Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Musical interval perception with pulsatile electrical stimulation of profoundly deaf ears Pijl, Sipke 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1994-894415.pdf [ 2.35MB ]
Metadata
JSON: 831-1.0099227.json
JSON-LD: 831-1.0099227-ld.json
RDF/XML (Pretty): 831-1.0099227-rdf.xml
RDF/JSON: 831-1.0099227-rdf.json
Turtle: 831-1.0099227-turtle.txt
N-Triples: 831-1.0099227-rdf-ntriples.txt
Original Record: 831-1.0099227-source.json
Full Text
831-1.0099227-fulltext.txt
Citation
831-1.0099227.ris

Full Text

MUSICAL INTERVAL PERCEPTION WITH PULSATILE ELECTRICALSTIMULATION OF PROFOUNDLY DEAF EARSbySIPKE PIJLB.Ed., The University of British Columbia, 1975M.A., Western Washington University, 1977A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIES(Neuroscience Program)We accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAFebruary 1994® Sipke Pijl, 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.__________________________Department of URO(t.LcThe University of British ColumbiaVancouver, CanadaDate 1 i99L.DE-6 (2/88)ABSTRACTThis research examines, in a musical context, themeasurement of pitches heard by Nucleus cochlear implantrecipients upon systematic variation of electrical pulserates, delivered to single intracochlear electrodes at acomfortable listening level. Stimuli were configured by acomputer in tandem with the Boys Town National InstituteInterface for psychophysical research with the Nucleuscochlear implant. Seventeen subjects participated in a 30-item tune recognition test (Experiment I). Many subjectsidentified a substantial number of items.Three subjects underwent a more detailed investigation todetermine whether pitches resulting from pulse rate variationwere sufficiently salient for musical interval perception.The results of a closed—set melody recognition test(Experiment II) suggested that recognition was possible on thebasis of melody, i.e., even in the complete absence ofrhythmical information, and that recognition was possible overa range of pulse rates. However, these results did notdetermine whether performance was based on ordinal propertiesof the pitches, or whether successive pitches definedidentifiable musical intervals.Intonation quality judgements (Experiment III) ofintervals ranging in size from a minor 3rd to a 5th providedevidence that the frequency ratios which characterizeacoustical musical intervals also apply to electrical pulserate pitch. Further evidence of musical ratio recognition wasobtained using the method of adjustment (Experiments IV andV). At least 2 out of 3 subjects were able, by means of theadjustment of a variable pulse rate, to reconstruct selectedmusical intervals abstracted from melodies well—known to thesubjects. Two subjects, furthermore, were able to transposethese melodic patterns to higher and lower pulse rates, in amanner similar to that demonstrated by normal—hearing subjectswhen listening to musical intervals.These results suggest that temporally mediated pitchesare capable of conveying ratio pitch information, in the sensethat equal ratios of pulse rates appear to produce equalmusical pitch intervals. These findings lend support totemporal theories of musical pitch and interval perception.iiTABLE OF CONTENTSABSTRACT iiTABLE OF CONTENTS iiiLIST OF TABLES VLIST OF FIGURES viLIST OF ABBREVIATIONS viiiDEFINITION OF TERMS ixACKNOWLEDGEMENTS XiiiDEDICATION xivINTRODUCTION 1METHODS: General 13Experimental Setup 18Preliminary Psychophysical Measurements 22Experimental Subjects 25Results 28Discussion 32EXPERIMENT I: Open-Set Tune Identification 38Methods 39Results 42Discussion 48iiiEXPERIMENT II: Closed-Set Melody Identification 55Methods 56Results 60Discussion 64EXPERIMENT III: Intonation Quality Judgements 69Methods 70Results 76Discussion 90EXPERIMENT IV: Musical Interval Reconstruction 100Methods 101Results 106Discussion 111EXPERIMENT V: Musical Interval Transposition 122Methods 122Results 128Discussion 143GENERAL DISCUSSION 152REFERENCES 171APPENDICES1. Open-Set Tunes: Experiment I 1842. Pulse Rates for 7-Note Melodies: Experiment II 187ivLIST OF TABLESTable Page1. Subject Data 232. Intonation Quality Judgements: points ofsubjective equality (PSE) and standarddeviations 783. Method of Adjustment: interval reconstructiondata 1074. Method of Adjustment: interval transpositiondata for Subject 7 1305. Method of Adjustment: interval transpositiondata for Subject 10 - ascending intervals 1346. Method of Adjustment: interval transpositiondata for Subject 11 - descending intervals 137vLIST OF FIGURESFigure Page1. Equipment set-up 212. Current amplitudes at comfortable loudness:effect of pulse width 293. Current amplitudes at comfortable loudnessfor 4 subjects 294. Current amplitudes at comfortable loudness:Subject 10 versus Subject 11 305. Current amplitudes at comfortable loudness:replicability of measurements 306. Open-set tune identification:percent correct scores 447. Open-set tune identification:Number of positive identifications ofeach tune 458. Tune identification performance versusspeech perception scores 479. Closed-set melody identification:percent correct scores 6110. Schematic representation of intonationquality judgements 7211. Intonation Quality Judgements: interval of a 5th 7912. Intonation Quality Judgements: minor 3rd 8013. Intonation Quality Judgements: interval of a 4th 8114. Intonation Quality Judgements: major 6th 8215. Point of subjective equality and standarddeviations for intervals from a minor 3rd toa major 6th 83viFigure Page16. Intonation Quality Judgements: effect ofelectrodes on interval of a 5th 8717. Intonation Quality Judgements: effect of pulserate of starting note on interval of a 5th 8918. Interval reconstruction: mean interval sizeand standard deviations 10919. Interval reconstruction: percentage ofadjustments 0-2 semitones from target 11220. Interval transposition: mean interval size andstandard deviations — intervals of a 5th andminor 3rd. Subject 7 13121. Interval transposition: mean interval size andstandard deviations - interval of a 4th.Subject 7 13222. Interval transposition: mean interval size andstandard deviations — ascending intervals.Subject 10 13523. Interval transposition: mean interval size andstandard deviations — descending intervals.Subject 10 13824. Interval transposition: differences in intervalsize related to adjustment of upper or lowernote 14025. Interval transposition: percentage ofadjustments 2 semitones or less from target 144viiLIST OF ABBREVIATIONSANOVA Analysis of VarianceANSI American National Standards InstituteBTNI Boys Town National InstituteE electrodeHz HertzkHz kiloHertzMHz MegaHertzmsec millisecondspps pulses per secondPSE points of subjective equalityS SubjectSD standard deviationsec microsecondsmicroampsviiiDEFINITION OF TERMS(Items arranged in logical sequence)Musical pitch: That aspect of pitch (related to chroma, notpitch height) that is capable of conveying musicalinterval information. The musical scale specifiesexactly the frequency ratios that are to be used toestablish musical intervals. These ratios are thus onthe stimulus side of the “equation”, and not on thephenomenal side of pitch. Musical intervalscharacterized by tones in equal frequency ratios arejudged to be subjectively equivalent by musicians, butnot necessarily or generally by others, except aslisteners acquire experience with music. For anylistener, subjectively equal musical intervals do notnecessarily represent equal nonmusical pitch (pitchheight) differences. The latter differences depend onthe relative position of the intervals on the frequencyscale.Musical scale: The full set of discrete musical intervalswhich are permitted. In the music of Western culture,these steps are defined by frequency ratios, and arederived from the octave as the basic interval. The unitof the Western musical scale is the semitone, which forthe equal temperament scale, is obtained by dividing theoctave into 12 equal frequency intervals. Pairs of tonesseparated by a given number of semitones (and a givenfrequency ratio) are given the same name, such as minor3rd, major 6th, and so forth, regardless of where theyoccur in the musical scale. The ratios used for all thecalculations in the experiments reported in this paperare based on the equal temperament scale.Musical intervals: Tones in one of a set of standardfrequency ratios which ideally characterize the frequencyrelationships between musical notes. Intervalscharacterized by identical frequency ratios are generally(at least for musicians) perceived as being musicallyequivalent. In the music of Western European culture,interval size is measured in semitoies. One semitoneequals one—twelfth of an octave (21/12), or a frequencyratio of approximately 1:1.059. The intervals commonlyreferred to in the experiments detailed in this paper,together with the accompanying frequency ratio and numberof semitones, are described briefly as follows:ixName of Interval Ratio qm -i tricsemitone 21/12 or 1:1.059 1major second 22/12 or 1:1.122 2minor third 23/12 or 1:1.189 3major third 24/12 or 1:1.257 4fourth 25/12 or 1:1.335 5tritone 26/12 or 1:1.414 6fifth 27/12 or 1:1.498 7major sixth 29/12 or 1:1.681 9octave 1:2 12Melody: For this paper, melody is defined as an orderedseries of musical intervals, in the absence of rhythmicalor phrase length information. The magnitude anddirection (i.e., ascending or descending) of suchintervals are precisely specified. While melodies playedin one frequency range may be transposed to otherfrequency ranges, melodies retain their identity onlywhen the sequence, the direction, and the magnitude ofthe intervals (i.e., the ratio relationships between thefrequencies of the notes) are preserved.Tune: For this paper, tune is arbitrarily defined as a melody(i.e., an ordered series of intervals) with a specifiedrhythmical pattern. The preservation of rhythmdistinguishes tune from melody, as used here, in thatwith melody, only the series of intervals (withoutrhythmical or phrase-length information) is available tothe listener.Melodic contour: A generalized form of interval information,and refers to the overall pattern or shape of a melody.Contour is defined by the directional relationshipsbetween temporally adjacent notes, without precisespecification of the magnitude of the pitch changes(i.e., interval size). Contour, therefore, indicateswhether adjacent notes in a melody are higher or lowerthan contiguous notes (i.e., the sequence of ups anddowns in pitch). For example, the contour of the firstphrase of “Twinkle, Twinkle, Little Star” could berepresented as= + = +where=,+, and — represent unisons, ascending, anddescending intervals, respectively. Melodies may shareidentical contours, but differ in the precise size of theintervals (e.g., “Twinkle, Twinkle, Little Star” versusthe Andante from Haydn’s Surprise Symphony).xMusical transposition: A procedure by which a melody or othermusical entity is shifted up or down on the musicalscale, with preservation of the constituent musicalintervals, as defined by the number of semitones (themathematical frequency relationships) between the notes.Transposed familiar melodies are readily recognized asmusical equivalents, even by musically unsophisticatedlisteners.Intonation: The adjustment of the frequency of one tone untilit is in some specified ratio relationship to a secondtone. The permissible ratios are defined by the musicalscale, and may represent a unison interval, a fifth, afourth, and octave, or any other desired interval. Whenmusicians listen for the precision of the frequencyadjustment (ratio properties) of musical intervals, theyare said to be performing intonation quality orintonation accuracy judgements. In judging musicalinterval size, musically unsophisticated listeners arethought to rely more on pitch height (the vertical axisof the pitch helix) than on tone chroma (the preciseposition of a tone within the octave) or the ratioproperties of the stimulus frequencies. The intonationquality judgements performed by such listeners must,therefore, provide a much cruder index of target intervalsize than when musicians perform such tasks (Burns andWard 1978).Pitch height: The helical model of pitch represents pitchheight as vertical dimension of the spiral, and chroma(C, G, etc.) by the angle of rotation around the spiral.One rotation represents a circle consisting of the 12notes of the chromatic scale. Thus, two C’s an octaveapart have the same chroma, but differ in pitch height,whereas a C and D adjacent to each other on the musicalscale are similar in pitch height, but differ in chroma.A continuous increase in frequency will, according tothis representation, be heard as a continuous increase inpitch height (a nonmusical aspect of the helix model),along with repetitions of tone chroma as each octave istraversed. Pitch height is thought to be an importantfactor governing the perceived relations among tones whenthey are presented outside a musical context or whentones are presented to musically unsophisticatedlisteners. Chroma is thought to be relatively moreimportant to musically sophisticated listeners or whentones are presented in a musical context (Krumhansl andShepard 1979).xiClosed-set: A forced-choice test paradigm, in which subjectsare provided with a limited set of stimulus—responsealternatives.Open-set: A test paradigm in which subjects are not providedwith any specific response alternatives.xiiACKNOWLEDGMENTSThe preparation of this work was greatly facilitated bythe help and encouragement of the following persons, each ofwhom provided a unique and essential contribution to itscompletion:I am grateful to Dr. D.W.F. Schwarz, for serving as mydissertation advisor and Committee Chairman, for hisencouragement and valuable discussions regarding this project,and for helpful comments on a preliminary version of thispaper. Many thanks to Professor D. Greenwood, for his manyconstructive comments on an earlier draft of this thesis.I am indebted to Dr. P.J. Doyle, Professor and Head (Emeritus)of the Division of Otolaryngology at the University of BritishColumbia, for encouraging me to undertake this project.My deepest gratitude to Mrs. J.H. Mitchell, President ofthe John Hardie Mitchell Family Foundation, for her generosityin providing the funding essential for the completion of thisproject.A very special thanks to all the cochlear implantpatients who participated in this work. Especially, I amgrateful to the three patients who contributed