain of the chick (Gallusdomesticus), Baltimore: Johns Hopkins University Press.Langner, G. (1983) Evidence for neuronal periodicity detection in the auditory system of theguinea fowl: implications for pitch analysis in the time domain. Exp Brain Res 52: 333-355.Llinás, R.R. (1988) The intrinsic electrophysiological properties of mammalian neurons: insightsinto central nervous system function. Science 242: 1654-1664.Llinás, R.R., A.A. Grace and Y. Yarom (1991) In vitro neurons in mammalian cortical layer 4exhibit intrinsic oscillatory activity in the 10- to 50-Hz frequency range. Proc Nat AcadSci USA 88: 897-901.Manley, G.A. and 0. Gleich (1992) “Evolution and specialization of function in the avianauditory periphery.” In The evolutionary biology of hearing, ed. D.B. Webster, R.R. Fayand A.N. Popper. pp. 56 1-80. New York: Springer-Verlag.McCormick, D.A. and H.-C. Pape (1990) Properties of a hyperpolarization-activated cationcurrent and its role in rhythmic oscillation in thalamic relay neurons. J Physiol (London)431: 291-3 18.Metherate, R., C.L. Cox and J.H. Ashe (1992) Cellular bases of neocortical activation:modulation of neural oscillations by the nucleus basalis and endogenous acetylcholine. JNeurosci 12(12): 4701-4711.Michelsen, A. (1992) “Hearing and sound communication in small animals: evolutionaryadaptations to the laws of physics.” In The Evolutionary Biology of Hearing, ed. D.B.Webster, R.R. Fay and A.N. Popper. pp. 61-77. New York: Springer-Verlag.Moiseff, A. and M. Konishi (1981) Neuronal and behavioral sensitivity to binaural timedifferences in the owl. J Neurosci 1(1): 40-48.Moore, B.C.J. (1989) An introduction to the psychology of hearing, 3rd ed., 308 pp. San Diego:Academic Press.MUller, C.M. (1987) ‘y-Aminobutyric acid immunoreactivity in brainstem auditory nuclei of thechicken. Neurosci Lett 77(3): 272-276.MUller, C.M. (1988) Distribution of GABAergic perikarya and terminals in the centers of thehigher auditory pathway of the chicken. Cell Tiss Res 252: 99-106.Nuñez, A., F. Amzica and M. Steriade (1992) Voltage-dependent fast (20-40 Hz) oscillations inlong-axoned neocortical neurons. Neurosci 51(1): 7-10.Nuñez, A., R. Curró Dossi, D. Contreras and M. Steriade (1992) Intracellular evidence forincompatibility between spindle and delta oscillations in thalamocortical neurons of cat.Neurosci 48(1): 75-85.51Oertel, D., S .H. Wu and J.A. Hirsch (1988) “Electrical characteristics of cells and neuronalcircuitry in the cochlear nuclei studied with intracellular recordings from brain slices.” InAuditory function: Neurobiological bases of hearing, ed. G.M. Edelman, W.E. Gall andW.M. Cowan. pp. 3 13-336. New York: John Wiley & Sons.Pallotta, T. and A. Peres (1989) Membrane conductance oscillations induced by serum inquiescent human skin fibroblasts. J Physiol 416: 589-599.Parks, T.M., P. Collins and J.W. Conlee (1983) Morphology and origin of axonal endings innucleus laminaris of the chicken. J Comp Neurol 214: 32-42.Pfeiffer, R.R. (1966) Classification of response patterns of spike discharges for units in thecochlear nucleus: tone-burst stimulation. Exp Brain Res 1: 220-235.Pinault, D. and M. Deschênes (1992) Voltage-dependent 40-Hz oscillations in rat reticularthalamic neurons in vivo. Neurosci 51(2): 245-258.Rhode, W.S. and R. Kettner (1986) Physiological study of neurons in the dorsal andposteroventral cochlear nucleus of the unanesthetized cat. J Neurophys 57(2): 414-442.Ritz, L.A. and W.E. Brownell (1982) Single unit analysis of the posteroventral cochlear nucleusof the decerebrate cat. Neurosci 7: 1995-2010.Rose, J.E., L.M. Kitzes, M.M. Gibson and J.E. Hind (1974) Observations on phase-sensitiveneurons of anteroventral cochlear nucleus of the cat: non-linearity of cochlear output. 3Neurophys 37: 218-253.Sachs, M.B., R.H. Lewis and E.D. Young (1974) Discharge patterns of single fibres in thepigeon auditory nerve. Brain Res 70: 431-447.Sachs, M.B. and J.M. Sinnott (1978) Responses to tones of single cells in nucleusmagnocellularis and nucleus angularis of the redwing blackbird (Agelaius phoeniceus). JComp Physiol A 126: 347-361.Schwarz, D.W.F. (1992) Can central auditory neurons reproduce sound waveforms? An analysisof the neurophonic potential in the laminar nucleus of the chicken. J Otolaryng 2 1(1): 30-38.Schwarz, D.W.F. (1992) Sound delay lines in the nucleus laminaris of the chicken. 3 Otolaryng21(3): 202-208.Schwarz, D.W.F., A. Dezsö and P.R. Neufeld (1993) Frequency selectivity of central auditoryneurons without inner ear. Acta Otolaryngol (Stockholm) 113: 266-270.Schwarz, D.W.F., I.E. Schwarz and A. Deszö (1994) Retrograde labeling of neurons in theauditory brainstem following injection of tritiated glycine into the inferior colliculus in thechicken, Gallus domesticus. 3 Otolaryng, submitted.Schwarz, D.W.F., I.E. Schwarz and A. Dezsö (1992) Cochlear efferent neurons projecting toboth ears in the chicken, Gallus domesticus. Hear Res 60(1): 110-114.Serafin, M., A. Khateb, C. de Waele, P.P. Vidal and M. MUhlethaler (1992) Medial vestibularnucleus in the guinea-pig: NMDA-induced oscillations. Exp Brain Res 88: 187-192.52Smith, D.J. and W. Rubel (1979) Organization and development of brain stem auditory nuclei ofthe chicken: dendritic gradients in nucleus laminaris. J Comp Neurol 186: 2 13-240.Smith, Z.D.J. (1981) Organization and development of brain stem auditory nuclei of the chicken:dendritic development in N. laminaris. J Comp Neurol 203: 309-333:Srulovicz, P. and J.L. Goldstein (1983) A central spectrum model: a synthesis of auditory-nervetiming and place cues in monaural communication of frequency spectrum. J Acoust SocAmer 73(4): 1266-1276.Steriade, M., M. Deschênes, L. Domich and C. Mulle (1985) Abolition of spindle oscillations inthalamic neurons disconnected from nucleus reticularis thalami. J Neurophys 54(6):1473-1497.Steriade, M. and R.R. Llinás (1988) The functional states of the thalamus and the associatedneuronal interplay. Physiological Rev 68: 649-74 1.Sullivan, W.E. and M. Konishi (1984) Segregation of stimulus phase and intensity coding in thecochlear nucleus of the barn owl. J Neurosci 4: 1787-1799.Takahashi, T. and M. Konishi (1983) Functional organization of the barn owl’s inferiorcolliculus. Soc Neurosci Abstr 9: 212.Takahashi, T., A. Moiseff and M. Konishi (1984) Time and intensity cues are processedindependently in the auditory system of the owl. J Neurosci 4: 178 1-1786.Takahashi, T.T. and M. Konishi (1988) Projections of the cochlear nuclei and nucleus laminaristo the inferior colliculus of the barn owl. 3 Comp Neurol 274: 190-211.Takahashi, T.T., H. Wagner and M. Konishi (1989) Role of commissural projections in therepresentation of bilateral auditory space in the barn owl’s inferior colliculus. J CompNeurol 281: 545-554.Tsuchitani, C. and J.C. Boudreau (1964) Wave activity in the superior olivary complex of the cat.3 Neurophys 27: 8 14-827.Volman, S.F. and M. Konishi (1989) Spatial selectivity and binaural responses in the inferiorcolliculus of the great horned owl. J Neurosci 9(9): 3083-96.Wagner, H. (1990) Receptive fields of neurons in the owl’s auditory brainstem changedynamically. Eur J Neurosci 2: 949-959.Wagner, H. (1992) On the ability of neurons in the barn owl’s inferior colliculus to sense briefappearances of interaural time difference. 3 Comp Physiol A 170(1): 3-11.Wagner, H., T.T. Takahashi and M. Konishi (1987) Representation of interaural time differencein the central nucleus of the barn owl’s inferior colliculus. 3 Neurosci 7(10): 3105-3116.Warchol, M.E. and P. Dallos (1989) Neural response to very low-frequency sound in the aviancochlear nucleus. J Comp Physiol A 166: 83-95.Webster, W.R. (1977) Chopper units recorded in the cochlear nucleus of the awake cat.Neurosci Lett 7: 261-265.Yin, T.C. and J.C.K. Chan (1988) “Neural mechanisms underlying interaural time sensitivity totones and noise.” In Auditory Function: Neurobiological Bases of Hearing, ed. G.M.Edelman, W.E. Gall and W.M. Cowan. pp. 385-430. New York: John Wiley and Sons,Inc.Young, E.D., J.-M. Robert and W.P. Schofner (1988) Regularity and latency of units in theventral cochlear nucleus: implications for unit classification and generation of responseproperties. J Neurophys 60: 1-29.53