tirelessly oftheir time, effort and energies. In particular, I thank PeterLaschitz, who spent many hours in our laboratory, evaluatingand adjusting musical intervals, and participating ininnumerable pilot experiments. Also, thanks to Lewis Dronfeldand Sonja Reid. Each of them provided some unique insightsinto the pitches perceived with their cochlear implant.Thanks to Dr. Bob Shannon, who devised and implementedthe interface used for this research and provided us withsample software routines.My sincere appreciation to my son Michael, for hisextensive technical assistance and for writing the computerprograms which formed the backbone of this work.My deepest appreciation to my wife Lois, for her patienceand emotional support throughout the lengthy preparation ofthis work, and for her diligence in overcoming numerousobstacles in the preparation of the figures and the final copyof the manuscript.Thanks to Marion Hawryzki for proofreading themanuscript. Thanks also to my daughter Emily, my family andfriends for their patience, understanding and support.xiiiDedicated to the memory of my FatherWho sparked my love for musicxiv1MUSICAL INTERVAL PERCEPTION WITH PULSATILE ELECTRICALSTIMULATION OF PROFOUNDLY DEAF EARSINTRODUCTIONThe relative contributions of temporal and cochlear placemechanisms in the perception of pitch have been debated forwell over 100 years. Place coding of frequency refers to adifferential distribution of activity across the neuronalarray, as a function of tone frequency. In the normal ear,this results from the mechanical and spatial analysis of thewaveform along the cochlear partition. Temporal coding offrequency results from the synchronization of auditory nervefiber responses to the mechanical oscillations of the basilarmembrane and its associated structures.The place theory of pitch encoding was advanced in themid-nineteenth century by Helmholtz. Helmholtz found supportfor his theory in three, then recent, developments in auditoryscience. Helmholtz theorized that the rods of Corti,discovered by Corti in 1851, formed a series of finely tunedresonators, responsive to progressively lower frequencies in abasal to apical direction. In a later publication, Helmholtzasserted that the transverse fibers of the basilar membrane,rather than the rods of Corti, were the actual resonatingelements. The length, tension, and mass of these fibers werethought to vary from base to apex. A second influence wasOhm’s theory of complex sounds, which stated, in essence, that2any complex periodic sound could be specified as the sum of aseries of sine wave components in appropriate amplitude andphase relationships. A third major influence was Muller’sdoctrine of the specific energy of nerves. Helmholtz extendedthis doctrine, theorizing that each resolvable frequencycomponent within a stimulus would be associated with aspecific cochlear place and corresponding neuron. Such aneuron was thought to respond with an intensity—dependentincrease in discharge rate. Thus, the excitation of specificneurons was thought to be associated with specific pitches.With the subsequent work of von Bekesy, beginning in1928, the tuned—resonance concept was largely replaced by thetraveling wave theory, which incorporates the notion ofbasilar membrane motion rather than sympathetic resonance ofindividual transverse fibers. Bekesy’s observations showedthat movement of the stapes resulted in displacements of thecochlear fluids, with resultant displacements of the scalamedia alternately towards the scala tympani and the scalavestibuli. Thus, the vibratory motion of the stapes initiateda traveling wave along the basilar membrane, moving from baseto apex, with a progressive phase lag at successively apicallocations. The locus of maximum displacement and theamplitude characteristics of the traveling wave were shown tobe determined by the mechanical properties of the cochlearpartition, of which the stiffness was found to decrease, andthe width to increase, from base to apex. More recentphysiological data have shown that the mechanical responses in3the intact cochlea are highly sensitive, and more sharplytuned than found by Bekesy in cadaver ears, and that theseresponses may be aided by active processes.The traveling wave is known to reach a maximum amplitudeat a locus consistent with the mechanical properties of themembrane at that location. Thus, for high frequencies, themaximum amplitude is realized within the basal turn, whereasfor low frequencies, the maximum amplitude of displacementoccurs more apically. The locus of the population of cochlearneurons activated at low stimulus intensities is therefore astrong function of the stimulating frequency. Individualneurons, therefore, possess frequency selectivity (and a“characteristic frequency” or CF, to which the response occursat the lowest intensity) by virtue of their unique positionalong the cochlear partition. High frequency fibers innervatethe basal turn, and low frequency fibers the more apicalregions. While a low-intensity pure tone stimulus elicitsincreased activity in a small population of cochlear neurons,increased intensity of the stimulus results in a spread ofneural excitation to the increasing effective displacement ofthe basal portion of the displacement envelope. This resultsin a recruitment of neurons with higher characteristicfrequencies.The “tonotopic” frequency organization, established inthe cochlea, is maintained in the cochlear nuclei and athigher levels of the auditory system. This place-specificfrequency analysis is therefore one mechanism by which4frequency information can be transmitted to the centralauditory system. However, demonstrations that the pitch ofcomplex tones does not depend on the presence of thefundamental frequency (Fletcher, 1928; Schouten, 1938;Licklider, 1954) have presented a serious obstacle for simpleplace pitch theories (reviewed in Greenberg, 1980; Sachs andMiller, 1985)The second type of frequency information available to thebrain is encoded within the temporal structure of neuralresponses. The temporal or periodicity theory was advanced bySeebeck in 1841, by Wundt in 1880, and later refined byRutherford in 1886 (reviewed in Plomp 1976). In these earlyconceptualizations, the acoustic waveform was thought toundergo little, if any, modification in the cochlea. Thus,the pitch of a tone was thought to be based on the frequencyor rate of neural discharge, rather than on the identity of aparticular set of active nerve fibers. These early temporaltheories were limited in their ability to account for encodingof frequencies greater than several hundred Hertz, as theupper limit of firing for individual neurons is considered tobe less than 500 spikes per second. However, the recording byWever and Bray, in 1930, of frequencies as high as 5 kHz, withmacroelectrodes in the auditory nerve, led to the initialsuggestion that this potential might represent the aggregateresponse of a population of nerve fibers, each locked in phaseto the stimulating frequency. Although subsequent studieshave shown that the cochlear microphonic contributed to the5recordings of Wever and Bray, these findings nevertheless ledto a revival of interest in the role of temporal mechanisms inthe perception of pitch.A mechanism by which the pitch of complex tones could beencoded in the temporal responses of auditory nerve fibers wasproposed by Schouten (1940). In this scheme, the pitch ofcomplex tones resulted from the preservation of theperiodicity of the stimulus waveform in the neural responses,because of the interaction of high, unresolved harmonics inthe basal region of the cochlea. Further refinements ofSchouten’s proposal, suggesting that neural responses occurredon the high-amplitude peaks of the fine structure of thewaveform, were necessitated by evidence of pitch shifts, whichwere found to occur when the components of a complex tone werefrequency—shifted by an equal amount, even though the waveformperiodicity remained unchanged. Evidence regarding therelatively greater importance of low order harmonics (Ritsma1967) in the perception of pitch (i.e., those which arerelatively more resolved in the cochlea), has necessitatedfurther revisions of temporal theories, such as contributionsfrom units tuned to frequencies midway between contiguousresolved components (Javel 1980). In spite of a wealth ofphysiological and psychoacoustical data, the precise roles andcontributions of temporal and place mechanisms in theperception of pitch remain poorly understood. We shall nowconsider some of the evidence for the importance of temporalinformation.6The precision of phase-locking and the preservation ofthe temporal features of the stimulus waveform in theresponses of auditory nerve fibers (Rose, Brugge, Anderson,and Hind 1967; Javel 1980; Sachs and Young 1980; Greenberg,Geisler, and Deng 1986), as well as the persistence of thesefeatures into the auditory brainstem (Langner 1983; Takahashi,Moiseff, and Konishi 1984; Sullivan and Konishi 1986; Langnerand Schreiner 1988; Schreiner and Langner 1988), suggest thattemporal information may play an important role in providingcues regarding stimulus frequency (Javel, McGee, Horst, andFarley 1988). However, there exists, in addition, cogentpsychoacoustical evidence regarding pitch shifts which cannoteasily be explained on the basis of temporal mechanisms (Daviset al. 1950)A variety of data from normal—hearing subjects havesupported the existence of temporally based pitch. Stimuliconsisting solely of high, unresolvable harmonics have beenfound to evoke residue pitches corresponding to the absentfundamental (Moore and Rosen 1979; Houtsma and Smurzynski1990). Additional evidence for the importance of temporalinformation in pitch perception has been obtained fromexperiments utilizing amplitude-modulated white noise (Millerand Taylor 1948; Small 1955; Harris 1963; Pollack 1969;Patterson and Johnson—Davies 1977). Burns and Viemeister(1976, 1981) demonstrated that the pitches evoked bysinusoidally amplitude-modulated noise were sufficientlysalient to enable subjects to recognize simple melodies and7musical intervals, when the constituent “notes” correspondedto low modulation frequencies.Further evidence for temporal mechanisms in pitchperception can be found in time separation and phase—shiftphenomena. Cramer and Huggins (1958), for example, showedthat dichotic white noise, of which a small frequency regionpresented to one ear had been phase—shifted, resulted in theperception of a faint pitch which corresponded to thefrequency of the phase transition. Others (Small andMcClellan 1963; Warren and Bashford 1988) have shown that whena pulse train is complemented with a delayed replica ofitself, a pitch is heard which corresponds to the reciprocalof the delay (t) between the two pulse trains. Pitch matchesbetween periodic all-positive polarity pulses and periodicpatterns of positive and negative polarity pulses (Flanaganand Guttman 1960; Warren and Bashford 1988) further supportthe existence of pitches which are based on exclusivelytemporal mechanisms, at low pulse rates (below 200 pps). Athigher pulse rates, pitch matches were made on the basis offundamental frequency.Similarly, Pierce (1991) showed that, at low repetitionrates (up to about 250 Hz for bursts of a 4978 Hz tone),sequences of tone bursts with the same sign and sequences oftone bursts with alternating sign were perceptuallyindistinguishable, in spite of differences in fundamentalfrequency and harmonic spacing. The pitches resulting fromthese stimuli, at low repetition rates, have been shown to be8sufficiently salient to convey musical interval information(Pierce 1991)With acoustical stimuli and normal-hearing listeners, ithas been difficult to devise experiments that unequivocallydissociate place and timing information (Pierce, Lipes, andCheetham 1977). When noise is added to itself following adelay, the power spectrum displays peaks at 1/t Hz, and atintegral multiples of 1/t Hz. Although interrupted oramplitude—modulated white noise has an essentially flat long-term spectrum, fluctuations in the short—term spectrum maypermit limited spectral cues related to modulation rate which,without appropriate precautions (filtering or masking), couldcontribute to pitch perception (Pierce, Lipes, and Cheetham1977). While these alterations in the statistical propertiesof the noise are thought to be small, especially withsinusoidal modulation, and of limited significance indemonstrations of temporally based pitch (Moore and Glasberg1986), these possibilities introduce uncertainties regardingthe utilization of temporal features. Furthermore, even inthe absence of spectral peaks and valleys in the acousticstimuli themselves, experimental findings could becontaminated by the presence of distortion products generatedwithin the ear. Horst, Javel, and Farley (1990) have shown,in the phase—locked responses of auditory nerve fibers, thepresence of nonlinearities which introduce distortion productsat integer multiples of the signal frequencies and atfrequencies lower than those contained in the signal.9With direct electrical stimulation of residual auditoryneurons in deaf subjects, however, there is no spectralanalysis of the signal (Kiang and Moxon 1972; Merzenich et al.1973; Shannon 1983; Javel, Tong, Shepherd, and Clark 1987; vanden Honert and Stypulkowski 1987a), and the locus of neuronalstimulation is dependent exclusively on the distribution ofcurrent relative to the residual neuronal array. For bipolarstimulation, this distribution is a complex function of thebipolar electrode location, the electroanatoiny of the residualcochlear apparatus, and of current amplitude and pulseduration (van den Honert and Stypulkowski l987a). The tightrelationship between the temporal characteristics of theelectrical stimulating waveform and the neuronal response(Hartmann, Topp, and Klinke 1984; Javel et al. 1987; Parkins1989) suggest that electrically stimulated subjects present anunique opportunity to investigate periodicity-related pitchmechanisms.Although a substantial body of literature has accumulatedover the past 30 years supporting the presence of rate—basedpitch effects with electrical stimulation (Eddington, Dobelle,Brackmann, Mladejovsky, and Parkin l978a, 1978b; Simmons etal. 1979, 1981; Tong, Clark, Blamey, Busby, and Dowell 1982;Tong and Clark 1983, 1985; Hochmair-Desoyer, Hochmair, Burian,and Stigibrunner 1983; Shannon 1983; Eddington and Orth 1985;Pfingst 1985; Townshend, Cotter, Van Compernolle, and White1987; Shallop, Beiter, Coin, and Mischke 1990), thequalitative aspects of these pitch percepts and their10relevance to communication have rarely been addressed (Muller1983; Pfingst 1985; Schubert 1983). Most studies ofelectrical rate pitch have employed conventionalpsychophysical procedures such as difference limen assessment,pitch matching, and magnitude estimation of pitch on anarbitrary scale. While these procedures have been useful indocumenting some elementary relationships between electricalfrequency and subjective pitch, they do not place pitch withina meaningful context. Even with normal—hearing subjects,pitch scales resulting from approaches such as magnitudeestimation, interval equisection and quasi—fractionation (asemployed by Stevens and Volkmann in 1940), have been shown tobe of little relevance to pitch perception in the musicalsense (Ward 1970). Psychophysical studies in normal-hearingsubjects have also failed to demonstrate a correlation betweenthe size of the just-noticeable-difference for pitch of puretones and the ability to learn or reproduce simple tunes (Wing1948)Although a number of experiments between 1966 and 1979suggested that limited recognition of common melodies could beachieved with electrical stimulation of ears withoutfunctional receptor cells (Simmons 1966; Eddington et al.1978a, 1978b; Chouard 1978; Moore and Rosen 1979), otherinvestigators have failed to confirm these findings (Gfellerand Lansing 1991). Thus, while it appears clearly establishedthat pitch increases systematically with electrical frequencyor pulse rate up to a maximum of 300 to 1000 Hz, it is not11clear whether these pitches permit the perception of melody orthe recognition of musical intervals, as defined by electricalfrequency or pulse rate ratios.In the investigations which follow, Experiment I exploredthe ability of 17 musically unsophisticated cochlear implant(Nucleus) recipients to identify common tunes “played” bysystematic variation of pulse rate on single apicalintracochlear electrodes, using low pulse rates. ExperimentsIl-V employed three musically unsophisticated Nucleus implantrecipients in a more detailed examination of the musical pitchand interval properties of electrical pulse rate pitches.Experiment II assessed the ability of these subjects toidentify melodies from a closed set, in the absence ofrhythmical and melody-length information. These briefmelodies were played on electrodes in a variety of locationsalong the electrode array, and over a range of pulse rates.Experiments III to V examined the relationships between thesizes of melodic pitch intervals for electrical pulse ratepitches and the frequency ratios which characterize acousticalmusical intervals. In Experiment III, this was accomplishedby means of intonation quality judgements, where the subjectsrated a randomized series of pitch intervals as “flat”, “intune”, or “sharp”, relative to their memory of a specificmusical interval from a well—known melody. In Experiments IVand V, the method of adjustment was employed to assess theability of the same subjects to reconstruct and transposemusical intervals resulting from changes in pulse rate onsingle intracochlear electrodes.1213METHODS: GENERALINTRODUCTIONPsychophysical results with cochlear implant patientshave shown that there are several variables which influencethe pitch of electrical stimuli. Amongst the most widelyinvestigated are those which pertain to the locus ofstimulation (electrode position), and to the signal frequencyor repetition rate. These pitchlike percepts are commonlyreferred to as “place pitch” and “rate pitch”, respectively.Many studies have confirmed that gradations of pitchli]ceor timbrelike percepts result from electrical stimulation indifferent regions of the cochlea. These sensations have beenfound to vary in accordance with the tonotopic organization ofthe cochlea (Eddington et al. 1978a; Tong, Millar, Clark,Martin, Busby, and Patrick 1980; Simmons et al. 1981; Shannon1983; Townshend et al. 1987). Electrode position may, inaddition, interact in a complex fashion with pulse rate orfrequency. Shallop et al. (1990), for example, reported thatincreases in pulse rate had a much greater effect on pitchestimates for apically than for basally situated electrodes.Shannon (1983) reported pitch estimates consistent withincreases in electrical sinusoidal frequency for both basaland apical electrodes, although the asymptotic pitch estimatesremained higher for basal than for apical electrodes.14Stimulus waveform and mode of stimulation (monopolar orbipolar) have also been reported to influence perceived pitch.Shannon (1983) found the expected pattern (of higher pitchestimates upon stimulation of basal electrodes, and lowerpitch estimates for apical electrodes) to be quite specific tothe mode of stimulation.Pitchlike phenomena related to the rate rather than theplace of stimulation have also been described extensively(Merzenich et al. 1973; Hochmair-Desoyer et al. 1983;Shannon 1983; Tong and Clark 1983; Townshend et al. 1987;Shallop et al. 1990). However, it is frequently difficult todifferentiate pitch-related phenomena from those related totimbre. Simmons (1966), for example, reported thatstimulation at rates greater than 50 Hz (both pulsatile andsinusoidal) resulted in auditory percepts which were describedas “rough”, “rattling”, and “crackling”, whereas stimuli above100 Hz were described as “smooth” and “steady”. Between 20 Hzand 100 Hz, the pitch changes were reported to be small, andoccasionally, nonexistent. For repetition rates between 100and 400 Hz, pitch rankings were consistently in the expecteddirection. Further increases in rate, beyond 400 Hz, yieldedanecdotal descriptions corresponding to pitches normallyevoked by acoustic stimulation with 2—4 kHz sounds, or nochange in pitch. Similar rapid increases in pitch withincreases in frequency or repetition rate have been reportedby others (Tong, Black, Clark, Forster, Millar, Q’Loughlin,15and Patrick 1979; Shannon 1983; Tong, Blarney, Dowell, andClark 1983).However, electrical pulse rate or frequency and electrodeposition are not the only stimulus parameters which have aneffect on pitch or timbre. Variation of pitch and timbre havealso been reported with changes in stimulus amplitude (Shannon1983). Simmons et al. (1979, 1981), for example, reportedanecdotally that a low—intensity stimulus on a given electrodecould sound like “a knock on a large sheet of plywood”,whereas a high—intensity stimulus on the same electroderesembled “a small triangle being struck”. These changes werequite specific to certain electrodes, while on otherelectrodes, increases in amplitude resulted in only loudnessincreases, without concomitant changes in pitch or otherperceptual attributes. Asymmetrical spread of current orspatial asymmetry of surviving neurons relative to thestimulating electrodes have been considered as potentialexplanations for these phenomena.The potential effects of stimulus amplitude on pitch ortimbrelike percepts necessitate a further consideration ofloudness. With electrical stimulation of hearing, the primaryphysical determinant of loudness is the amount of chargedelivered per unit time (Eddington et al. l978a; Tong et al.1983). When pulse width is increased, the amount of chargedelivered and the amount of time available to charge thecapacitance of the neuronal membranes are also increased.Thus, stimulus amplitude, pulse width, and stimulus duration16have been shown to be the parameters which have the greatesteffect on psychophysical detection thresholds and on perceivedloudness (Eddington et al. 1978a; Pfingst, DeHaan, andHolloway 1991; Shannon 1992). Loudness has also been shown tobe an important factor in the performance of cochlear implantpatients on psychophysical tasks such as the measurement ofgap detection thresholds and temporal difference limens, inthat smaller temporal differences can be appreciated at higherintensity levels (Shannon 1989; Tyler, Moore, and Kuk 1989).Loudness judgements may, furthermore, be complicated bypsychological factors such as “pleasantness”, “noisiness”,“shrillness”, as well as by the probable unnaturalness ofauditory sensations that result from electrical stimulation.The lower and upper limits of the operating range ofcochlear implant electrode pairs are generally defined by thethresholds of detection and of loudness discomfort,respectively (Pfingst 1984). Between these two extremes liesthe dynamic range for electrical stimulation. The amount ofcurrent required to generate hearing sensations in profoundlydeaf ears has been shown to vary considerably between patientsand electrode configurations, and to correlate negatively withboth nerve fiber survival and proximity of the electrodes tothe surviving nerve fiber population (Pfingst and Sutton1983)In spite of extensive verification of the generality of“place” and “rate” pitchlike effects with electricalstimulation of hearing, the qualitative aspects of these17phenomena remain insufficiently characterized. Part of thedifficulty, no doubt, lies in the elusive subjectivity andcomplexity of the pitch percept, as well as in inadequatedefinitions of what is meant by pitch. Pitch is broadlydefined by the American National Standards Institute (ANSIS3.20-1973) as “that attribute of auditory sensation in termsof which sounds may be ordered on a scale from low to high.”This definition, however, addresses solely the ordinalproperties of pitch. Furthermore, many implanted subjectshave little sophistication in describing hearing sensations,especially following prolonged auditory deprivation.Further complicating the issue is the ambiguity of the“high” and “low” descriptors generally applied to pitch.These descriptors, in common usage, are frequently appliedalso to loudness and timbre. These attributes, like pitch,can also be scaled, matched, labeled, and ranked from low tohigh. It is clear that conventional psychophysical tasks suchas pitch matching, pitch ranking, and magnitude estimation areinadequate for the dissociation of pitch and timbre. Pitch,in the musical sense, implies more than the mere ability ofthe listener to make “higher” versus “lower” judgeinents(ordinal pitch). Musical pitch, in addition, implies thatmusically intelligent listeners should be able to assign atone or stimulus to a position relative to another tone on amusical scale. In short, such listeners should be able toidentify, label, or adjust musical intervals. Houtsma (1984)has argued that any useful definition of pitch should be based18on the ratio properties of the pitch percept. For the purposeof this paper, the operational definition of pitch will bethat adopted by Burns and Viemeister (1981), i.e., that aspectof auditory sensation which carries melodic information.While loudness equalization is of paramount importance inthe accurate measurement of pitch difference limens (Pfingst1988), it is not certain whether exhaustive loudnessequalization is essential in melody recognition and musicalinterval adjustment tasks. To preclude the availability ofgross loudness cues in the experiments which follow, stimuliwere drawn from a continuum of pulse rates which wereequalized for loudness. While a procedure which randomizedstimulus amplitude within the dynamic range of individualsubjects might have been more desirable, this was precluded bysoftware limitations, as well as by potential interactionsbetween stimulus intensity, timbre, and pitch. A procedurewas therefore devised in which subjects, using the method ofadjustment at a number of particular pulse rates, establishedthe current amplitudes required for a “comfortable” (definedas neither too loud for extended listening, nor too soft so asto be difficult to hear) and equal loudness over the range ofpulse rates tested.EXPERIMENTAL SETUPThe Nucleus Ltd./Cochlear Corporation cochlear implantdelivers a pulsatile electrical signal via an array of 2219electrodes inserted into the scala tympani (Blarney, Dowell,Clark, and Seligman 1987). The electrode array (Model C122M)consists of 32 platinum bands spaced 0.75 mm apart, situatedon the distal 25 mm of a silastic carrier. The 22 distalmostbands (occupying the distalmost 17 mm of the array) serve asthe functioning electrodes. Typical insertion distancesresult in the delivery of current pulses to cochlear regionswith characteristic frequencies of 1-16 kHz (Greenwood 1961,1990). The electrodes are arbitrarily numbered from 1-22 in abasal to apical direction, with the electrode number referringto the basal member of a bipolar electrode pair(active/ground). These bands are independently connected toand driven by the receiver—stimulator package, which issurgically embedded in the mastoid bone, and consists of ahermetically sealed titanium capsule. The receiver—stimulatorcontains a Complementary Metal Oxide Semiconductor (CMOS)integrated circuit which decodes the signal and routes theinformation to selected electrodes. The data streamoriginating from either the patient’s wearable speechprocessor or from an appropriate test device is transmittedtranscutaneously to the implanted receiver—stimulator packageby means of an externally worn transmitter coil, using a radiofrequency (2.5 MHz) signal.The clinical test device (Dual Processor Interface) andsoftware (DPS659A) used in the routine psychophysicalassessment and programming of the processor worn by thepatient does not permit detailed control over stimulus20parameters. Therefore, for the purpose of these experiments,the stimulus pulses delivered to the implantedReceiver/Stimulator were configured by a special computerInterface for Psychophysical Research with the NucleusCochlear Implant, designed and implemented at Boys TownNational Institute (BTNI). The technical specifications ofthis interface have been detailed elsewhere (Shannon, Adams,Ferrel, Palumbo, and Grandgenett 1990). The BTNI interfaceused in these experiments was an outboard version, external tothe host 386 computer. The interface was connected to thecomputer via the parallel printer port. The host computertransmitted a specified byte stream to the interface, and theinterface then generated the appropriate burst sequence fortransmission to the internal receiver/stimulator (Figure 1).Pulse rate, amplitude, and waveform parameters generatedby the experimental setup (the host computer and the BTNIinterface, and the transmission cable/coil) were verified by atest system consisting of a Receiver/Stimulator unit (Implant-in—a—Box, consisting of a Nucleus C122M internal deviceidentical to that implanted into the experimental subjects, ina plastic enclosure) and a Tektronix 2232 100 MHz DigitalStorage Oscilloscope.The sample software routines provided with the BTNIinterface were rewritten and expanded, using Pascal, in orderto permit the delivery of sequences of pulse trains atspecified rates and amplitudes. Pulse parameters wereindividualized for each subject, so that amplitudes were21Figure 1. Equipment setup with the Boys Town NationalInstitute interface for psychophysical research with theNucleus cochlear implant.TransmitterCoil22within the linear operating range of the device, and aredetailed in Table 1. The purpose of the preliminarypsychophysical measurements was to establish an equal andcomfortable loudness level over a range of pulse rates onselected electrodes. These amplitude values were thenaccessed by all subsequent routines, which permitted thedelivery of either continuously variable pulse rates, orspecific sequences of pulse rates organized into tunes,melodies or musical intervals. All randomization of stimuliand scoring procedures were performed automatically by thecomputer. Subjects were provided with written instructionsfor each of the experiments. For all the experimentalprocedures which follow, the subjects were connected directly,via a standard tricord cable and external transmitter coil, tothe BTNI Interface.PRELIMINARY PSYCHOPHYSICAL MEASUREMENTSThe method of adjustment was used to establish acomfortable listening level on selected electrodes, forapproximately a 4—octave range of pulse rates, from 54 to 1096pps. Spatial separation of active and ground electrodes andpulse widths were individualized for all subjects, due todifferences in current requirements of individual subjects(Shannon 1989; Blarney, Pyman, Clark, Dowell, Gordon, Brown,and Hollow 1992). These parameters, once established for eachsubject, were maintained constant throughout the remainder of23TABLE 1SUBJECT DATA*AGE DUR. YRS PW ELEC.S AGE ONSET DEAF. EXP. ETIOLOGY MUS E jsec SEPAR. %1 68 15 1 yr 1 Unknown 0 12 100 1.50 732 59 39 1 yr 2 Unknown 2 18 205 2.25 893 54 32 4 yrs 5 Labyrinthitis 2 18 150 .75 834 35 32 2 yrs 2 Unknown 1 20 150 1.50 645 40 34 1 yr 5 Trauma 0 19 100 .75 966 37 19 2 yrs .5 Unknown 0 18 100 2.25 747 35 30 2 yrs 2 Unknown 0 18 150 1.50 918 58 48 9 yrs 1 Trauma 0 18 100 1.50 889 44 20 24 yrs 4 Trauma 0 20 205 1.50 2310 70 8 1 yr 2 Otitis Media 2 18 250 2.25 8811 43 37 4 yrs 1 Meningitis 2 18 100 2.25 8412 77 57 1 yr .5 Unknown 0 18 100 1.50 5513 33 16 13 yrs 5 Meningitis 2 18 150 2.25 8214 38 6 9 yrs 2 Unknown 2 19 100 2.25 7715 58 7 5 yrs 5 Otoscierosis 2 14 250 2.25 5016 64 42 20 yrs 5 Unknown 0 20 150 1.50 6817 50 46 1 yr 3 Unknown 2 20 150 1.50 88*Legend:S: Subject.Age: Age in years at time of experiments.Age Onset: Age in years at onset of hearing loss.Dur. Deaf.: Duration of total deafness priorimplantation.Yrs Exp: Duration of experience with cochlear implant.Mus: Self-rated musical ability prior to deafness:0=No interest in music; 1=Interest inlistening to music; 2=Played a musicalinstrument.E: Basal member of electrode pair used for testing.Electrodes are numbered in a basal to apicaldirection, from 1-22.PW sec: Duration (in microseconds) of each phaseof biphasic current pulses. Interpulseinterval 40 tsec.Elec Separ: Distance in millimeters between electrodepairs.%: Average keyword correct score on CID EverydaySentences, Iowa Sentences, and BKB Sentences(Cochlear Corporation, Tape RecordedVersion).24the experiments (Table 1). The interval between the negativeand positive polarity portions of the biphasic waveform was 40jsec for all subjects. This value of interpulse interval wasarbitrarily selected merely because it was the interpulseinterval used for clinical applications of the device. Trainsof pulse rates were presented at a rate of 1 per second, witha 50% duty cycle. The relatively long duration (500 msec) wasintended to permit subjects sufficient time to estimate pitch,and was consistent with stimulus duration in acousticalstudies requiring musical pitch judgements. Subjects wereinstructed to adjust the amplitude of the signal (using the“up” or “down” arrow keys of the computer keyboard) to acomfortable loudness. They were instructed to do this bybracketing, i.e., by making the sound alternately too loud andtoo soft before deciding on a comfortable loudness level. Thesubjects were instructed to press the <Enter> key uponcompletion of the trial for a given pulse rate. The computerthen logged the appropriate amplitude stepnumber for thatpulse rate, and automatically proceeded to test the nexthigher pulse rate. Amplitude stepnumbers were converted tomicroamperes by means of a table, stored in computer memory,of actual measurements made by the manufacturer on each deviceprior to implantation. A total of 16 pulse rates were tested,increasing by an arbitrary factor of 1.22 from 54 to 1096 pps.The entire set of rates was then replayed, both in sequenceand in random order, and inequalities in loudness eliminated25by further adjustment, until all the pulse rates in thesequence were judged to be equally loud.EXPERIMENTAL SUBJECTSThe small number of local cochlear implant recipientsprecluded recruitment of subjects with documented musicalcompetence. However, since performance of cochlear implantrecipients on speech recognition tasks is known to varywidely, it was conceivable that optimal results would notnecessarily be obtained from the most musical subjects.Fourteen unpaid volunteer subjects who were returning to theLaboratory for routine follow-up testing participated in theopen—set tune recognition test. Subjects were questionedabout their interest and participation in musical activities,prior to deafness. Details regarding subjects are presentedin Table 1. All subjects (except S-9) were reported, atsurgery, to have a minimum of 22 intracochlear electrodes.Three subjects (S-7, S-b, S-il) participated inExperiments 11—V. These subjects were paid, and were notrecruited on the basis of their musical ability. None of thesubjects understood the concept of an octave. These subjectswere readily available for the experiments, and had a generalinterest in the project. All three subjects had excellentspeech perception results with the implant. A brief musicalhistory of these three subjects is detailed below.26SUBJECT 7 (S-7)This 35 year old male reported no particular interest inmusic prior to deafness. He was, however, familiar with avariety of common tunes. Although he had never played amusical instrument, sung, or participated in any significantmusical activity, he was nevertheless a highly motivated andcommitted subject, with an interest in exploring hisperceptual capabilities.SUBJECT 10 (S—b)This 70 year old male had sung as a cathedral choristerbetween the ages of 8 and 15 years, and was able to recallboth the texts and tunes of a variety of classical choralselections. His father had been a semiprofessional tubaplayer. Subject 10 had received a few violin lessons as achild, but had not played any musical instruments sincechildhood. In spite of his gradually worsening hearing sincethe age of 8 years, he reported having had an intense interestin classical music during most of his adult life, andparticipated extensively in ballroom dancing activities.Postoperatively, he indicated little or no enjoyment of themelodic and harmonic aspects of music, but was able tocontinue dancing. Subject 10 was conversant with the terms“sharp” and “flat” as they relate to interval size, andapplied these terms voluntarily and readily to musicalintervals played at low pulse rates. For example, whenadjusting the lower note of a musical interval with a fixed27upper note, he readily described the interval as soundingprogressively “sharp”, even though it was the lower note ofthe interval which he had adjusted “flat”.The electrode array for this subject was located in thescala vestibuli. Extensive osteoneogenesis following theremoval of an identical previous cochlear implant from thescala tympani precluded reinsertion via the round window. Theremoval of the first device was necessitated by postoperativeinfection. Except for the higher current requirements withthe scala vestibuli device, audiological results werecomparable to those achieved previously in the same patientwith the scala tympani implant (Pijl and Noel 1992).SUBJECT 11 (8-11)This 43 year old female reported a considerable interestin singing prior to the onset of deafness. Although she didnot read music or play any musical instruments, she reportedhaving been able to play, by ear, simple tunes, with a degreeof accuracy, on a piano keyboard. She was familiar with awide variety of common tunes. With her cochlear implant, shereported little or no enjoyment of music, and indicated onlyoccasional recognition, following a large number ofrepetitions, of tunes played on musical instruments by herchildren. She felt that her occasional recognition of commontunes in everyday situations was based primarily on rhythmicalcues.28RESULTSThe results of the preliminary loudness equalizationprocedures for several subjects are shown in Figures 2 to 5.The current levels required to generate comfortable hearingsensations ranged from approximately 400 to 1600 )LA, dependingon the subject, the spatial separation between the active andground electrodes, and the pulse width employed. Directcomparison of current amplitudes for different subjects isprecluded by intersubject differences in stimulationparameters (Table 1). Reductions in pulse width resulted inmarked elevations in the amplitudes required for comfortablehearing levels, when the active and return electrodes werekept constant (Figure 2). For most subjects, the amplitudefunction showed a decrease as pulse rates were increased(Figures 3 and 4). The slope of this function was steep forsome subjects (S—3, S—b), especially at low pulse rates, andshallow for others (S-7, S-14). In order to maintain an equalloudness sensation, one subject (S-il) required higheramplitudes as pulse rates were increased. The currentamplitudes required for comfortable hearing sensations werefound to be reasonably consistent from session to session(Figure 5).Anecdotally, pulse rates below 100 pps were oftendescribed as a “telephone ringing” or “motorboat”, dependingon the basal or apical location, respectively, of theactivated electrode. Subjects reported that pulse rates above29Figure 2. Current amplitudes at comfortable loudnessfor pulse rates from 54 to 1096 pps over a range ofpulse widths, from 150 to 600 isec. Subject 7,Electrode 18. Electrode separation = 1.5 mm.PW = pulse width/phase.Figure 3. Current amplitudes at comfortable loudnessfor pulse rates from 54 to 1096 pps for 4 subjects.Additional data regarding subject and stimulationparameters in Table 1. S= Subject; E= Electrode.800700 — --_60050020054 66 81 89 ili log 11 21 20 3Q 492 601 730 897 1096PULSE RATEPW=150 -- PW=300 - PW=600700620040060460380300[E1IE18_-A-- 541E20 -- S31E18 -- 514/E19301600140012001000800600400200-.- S-11/E-I8IPWIOO —w-- S10iE-18IPW25fjFigure 4. Current amplitudes at comfortable loudness fortwo subjects with electrode separation of 2.25 mm.S=Subject; E=Electrode; PW=Pulse width/phase.160014001200100080060040020054 81 121 181 270 403 601 89766 99 148 221 330 492 735 1096PULSE RATEFigure 5. Current amplitudes at comfortable loudnessfor Subject 7 over 10 sessions. Markers and verticalerror bars represent the mean and ± 1 standard deviation.Electrode 18; pulse width 150 jsec. Electrode separation1.5 mm.54 81 121 181 210 403 601 89766 99 148 221 330 492 735 1096PULSE RATE-f I I f I31100 pps tended to sound more pleasing (“more musical” and“smoother”) than lower pulse rates. At rates of 400—500 ppsand above, some subjects (especially S—li) frequently reportedhearing transient onset pitches followed by a sustained noisewith no definite pitch. Other subjects (particularly S-b)reported hearing occasional “double” pitches, which consistedof both a high-pitched and a simultaneous low-pitchedcomponent. However, these phenomena appeared to occurinconsistently, and were often not replicable, even uponrepetition of the identical pulse rate sequence. Othersubjects reported inconsistent decreases in pitch at pulserates above 600 pps.Above 200 pps, especially with wider pulse widths, somesubjects reported hearing only a transient, rapidlydisappearing sound. One subject described this as resemblingthe sound of a kettle drum, and another, as the sound of apencil tapping. For 1000 sec/phase pulses, this occurred atapproximately 200 pps, whereas for 300 sec/phase pulses, thislimit was usually near 1000 pps. These upper limits appearedto be similar for all subjects and electrodes tested.Apical electrodes were reported by some subjects toresemble the sound of a tuba, while basal electrodes weredescribed as sounding “like a telephone ringing”. Subjectswho participated in pilot experiments described the apicalelectrodes as sounding “more musical” than the basalelectrodes. Two subjects (S-2 and S-l4), both of whom hadsome musical background, volunteered that the place pitches32resulting from systematically sweeping across electrodes (1electrode/sec) in a basal—to—apical or apical—to—basaldirection at a constant pulse rate (250 pps) resulted in “abetter musical scale” than did slow (1/sec) pulse rate sweepson single electrodes. Only two subjects volunteered that thepitch changed with increases in amplitude. Subject 7indicated that only at high pulse rates (above 700-800 pps)did the pitch appear to increase with amplitude. Subject 5described a pitch decrease with increases in amplitude at allpulse rates.DISCUSSIONThe current amplitudes required for comfortable hearingsensations were comparable to those reported with similarlyspaced electrodes in both humans and animals (Clark, Black,Dewhurst, Forster, Patrick, and Tong 1977; Tong et al. 1982;Tong et al. 1983; Shannon 1989; Blamey et al. 1992).Interindividual and interelectrode differences in currentamplitudes required for auditory thresholds have been reportedto be associated with differences in nerve fiber survival andproximity of the electrodes to the neural targets (Pfingst andSutton 1983). Increased distance between stimulatingelectrodes and neural targets may explain the high currentlevels required by S—b, whose electrode array was located inthe scala vestibuli. Intersubject comparisons of stimulationrequirements in our data were precluded by individual33differences in electrode spacing and pulse width (Shannon1989; Blarney et al. 1992)When pulse width equals the period of electricalsinusoids, the current amplitudes required for detectionthresholds and for the upper limits of the dynamic range havebeen reported to increase systematically with increases inpulse rate, especially above 100 pps (Shannon 1983; Pfingst1984). However, with pulses of constant width, such as thoseutilized in our study, psychophysical detection thresholds andthe upper limits of the dynamic range have been reported todecrease as pulse rate is increased (Mladejovsky, Eddington,Dobelle, and Brackrnann 1975; Pfingst, Donaldson, Miller, andSpelman 1979; Tong et al. 1983; Pfingst 1984), presumably dueto an increase in charge transferred per unit time.While we did not specifically examine hearing thresholdsand maximum acceptable loudness, our results confirmed thatfor most subjects, the current amplitudes required for acomfortable loudness decreased somewhat as pulse rate wasincreased. The pulse rate versus amplitude function for S—bshowed a particularly steep slope. In contrast, S—li requiredan increase in current at higher pulse rates, to maintain anequal loudness sensation. It is possible that the upslopingcurve for S—li was a manifestation of increasingly rapidadaptation at higher pulse rates, perhaps associated with therelatively long duration (500 msec) stimuli. Thesedifferences between subjects may reflect differences in nervesurvival. Differences in the slope of the frequency/amplitude34function for psychophysical detection thresholds withdifferent degrees of cochlear pathology have been reported forimplanted monkeys, by Pfingst, Sutton, Miller and Bohne(1981). In spite of the subjectivity of the comfortableloudness criterion, patients demonstrated no difficultyreplicating comparable results in different sessions.The perception of intermittency for very low electricalpulse rates has been reported by other investigators (Simmons1966; Simmons et al. 1981; Shannon 1983), and is consistentwith gap detection thresholds of 2-5 msec reported for Nucleusimplant recipients (Shannon 1989), corresponding to interpulseintervals at pulse rates of 500 and 200 pps, respectively.The relatively more pleasing, musical quality of pulse ratesbetween 100-300 pps has also been reported previously (Simmons1966; Merzenich et al. 1973).The instability of auditory percepts observed with 500msec pulse trains at high rates may relate to the refractoryperiod of the electrically stimulated neuronal population.Van den Honert and Stypulkowski (1984), using bothintracellular and extracellular recordings of the cat auditorynerve activity in response to monophasic 100 sec closelyspaced electrical pulses, reported a graded decrease in theamplitude of the neuronal response to a second pulse asinterpulse intervals were decreased below 1.0 msec. Hartmann,Topp, and Klinke (1984) reported a similar decrease in firingprobability, due to the refractory period of neurons, wheninterpulse intervals were shorter than 5 msec. These findings35suggest that 200 to 1000 pps may constitute an approximateupper limit for eliciting equal-amplitude responses to bothpulses of a closely spaced pulse pair, and that closer spacingof pulses may generate only an “on” response to the firstpulse of a pulse train. This phenomenon could account for therapid loudness decrement noted with 500 msec stimulation athigher pulse rates. An analogous abolition of actionpotentials of auditory nerve fibers to continued (2-5 sec)stimulation with electrical pulse rates in excess of 600 ppswas reported by Javel et al. (1987), who attributed thisphenomenon to probable depolarization block.Alterations in pitch with changes in stimulus amplitude,such as those reported by S—5 and S—7, have also beenpreviously reported by other investigators. Shannon (1983),for example, showed an increase in pitch estimates for 1000 Hzmonopolar stimulation, as amplitude was increased. Theseamplitude-dependent pitch shifts were reported to be moreprominent than those associated with changes in the site ofstimulation. Townshend et al (1987) reported a pitchincrease for two patients, and a pitch decrease for a thirdpatient, as intensity of a 100 and 200 Hz stimulus wasincreased. Parkins (1989) has suggested that changes infiring rate which occur with changes in amplitude may providea partial explanation for these phenomena. It is equallypossible that the decrease in pitch with increased amplitude,such as that reported by S-5, could result from a predominanceof place pitch information at low stimulus amplitudes, and an36increasing prominence of rate—based pitches with the enhancedentrainment of spikes to the temporal pattern of theelectrical stimulus, as amplitude is increased. Amplitude—dependent pitch shifts could also result from spatialasymmetry in current spread or asymmetry of neural survival,relative to the stimulating electrode.The decreases in pitch reported by some subjects at highpulse rates are also consistent with psychophysical findingsof previous investigators (Hochmair-Desoyer et al. 1983;Shannon 1983). Physiological data (Parkins 1989) have shownthat at low pulse rates, interspike intervals are determinedlargely by repetition rate. At high pulse rates (2500 pps),neuronal responses were shown to remain fairly periodic, butinterspike intervals became a strong function of stimulusintensity (with decreasing interspike intervals as intensityis increased) rather than of interpulse intervals.Presumably, at high pulse rates, the interspike intervals aredetermined by interactions between the relative refractorystatus of neurons as they recover from the previous response,and the amount of charge delivered during the excitatory phaseof the stimulus waveform. Therefore, no further responses mayoccur until sufficient time has elapsed for the excitationthreshold to fall below the charge per phase delivered by thestimulus. The threshold nature of these responses (to all butthe initial pulse of a high frequency burst), as well as thesummation of jitter and possible variations in the neuronalrefractory curve preclude a high degree of synchronization tothe electrical waveform, such as that seen at lower pulserates.3738EXPERIMENT I: OPEN-SET TUNE RECOGNITIONRecognition of melody (a sequential pattern of pitches)requires more than simple pitch perception. It requiresextraction, by the listener, of certain relational propertiesbetween the constituent stimulus elements in a tonal sequence.Experiment I investigates, in a group of cochlear implantrecipients not selected for pre—deafness musical ability orpostoperative speech perception results, the recognition ofcommon tunes, “played” by systematically varying pulse rate onsingle apical intracochlear electrodes. For the purpose ofthis paper, “tune” was arbitrarily defined as a sequence ofmusical intervals of which the rhythmical pattern waspreserved, whereas “melody” was reserved for sequences ofmusical intervals without rhythmical or phrase-lengthinformation (i.e., pitch changes only). While it was fullyrecognized that the subjects who participated in Experiment Icould be basing their identifications on rhythmical cues,Experiment I nevertheless closely resembles the realisticsituation of a subject listening to music, where sequences ofpitches occur within a rhythmical context. The task ofrestricting discriminative cues solely to pitch changes wasaddressed in subsequent experiments.Variation of pulse rate was expected to induce temporallypatterned firing (Parkins 1989) in a relatively restricted39(Tong et al. 1982; van den Honert and Stypulkowski 1987a)group of cochlear neurons, and to result in some variation ofpitch (Simmons 1966; Merzenich et al. 1973; Eddington et al.1978a, 1978b; Simmons et al. 1979, 1981; Dillier, Spiliman,and Guentensperger 1983; Hochmair-Desoyer et al. 1983;Shannon 1983; Tong and Clark 1985; Townshend et al. 1987;Shallop et al. 1990).METHODSSubjects (Table 1) were 17 Nucleus implant recipients,not selected for pre—deafness musical or postoperative speechperception abilities, who were returning to the Laboratory forroutine follow-up testing. Three of the 17 subjectsparticipated, in addition, in Experiments II to V. Thesubjects ranged in age from 34 to 76 (mean 51.6 years).Duration of preoperative total deafness ranged from 1 to 24years (mean 5.5 years).During the test session, the subjects were connected bymeans of a cable and a transmission coil directly to theoutput of the computer and the BTNI interface. The currentamplitudes required for a comfortable loudness at the pulserates used in all the experiments were interpolated from theamplitudes established at discrete pulse rates during thepreliminary psychophysical measurements. The stimulus itemsconsisted of the first 10-20 notes of 30 well-known tunes.The melodic lines were established by a systematic variation40of pulse rate on single electrodes in the apical or midportionof the electrode array. Pilot experiments suggested thatapical electrodes sounded more pleasant and musical than basalelectrodes. A typical test session, including the preliminarypsychophysical measurements, required approximately 1—2 hours.Pulse rates for each tune were calculated by using thefrequency ratios appropriate to each of the twelve equal—tempered semitones of the octave. When the lower pulse ratewas designated as f0 and the upper pulse rate as f, the ratiobetween the two pulse rates was defined by the formulaf/fo=21/l2where n equals the number of semitone steps within thespecified musical interval and s is a number that determinesthe size of the semitone steps. The scaling factor s, whensmaller or larger than unity, results in a compression orexpansion, respectively, of all the nonunison intervals in agiven melody. The computer was then programmed to calculatethe appropriate pulse rates for the notes of the tunes, to thenearest integer number of pulses per second, once a scalingfactor (s) and a base pulse rate (f0) were specified. ForExperiment I proper, the base pulse rate was constant at 100pps and the scaling factor (s) was constant at 1.0. Pulserates and rhythmical patterns for each tune, as played to thesubjects, are detailed in Appendix 1. Two subjectsparticipated, in addition, in an informal experiment whichassessed the recognition of familiar tunes when interval sizes41were systematically altered, so that n was multiplied byscaling factors (s) of 0.5 or 2.0, resulting in, respectively,a compression or expansion of interval size.The subjects were instructed that the test would consistof the first 1-3 phrases of 30 common tunes, and that theywere to write either the title of the tune or some of thelyrics on the response sheet. If they did not recognize thetune, they were to write “unfamiliar” on the response sheet.The subjects were not informed of the identity of any of theitems, or of the potential pool of tunes. Each subject waspermitted a maximum of two repetitions of each item. Nofeedback was provided. Following the test session, subjectswere presented with a list of the titles and first lines ofall the tunes, and asked to indicate the tunes with which theywere familiar. Specifically, they were asked to indicatewhich tunes they thought they would have recognized, prior todeafness, if the tune had been played on a musical instrument.While not an objective measure, this procedure was intended tocorrect for the anticipated poor scores of nonmusicalsubjects, who might be familiar with only a small number ofthe test tunes. For each subject, two percentage correctscores were calculated. The first was a percent correct scorebased on the total 30 test items. The second was a percentcorrect score based on the number of tunes with which subjectsreported themselves to be familiar. These two scores werearbitrarily termed the absolute and relative (i.e., relative42to the number of tunes familiar to individual subjects)scores, respectively.To determine whether identification was likely solely onthe basis of rhythmical information, an additional informalexperiment was performed in which the 30 tunes were played to5 normal—hearing observers, using a constant 100 Hz squarewave monotone, via the acoustic monitor of the BTNI interface.Thus, only the rhythmical information of the tunes waspreserved. While a similar control involving electricallystimulated subjects would have been desirable, this wasprecluded by a lack of time and subjects. It is recognizedthat this procedure also does not yield information regardingthe influence of rhythm in the normal situation of a subjectlistening to music, as rhythm may have a multiplicative orsynergistic effect on recognition, when combined with melody.RESULTSIt is difficult to specify a precise level of chanceperformance on open—set tune or musical phrase recognitiontests, particularly when the instructions to the subjectsspecify “familiar” tunes. Subjects are likely to select theirresponses from their repertoire folk tunes, nursery rhymetunes, children’s songs or Christmas tunes, and thereby tolimit the number of available response alternatives. Previousresearch has shown that when normal—hearing subjects are askedto identify familiar tunes which have been intentionally43distorted, most of the incorrect guesses tend to be drawn fromthe same general category of popular traditional Americansongs (Deutsch 1972). Thus, a “familiar” tune recognitiontest cannot be considered to be entirely an open-set paradigm,because of the limited number of response alternativesavailable to the subjects.It can be assumed, however, that a chance performancelevel would be close to 0%. The results of Experiment I areshown in Figure 6. Approximately one-half of the subjectsscored 40% or better, and only 3 scored less than 10%.Absolute scores (based on 30 test items) ranged from 0% to67%, with a mean score of 33.7%. Of the 8 subjects who scoredbetter than 40%, 5 had no previous experience with the stimuligenerated by the experimental setup, except for thepreliminary psychophysical measurements. The other threesubjects with high scores had received several hours ofexposure during pilot studies. The only subject who failed toidentify any of the test items had been deaf for 25 yearsprior to implantation. The frequency of identification ofindividual tunes is shown in Figure 7.Relative scores (relative to the number of tunes thesubjects indicated they could or should have recognized priorto deafness, had the tunes been played by a musicalinstrument) ranged from 0% to 84%, with a mean score of 44.1%.The relative scores were always equal to or better than theabsolute scores. When asked about the sound quality, somesubjects complained that the tunes sounded artificial and44IC)Ui0C)IzUiC)Uia.ABSOLUTE SCORE RELATIVE SCORE(#CORRECT/30)Xl 00 (#CORRECT/#FAJAR)Xl 00Figure 6. Open—set tune recognition: percentage ofitems correctly identified by each of 17 subjects.Percent correct scores were calculated both on the basisof the entire set of 30 test items (absolute score), andon the basis of the number of tunes with which thesubjects reported themselves to be familiar (relativescore). Details regarding subjects in Table 1.1 2 3 4 5 6 7 8 91011121314151617PATIENT NUMBER450z0I-C)LIIzw0IC)waa0C)II.0awDz‘ErIt__.1 anI P.’‘yr-Fr1.4-1 2 3 4 5 6 7 8 91011 12131415161718192021 222324252627282930TUNE NUMBERFigure 7. Open—set tune recognition experiment:number of positive identifications for each test tune.Details regarding tunes in Appendix 1.46unmusical, while others reported them to sound clear andpleasing. Four of the subjects volunteered that theiridentifications were based mostly on rhythmical cues, ratherthan on pitch cues.It is obvious from an examination of the data (cf. Figure6 and Table 1) that musical interest or achievement alone,prior to deafness, cannot account for differences inperformance on the tune identification test. The 3 subjectswith the highest scores reported an historical lack ofinterest in music. However, 6 of the 9 subjects with a tuneidentification score of 40% or greater reported having playeda musical instrument. In contrast, five of the 7 subjects whoscored 20% or less reported no musical interest. The subjectswho had played a musical instrument (N=8) achieved a meanscore of 44.9%, whereas the subjects who had no musicalinterest (N=8) had a mean score of 24.6%. The speechperception scores (Table 1) of the subjects with some musicalhistory were slightly higher (80.1%) than those for thesubjects with no history of musical interest (mean 70.1%).The relationship between the tune identification scores andspeech perception scores was plotted in Figure 8. The twosets of scores were found to be moderately correlated (r=.53).The tune identification experiment was repeated,informally, with two subjects, using either expanded orcompressed interval sizes, with an s value of either 0.5 or2.0 randomly assigned to each tune. The subjects (S-7 and S47100ILU80IXLUXX600z X040XXX1XzLU X X20 XLUz XI-0 I I I Xi I I I I I I I I0 10 20 30 40 50 60 70 80 90 100SPEECH PERCEPTION: PERCENT CORRECTFigure 8. Scatter plot of the relationship between tuneidentification (Figure 6) and speech perception (Table 1)scores.4811) identified the mistuned tunes as well as their properlytuned counterparts, and responded with incredulity, wheninformed of the mistuning, following completion of theexperiment.In the other informal experiment, five normal—hearingcontrols failed to identify any of the 30 tunes of ExperimentI, when these tunes were played as rhythmical patterns withoutpitch changes, using a constant 100 Hz square wave monotone,via the acoustic monitor of the BTNI interface.DISCUSSIONThe results suggest that tune recognition in profoundlydeaf subjects is possible with pulsatile electricalstimulation on single intracochlear electrodes. Thesefindings are in agreement with the results of a number ofprevious experiments involving electrically stimulated deafsubjects, suggesting that pitches sufficiently salient formusical perception are possible on the basis of solelytemporal information. Eddington et al. (1978a, 1978b)reported limited tune recognition with pulsatile stimulation,using a single intracochlear electrode. Their single subjectcorrectly identified 3 out of 5 test tunes played at low pulserates on one electrode, but was unable to identify the tuneswhen they were played on other electrodes. Chouard (1978)claimed that postlingually deaf implanted patients weretypically able, with near 100% accuracy, to recognize popular49tunes. However, the experimental protocol was poorlydocumented. Moore and Rosen (1979) reported essentiallyperfect scores for a single deaf subject on a 10—itemalternative—forced-choice melody identification test, using anextracochlear electrode and an analog processing scheme. The16—note rhythmically identical tunes were presented live—voiceand low-pass filtered at 300 Hz. Their patient observed thateven though the sound quality resembled that of “a comb andpaper”, the melodies were nevertheless “in tune”. Thesefindings support our conclusions that musical pitchinformation sufficient for tune recognition can indeed beconveyed in the temporal discharge patterns of electricallystimulated auditory neurons.Similar data, consistent with the existence of musicalpitches based on temporal features of acoustic waveforms, havebeen reported in normal—hearing subjects. Burns andViemeister (1976) reported musical interval identification andclosed-set tune recognition utilizing both wide band andbandpass sinusoidally amplitude—modulated noise. In asubsequent experiment, Burns and Viemeister (1981) showed thateven musically naive listeners were able to identify, open—set, rhythmically identical, same-length melodies played bysinusoidally amplitude-modulated white or high pass noise.While the mean number of items correctly identified was lower(13.3* to 22.5%) than that obtained by our subjects, thisdifference may well be due to our preservation of rhythmicalinformation.50The observations of some subjects about the importance ofrhythmical information in their tune identificationperformance could be due to several factors. It is possible,for example, that the electrical pulse rate pitches were weakand difficult to hear, at least for some subjects. Themoderate correlation found between speech perceptionperformance and tune identification scores supports arelationship, found also by other investigators, between theability of implanted subjects to utilize temporal informationand their ability to understand speech (Shannon 1989; Tyler,Moore, and Kuk 1989). The unnatural sound quality resultingfrom electrical stimulation of the auditory system should notbe surprising, in view of the highly synchronized responses ofneurons over a range of characteristic frequencies, as well asthe absence of the delays normally imposed by basilar membranemechanics and hair cell transducer action, and the lack ofcongruence between place, period, and phase information (vanden Honert and Stypulkowski 1987b). Weakness of pitch couldexplain, for example, the failure of the subjects in theadditional informal experiment to detect gross mistuning ofthe melodies (i.e., the sequence of intervals) of the tunes.Alternately, this failure could result from their attendingsolely to the identification task, since the subjects were notinstructed to attend to the intonation quality of theintervals in the tunes. The poor scores and predominantreliance of some subjects on rhythmical information could alsoresult from a musical illiteracy of the subjects or the high51variability of the size of the musical repertoire known toexist even among normal—hearing subjects. It is also possiblethat, following a period of total deafness, the memories ofthe melodies of common tunes for some subjects were no longerintact. It was not possible to examine the effects ofprolonged deafness on tune identification, as only three ofthe 17 subjects had been deaf for more than 10 years.An important question remains, however, whether thesubjects in Experiment I could have been responding solely tothe rhythmical patterns of the tunes rather than to themelodies, i.e., to the sequence of intervals. Musicallyuntutored subjects reportedly listen to music in anonanalytical way, in which the sequence of pitches isinextricably linked to the rhythmical structure whichaccompanies the melody (Cross, Howell and West 1985; Jones,Summerell, and Marshburn 1987). The possibility of limitedrecognition on the basis of exclusively rhythmical informationhas been previously demonstrated by others. Deutsch (1972),for example, reported that 19% of subjects were able torecognize (open—set) Yankee Doodle when presented as a seriesof timed clicks. While identification of our tunes solely onthe basis of rhythmical information remains a possibility, itwould appear to be unlikely, in view of the inability ofnormal—hearing controls to identify the tunes when they wereplayed as mere rhythmical patterns, without variation inpitch. However, the fact that rhythm by itself was52insufficient for tune recognition does not mean that rhythm isalso unimportant when accompanied by melody.Thus, reliance on exclusively rhythmical information, inthe absence of pitch interval or musical contour cues, couldbe expected to result in poor recognition scores, especiallyin an open-set paradigm. While rhythmical information mayhave contributed to tune identification in Experiment I, asobserved by some of our subjects, our evidence suggests thatmelody played the greater role.Gfeller and Lansing (1991) found that 10 Nucleus implantrecipients obtained low scores when asked to identify shorttaped excerpts of solo renditions of nine familiar tunes,produced on acoustic instruments. These excerpts were playedover a soundfield speaker and processed by the body—worndevice normally used by the subjects. The Nucleus processorperforms a feature extraction upon incoming acoustic signals,in which the electrode pairs, selected on the basis of a placecode, are activated at a pulse rate equal to the fundamentalfrequency. In spite of the preservation of rhythmicalinformation, subjects (including, in addition, 8 Ineraidrecipients) of Gfeller and Lansing (1991) were able toidentify the tunes for only 5% of the total number of trials.Thus, these subjects appeared to be unable to use rhythmicalinformation to identify the tunes. This inability did notappear to be due to a general ineptitude in the use ofrhythmical cues, since these subjects reportedly achieved high53scores on the rhythmical subtests of the Primary Measures ofMusic Audiation (PMMA).The poor recognition scores of the subjects in theGfeller and Lansing (1991) study may be attributable to anumber of factors. With the Nucleus implant, for example, acomplex sound delivered to the microphone of the processorresults in a quasi—simultaneous activation of a number ofelectrodes selected on the basis of spectral peaks in theacoustic signal (Koch, Seligman, Daly and Whitford 1990), witheach electrode yielding a more or less distinct sound quality.It is conceivable that rate pitches resulting from the quasi-simultaneous activation of a number of electrodes may be moredifficult to discern than those resulting from simplevariation of pulse rate on a single electrode. Thus, it ispossible that, under these conditions, the more complicatedstimulation pattern resulting from the feature extractionprocess may override the temporal structure needed for pitchperception.It is noteworthy that several of our subjects alsoachieved relatively low scores (e.g., S—i, S—6, S—9, S—12),even with simple variation of pulse rate on a singleelectrode. It is possible that these subjects were less ablethan the remainder to use temporal information. It is notedthat the speech perception scores of these subjects tended tobe somewhat lower than those of the remaining 13 subjects.However, additional factors, including the extent of pre—deafness musical aptitude and experience, and the size of54individual musical repertoires, may also be significantfactors. In addition, the slow tempo of some of the testtunes (Appendix 1), relative to the tempo at which these tunesare conventionally played or sung, may have degradedperformance for some subjects. In the author’s experience,some normal-hearing subjects also have difficulty identifyingtunes when the tempo differs from that to which the subject isaccustomed (as for example, Lutheran chorales played or sungas a slowly moving cantus firmus in the chorale Cantatas ofJ.S. Bach). It is also possible that, following the prolongedauditory deprivation of musically naive subjects, theinternalized representations of melodies and the underlyingtonal schemata are no longer intact in long—term memory.While Dowling (1978) and others have convincingly argued thatmemory for melodies and the tonal framework on which thesemelodies are hung are amongst the most stable of sensoryschemata in cognitive psychology, the effects of long-termdeafness on such memories have not been investigated.55EXPERIMENT II: CLOSED-SET MELODY IDENTIFICATIONAlthough the results of Experiment I showed that subjectswere able to identify common tunes when played as a sequenceof pulse rates over single intracochlear electrodes, it wasnot clear to what extent performance was based on rhythm, oron a combination of rhythm and melody. Removal of rhythmicalinformation, while introducing a significant distortion fornaive listeners, who reportedly listen to music in anonanalytical way, which does not separate rhythm and melody,should permit the assessment of the role of pitch information.For this paper, melody has been defined as an ordered seriesof musical intervals, in the absence of rhythmical or phraselength information.Melodies may be described by the frequency intervals thatseparate temporally contiguous notes and by their melodiccontour. Contour is a generalized form of intervalinformation, and is defined by the directional relationshipsbetween temporally adjacent notes, without precisespecification of interval magnitude. In Experiment II, aclosed-set melody identification paradigm was utilized todetermine whether unequivocal melody identification waspossible in the absence of rhythmical information. ExperimentII, in addition, assessed whether this information was also56available at higher pulse rates and on more basally situatedelectrodes.METHODSThe stimulus set consisted of 7—note truncations of theinitial phrases of familiar tunes. For the purposes of thispaper, we will refer to these isorhythmical 7—note fragmentsas melodies. All the melodies were equal in length and devoidof rhythmical cues, in that all the note durations were 500msec, and each note was followed by a 200 msec silent period.While the rhythmical equalization procedure left the melody(i.e., the sequence of intervals) intact, this procedurenecessarily resulted in a rhythmical distortion of some of theoriginal tunes. The pulse rates (Appendix 2) were assigned ina manner comparable to that described for the tunes inExperiment I. Items not familiar to individual subjects weredeleted from the test set for those subjects. Subject 11 wasfamiliar with all 8 items. Subjects 7 and 10 were familiarwith only 5 or 6 of the items. All the melodies within theresponse set were constructed for similarity in respect tomelodic center of gravity and pitch range. The computer wasprogrammed to calculate the appropriate pulse rate values tothe nearest whole number for each note of the melody, once abase pulse rate (f0) had been specified. Melodies and basepulse rates were randomized between 75, 100, 150, 200, 300,57400 pps by the computer. One subject (S-7) was further testedwith a base pulse rate of 600 pps.Subjects 10 and 11 were not tested at higher pulse ratesbecause at high pulse rates, they appeared to be able toachieve high recognition scores, at least in part, on thebasis of non—pitch attributes of the sound. For example, atpulse rates above 600-800 pps, S-lU reported hearing only abrief onset sound at the beginning of each stimulus note,which he described as “the sound of a kettle drum”. It becameapparent during the initial trials that, at the highest pulserates, he was able to use this information to identify themelodies. Thus, at the highest pulse rates, he responded with“Twinkle, Twinkle, Little Star” whenever he heard (as “twobrief notes on a kettle drum”) the two sequential A’s at thehighest point in the pitch contour of this melody. Similarly,at these pulse rates, he responded with “0 Suzanna” wheneverhe heard (as “one short note on a kettle drum”) the highestnote (A) in the contour of this melody. Subject 11 reportedsimilar phenomena. Thus, for these two subjects, performanceat the highest pulse rates appeared to be based on non-pitchinformation. For this reason, S—b and S—lb were not testedat the highest pulse rates.During the pilot experiments, all three subjectscomplained about the instability of auditory percepts athigher pulse rates, and frequently commented that melodies atpulse rates above 200-300 pps failed to sound the same uponrepetition. Subject 11 described pulse rates above 100 pps as58having an onset sound with a transient, but definable pitch,which was followed immediately by a pitchiess, noiselikesound. These instabilities were identical to those observedduring the preliminary psychophysical measurements.Closed-set melody identification at the specified basepulse rates (f0) was assessed for electrodes located in thebasal, the midportion, and the apical regions of the array.The sequencing of the electrodes used during the tests wascounterbalanced between subjects and sessions, to minimize theeffects of fatigue and practice. Thus, during a givensession, the sequencing of electrodes for one subject mightprogress from apical to basal, while the sequence for othersubjects was in a basal to apical direction. The sequence ofelectrodes used in the tests was varied from session tosession. Each subject was provided with a numbered,individualized list of the test melodies. Within eachsession, each item on the list was presented, in randomizedorder, twice at each base pulse rate. The subjects wereinstructed to select a response from the list, and to pressthe corresponding number on the computer keyboard. Thisresulted in automatic scoring of responses and initiation ofthe next trial. No repetition of stimulus items waspermitted, and no feedback was provided. The percentage ofcorrect responses was calculated for each condition. Todetermine whether the magnitude of pulse rate musical intervalin the melodies represented a significant factor in melodyidentification performance, each melody was assigned to one of59three groups: those with at least one interval equal to orwider than a 4th (large intervals), those in which the largestinterval was a major or minor 3rd (medium sized intervals),and those with only small intervals (major 2nd or semitonesteps only).Normal—hearing subjects readily recognize transposedmelodies as musical equivalents, because the transpositionpreserves the precise sequence and magnitude of musicalintervals. When the sizes of the musical intervals ofmelodies are altered, melodies lose their identity, unlessrhythmical cues (absent in this experiment) or the melodiccontour are sufficiently distinctive. To determine whether S—7 was listening to the melodies as a sequence of musicalintervals, an informal ancillary experiment was devised, inwhich this subject performed the closed-set melody recognitiontask with an unfamiliar melody added to the stimulus set. Thesubject was not informed of the presence of the additionalmelody. It was expected that, if the subject were listeningto the stimuli as music (i.e., as a sequence of musicalintervals), he might readily declare the presence of anunfamiliar melody. If he were listening only for a similarlyshaped melodic contour, he was expected to match theunfamiliar melody to a same-contour melody in the responseset.60RESULTSAn analysis of variance (ANOVA) was performed on thepercentages of melodies correctly identified, with 6 levels ofpulse rates, 3 electrodes, and 3 levels of magnitude ofinterval size. The mean percentage correct scores for eachsubject are plotted in Figure 9. Subjects differedsignificantly in their performance [F(2,26l4) = 186.21,p<.0000l]. Subject 7 scored 3% to 56% higher than S-b and S11, with the greatest performance differences occurring at thehighest pulse rates. The scores of S—b and S—il were notsignificantly different. These findings were in sharpcontrast with the self-reported musical history of thesubjects, as both S-b and S-li reported a longstandinginterest in music prior to deafness. While S-7 did not reportany specific interest in music prior to deafness, he wasnevertheless familiar with a considerable variety of melodies(Experiment I).The pulse rate of the starting note had a significanteffect on subject performance [F(5,26l4) = 127.97, p<.0000l].At low pulse rates, subjects identified the melodies with 80-100% accuracy. Subjects 7 and 10 did so without apparentdifficulty. Subject 11, however, reported that the pitcheswere difficult to hear, even at low pulse rates. The decreasein scores at higher pulse rates was more marked for S—b andlb than for S-7. While mean scores at low pulse rates (75 and100 pps) were not significantly different from each other,61Figure 9. Closed—set melody recognition experiment:percentage of items correctly identified by Subjects 7,10, and 11, for electrodes situated in apical, basal,and middle portions of the electrode array, over arange of pulse rates. The numbers along the horizontalaxis do not represent a dimension, and represent thebase pulse rate (“c”), at octave multiples of 75 and 100pps. Pulse rate assignments for individual notes areshown in Appendix 2. Chance level of performance foreach subject is indicated by a horizontal line acrossthe panel.bSz7 100 150 200 300 400 6005-10908070b6050403010ñ 10 iSO 200 300 400 600b01I-.zUa.PULSE RATEELECIRE -- MIDDLO ELECTRODE -•- BASAL ELECTRODE62scores at higher pulse rates showed, in general, a monotonicdecline with increasing pulse rate. The superior performanceof S-7 at higher pulse rates resulted in a significantinteraction between subject and pulse rate variables[F(lO,26l4) = 18.581, p<.00001]. At low pulse rates, Subjects7 and 10 reported being able to hear all the pitch changes ineach melody, but at higher pulse rates, they indicated thatonly some of the pitch changes were detectable. Subjects alsovolunteered that only at low pulse rates did the test itemsresemble music.In the informal ancillary experiment, one of the subjectsperformed the closed-set melody identification task with anunfamiliar melody added to the stimulus set. Only when theunfamiliar melody was played at low pulse rates did S-7conclude that the item was not on the list of possibleresponses. This suggests that, at higher pulse rates, thesubjects were perhaps simply matching melodies on the basis ofpitch contour, rather than processing the pulse rates as asequence of musical pitch intervals. The possibility alsoremains that, at high pulse rates, subjects were using complexcognitive strategies by attending to nonpitch attributes (suchas “double notes”).The main effect of the electrode on which the melody wasplayed was also significant [F(2,26l4) = 14.944, p<.0000l].Scores for Subjects 10 and 11 were best on apical, and worston basal electrodes. The performance decrements for these twosubjects on the basal electrodes were most marked at the63higher pulse rates, accounting for significant interactionsbetween the electrode and pulse rate variables [F(l0,2614) =3.1907, p=.00044]. There were no significant differences inscores between electrodes in the basal portion and midportionof the array. The significant interaction between the subjectand electrode variables {F(4,2614) = 6.2104, p=.00006] was dueto differences between the performance of S—7, whodemonstrated only a slight decrease in scores at the highestpulse rate tested, and the other two subjects, for whom thedecrease in scores at higher pulse rates was marked. Thisaccounted for a 3—way interaction between subjects,electrodes, and pulse rates [F(20,2614) = 1.9284, p=.00787].All three subjects reported that the more apical electrodesyielded the more musical and pleasant sounding melodies.The size of the intervals in the melodies was notsignificant as a main effect [F(2,26l4) = 2.0017, p=.l3532],and failed to produce significant interactions with theelectrode and pulse rate variables. Thus, large intervals(such as 5ths and 4ths) did not appear to make melodies moreidentifiable than small intervals (such as major 2nds andsemitones). Although a significant interaction occurredbetween subjects and interval magnitude [F(4,26l4) = 5.922,p=.000lO], there appeared to be no consistent pattern to thisinteraction. For example, S—7 obtained the highest scoreswith melodies consisting of small intervals, and the lowestscores with large-interval melodies. Subject 10 obtained thehighest scores with medium—sized intervals and the lowest64scores with small intervals. Subject 11 obtained the highestscores with large—interval melodies, and the lowest scoreswith melodies consisting of medium-sized pitch intervals. Inview of the small number of melodies in each interval sizecategory, it is possible that these interactions resulted fromother, noninterval size variables, such as a greaterfamiliarity of subjects with some of the melodies, or agreater resistance of some melodies to the temporal distortionimposed by rhythmic equalization.DISCUSSIONThe findings suggest that systematic variations in pulserate on single electrodes can result in pitch perceptssufficiently salient to enable subjects to score well on aclosed—set melody recognition test, in complete absence ofrhythmical information. At low pulse rates, scores werecomparable for all electrodes. At high pulse rates, the morebasally situated electrodes yielded a greater performancedecrement than apically situated electrodes. This, togetherwith observations of the subjects regarding the more musicalquality of the apical electrodes, may reflect the importanceof a congruence between place and rate information in pitchprocessing. The similarity of the results for basal andapical electrodes at low pulse rates are also in agreementwith gap detection data of Shannon (1989), showing equivalent65gap detection thresholds at equal loudness levels, regardlessof the intracochlear location of the stimulating electrodes.There were significant differences in the ability ofsubjects to utilize pitches at higher pulse rates. Subject 7,for example, achieved high scores even at the highest pulserates tested, while 5—10 and 11 showed large performancedecrements at higher pulse rates. Subject 10 appeared, inaddition, confused by the rhythmical monotony of the melodies,even though he reported that the pitches, especially at lowpulse rates, were not difficult to hear. Subject 11 commentedthat all the pitches were difficult to hear, even at thelowest pulse rates. It is possible that the weakness of thepulse rate pitches, for this apparently relatively musicalsubject, may relate to the distribution, number, or functionalstatus of residual spiral ganglion cells, perhaps as a resultof deafness secondary to meningitis (Nadol and Hsu 1991).Previous investigators have also reported large individualdifferences in the abilities of cochlear implant recipients toutilize temporal information (Shannon 1989; Tyler, Moore, and[<uk 1989)The upper limits for pitches resulting from pulsatileelectrical stimulation in Experiment II appear to becomparable to the 850—1000 Hz upper limit of pitch perceptionobserved for sinusoidally amplitude—modulated noise by Burnsand Viemeister (1976, 1981), in tasks requiring musicalinterval recognition and dictation, as well as melodyidentification. This upper limit was hypothesized to reflect66the inability of the auditory system to follow rapid temporalchanges. Even for the musically trained observers of Burnsand Viemeister (1976, 1981), intersubject differences atmodulation frequencies above 300 Hz were considerable.Comparable individual differences, such as those between S—7on the one hand, and Subjects 10 and 11 on the other, wereobserved in our data. Tong and Clark (1985) reported similarupper limits and individual differences in the ability ofimplanted patients to identify electrical pulse rates. Thesedifferences, in our subjects as well as those of Tong andClark (1985), appeared to be unrelated to musical experience.For electrical stimulation rates above 500 Hz, theability of individual neurons to fire on a cycle-for-cyclebasis is limited by the neural refractory period (Javel et al.1987; van den Honert and Stypulkowski 1987b). While theabsolute refractory period has been estimated at approximately300 sec, the relative refractory period may extend to atleast 5 msec (van den Honert and Stypulkowski 1984). Athigher pulse rates, interval histograms for single units havebeen shown to become multimodal, with an increasingrepresentation of multiples of the stimulus period (van denHonert and Stypulkowski 1987b). It is possible that theincreasing intrusion of a variety of interspike intervals,related more to multiples of the stimulus period or to therefractory period of the neurons than to the interpulseintervals, results in a weakening of pitch at higher pulserates.67While results for Experiment II showed that, even in thecomplete absence of rhythmical information, subjects were ableto identify melodies from a closed set, it could not bedetermined from these results whether these pitches conveyedonly contour information, or both contour and intervalinformation. Contour refers simply to the ordinalrepresentations of the note frequencies, and thus designateswhether individual notes are higher or lower in pitch thantheir contiguous counterparts, independent of precise intervalsize (Watkins and Dyson 1985). Previous research has shownthat, especially in closed—set recognition paradigms, melodiccontour can serve as a significant cue in the retrievalprocesses or strategies utilized by subjects. Massaro,Kallmann, and Kelly (1980) and Dowling and Bartlett (1981)have suggested that in the closed—set format, subjects mayretrieve all items in the response set from memory, extractthe contour by covert rehearsal, and then compare the contourswith those of the stimulus items. The possibility should alsobe considered that, especially at higher pulse rates, oursubjects (or S-7, specifically) may have been applyingnonmusical strategies, such as listening for systematicchanges in non—pitch attributes, such as timbre, density,brightness, roughness, or fullness.In summary, the results of Experiments I and II suggestthat melody recognition is possible with pulsatile electricalstimulation of deaf ears, using single intracochlearelectrodes. While it is possible that rhythmical cues were68important in the retrieval process in the open—set paradigm(Experiment I), the alternative-forced—choice melodyrecognition (Experiment II) results for these 3 subjects showthat, even in the total absence of rhythmical information,subjects were able to perform at high levels. Performance,however, was best at low pulse rates, and a variable butsignificant deterioration occurred for all subjects at higherpulse rates. These upper limits of performance, whilediffering for different subjects, compare favourably to theupper limits of temporally mediated pitch in acousticexperiments (Burns and Viemeister 1976).69EXPERIMENT III: INTONATION QUALITY JUDGEMENTSContour information alone is known to be too imprecise toaccount for the accurate memory for melodies, even inmusically untrained subjects (Cross, Howell, and West 1985;Sloboda and Parker 1985). Precise interval size has beenshown to be important particularly in long—term memory forfamiliar melodies. Dowling and Fujitani (1971) showed thatwhile melodic contour may be more important than preciseinterval size in short—term memory for unfamiliar melodies,well—learned melodies are stored as a precise sequence ofmusical intervals. Thus, even unsophisticated listeners arecapable of discriminating between exact transpositions andless precise imitations of familiar melodies. Transpositionspreserve the precise sequence and sizes of musical intervalsfound in the original, whereas imitations preserve thecontour, but not the precise interval sizes of the originalmelody.The possibility remains that, in the melodyidentification tests, our subjects were perhaps merelyrecognizing the ordinal relationships among the pitches, orthat they were applying more complex cognitive strategies,attending perhaps to nonpitch attributes of the signal. Thus,while the pitches heard by these subjects could perhaps bescaled or ranked by means of higher/lower, up/down,70same/different psychophysical procedures, these pitches couldnot be assumed to be sufficiently salient to conveyinformation about musical intervals (Burns and Viemeister1981; Houtsma 1984). Musical pitch intervals arecharacterized by frequencies in very specific ratio relations(Houtsma 1984). Ward (1970) and Burns and Ward (1978) havepointed out that the precision shown by musicians in adjustingmusical intervals exceeds that shown by trained subjects inadjusting ratios of other auditory or nonauditory percepts.Nonmusicians, however, are known to be less precise, and arethought to be influenced more by pitch height rather than bycomplex hierarchical and ratio—governed stimulus relations(Siegel and Siegel 1977; Krumhansl and Kessler 1982; Monahan,Kendall, and Carterette 1987).To assess the response of cochlear implant subjects toeither the ratio properties or, alternatively, to the pitchextent (differences in pitch height) of the musical intervalswhich were represented by changes in pulse rate, an intonationquality test was devised.METHODSSubjects were required to label each stimulus pitchinterval, represented by changes in pulse rate on a singleintracochlear electrode, as “flat”, “sharp”, or “in tune”,relative to their memory for one of four specific intervalsexemplified in well—known melodies. For example, to determine71the pulse rate ratios that would best characterize thesubject’s memory of an ascending 5th, the subjects were askedto rehearse mentally the second interval (i.e., the firstnonunison interval) of “Twinkle, Twinkle, Little Star”, andthen to label each of 12 stimulus interval sizes as “in tune”,“flat”, or “sharp”, relative to their memory for the intervalin question (Figure 10). The 4 target intervals (minor 3rd,4th, 5th, major 6th) were abstracted from the initial phrasesof melodies well-known to individual subjects. Each of theseintervals was represented by only two pitches. In theconventional musical scale, these four intervals consist of 3,5, 7, and 9 semitones, respectively.Attempts were made to select melodies with at least somerepetitive notes (e.g. the repeated lower and upper notes ofthe 5th in “Twinkle”). Melodies used for the 4th were thefirst 4 notes of “0 Christmas Tree”, or the first 3 notes of“Away in a Manger”. The minor 3rd was exemplified by thefirst 8 notes of the chorus of “Jingle Bells”. The major 6thwas exemplified in the first two notes of “Jingle Bells” or“My Bonnie Lies Over the Ocean”. It is possible that theshort, two—note intervals such as those used for the major 6thwere more difficult than the 4- or 8-note sequences, becausethe subjects had less time to process the pitch information.Note duration was 500 msec, and notes were separated by 200msec.For a given block of stimuli, the pulse rate of the lowernote remained constant. Preliminary trials with a lower note72INTONATION QUALITY JUDGMENTSaoo1.891.781.691.591.50 IN TUNE1.411.331.261.191.121.061.00Figure 10. Intonation quality judgements:schematization of the stimuli and the task of thesubjects for an interval of a 5th, as exemplified in thefirst nonunison interval in “Twinkle, Twinkle, LittleStar”. The stimulus set consisted of 12 intervalsranging in size from a semitone (ratio of 1:1.06) to anoctave (ratio of 1:2) above the lower note of theinterval, in steps of one semitone. The target ratio(“in tune”) was 1:1.5. With acoustical stimuli,normal—hearing musical listeners label ratios largerthan this as “sharp” (“too high”), and smaller ratios as“flat” (“too low”)TOO HIGHTOO LOWCTWINKLE TWINKLE73pulse rate which was varied from trial to trial yieldedcomparable results, although requiring considerably greatereffort for the listeners. The pulse rate of the upper notewas varied to create a set of 12 interval sizes, eachdiffering from its adjacent counterpart by one semitone. Thisresulted in one correct target interval and 11 foils, rangingin size from a minor 2nd to an octave above the lower note.The pulse rates assigned to the lower note were different foreach target interval, to preclude the utilization of aconstant pitch reference across the entire set of 4 intervals.Specifically, the pulse rates for the lower notes were 137,145, 163, and 127 pps for the minor 3rd, the 4th, the 5th, andthe major 6th, respectively. Semitone step ratios werecalculated according to the equal—temperament musical scale,where tones separated by a semitone interval are defined by afrequency ratio of1:21/12 or 1:1.05946.The pulse rate for the second note of the intervals wascalculated by the formulaPulse Rate of First Note * 2n/l2where n equals integers from 1 to 12. For each targetinterval at a specified starting pulse rate, the 12 differentintervals were randomly presented to the subjects atamplitudes interpolated from the preliminary psychophysicalmeasurements.74Subjects were informed of the musical interval underscrutiny and the tune fragment which exemplified the testinterval. Subjects were then required to label each of thepitch intervals they heard as “in tune”, “sharp” (or “toohigh”), or “flat” (or “too low”), relative to their memory forthe test interval, and to respond by pressing the appropriatekeys (“F”, “T”, “S”) on the computer keyboard. No exemplarsof the target intervals were presented, and no feedback wasprovided regarding the “correctness” of the responses.A similar paradigm was used to investigate whether theratio relationships demonstrated for a 5th at low pulse rateswere transposable to lower and higher pulse rates. Threelower-note pulse rates (81, 163, 326 pps), separated byoctaves, were randomized from block to block.An additional series of blocks examined whether theability to detect gross mistuning of musical intervals at lowpulse rates was perhaps restricted to apical electrodes, orwhether similar results could be obtained via more basalelectrodes. Each block of trials consisted of 5 presentationsof each of the 12 intervals along the pulse rate continuum (60judgements). Each target interval or condition was assessedin 5 blocks (300 judgements). Following each block of trials,subjects received a 15 minute rest period.For each interval size, condition, and subject, thepercentage of “in tune”, “flat”, and “sharp” responses wascalculated. The percentages of “flat” and “sharp” judgementsfor each stimulus interval size category were used to75calculate the mean transition points from “flat” to “notflat”, and from “sharp” to “not sharp” (Woodworth 1938). Themean of these two points was taken as the point of subjectiveequality (PSE), or the subjective musical interval (i.e., theinterval equivalent to the memory of the subject for thisinterval, as represented in the melody in question) for eachsubject, interval size, and condition. In addition, astandard deviation was calculated, using the method describedby Woodworth (1938). This computation allowed relativelystraightforward comparison of the data from this experiment,using the method of constant stimuli, with those obtainedusing the method of adjustment (Experiments IV and V).Two additional informal pilot experiments were performed.The first determined whether subjects were able to assignintonation quality labels to intervals which were representedsolely by changes in cochlear place (i.e., electrode) ofstimulation. Thus, two subjects were asked to label the pitchchanges resulting from switching, for example, from electrode19 to electrode 18, or from electrode 19 to electrode 17, 16,or 15 (at a constant pulse rate), as “flat”, “sharp”, or “intune” relative to the interval of a 5th, using a paradigmsimilar to that for the pulse rate musical intervals describedabove. While the distances between the stimulated neuraltargets is not known, adjacent electrodes of the Nucleusimplant are spaced by 0.75 mm. Thus, switching from electrode19 to 18, from 19 to 17, 16, or 15 represented interelectrodedistances of 0.75, 1.5, 2.25, and 3.0 mm, respectively.76A second informal pilot experiment was intended todetermine whether covariation of place and rate of stimulationmight yield more certainly identifiable or more distinctlyperceived musical pitch intervals than those resulting solelyfrom variation of pulse rate on single electrodes. In thisexperiment, the lower pulse rate (100 pps) of a 5th was playedon an apical electrode, and the pulse rate of the upper noteof the interval was played on a second electrode located 2.25mm basalward, a distance corresponding to approximately one—half octave in the frequency region of 1 kHz in the normal ear(Greenwood 1961, 1990). The pulse rate of the upper note wasvaried randomly, from a semitone to an octave above 100 pps.The subjects were asked to label the intonation quality of theintervals as “flat”, “in tune”, or “sharp”.RESULTSThe results for the four musical intervals for eachsubject were plotted in Figures 11 to 14 as 3—categorycomplementary labelling functions, which show the percentageof items labelled as “flat”, “in tune”, and “sharp” for eachstimulus interval size. The ratios between the upper andlower pulse rates of each interval are shown along thehorizontal axis of each plot. Adjacent ratios along thehorizontal axis represent a distance of one semitone. Thusthe location, along the horizontal axis, of the maximumpercentage of “in tune” responses indicates the pitch change77favoured by each subject for a given pitch interval. Theslope and the width of the “in tune” functions reflect thetolerance of each subject for varying degrees of mistuning,and the height reflects the consistency of labellingbehaviour. The consistency of the judgements made by asubject, as well as the tolerance for mistuning, would beexpected to depend on the salience of the pitches heard, onthe ability of the subjects to label musical pitch intervals,and on the accuracy of their long—term memory for the correctinterval size.Musically trained subjects with normal hearing, whenperforming similar tasks with musical tones, produce functionsanalogous to those in Figures 11-14, but with very steepgradients and narrow, high maxima at very specific locationsalong the X-axis (Burns and Ward 1978).To facilitate direct comparison of the subjective sizesof the four intervals in Figures 11-14, the calculatedsubjective midpoints (PSEs) of each set of data (Table 2) wereplotted against the physical sizes of the target intervals (insemitones) in Figure 15, together with the standarddeviations. The bold horizontal lines in Figure 15 indicatethe correct size of each interval.At the low pulse rates tested, the interval sizesestimated by the subjects for the minor 3rd, the 4th, 5th, andmajor 6th compared favourably with the frequency ratios whichcharacterize analogous acoustical musical intervals (verticallines across each plot in Figures 11-14, horizontal bold lines78B. 5TH AT 100 PPS ON 3 ELECTRODESELECTRODE S-7 S.D. S-10 S.D. S-il S.D. MEAN MEAN S.D.18 6.64 (1.14) 7.64 (.75) 8.54 (1.04) 7.61 (.98)12 6.48 (1.23) 7.74 (.57) 7.22 (1.00) 7.31 (.93)5 7.18 (1.51) 7.22 (.49) 6.64 (1.31) 7.01 (1.10)MEAN 6.77 (1.29) 7.53 (.60) 7.63 (1.12) 7.31 (1.00)TABLE 2INTONATION QUALITY JUDGEMENTS*POINTS OF SUBJECTIVE EQUALITY (PSE) AND STANDARD DEVIATIONSA. FOUR INTERVALSINTERVAL S-7 S.D. S-10 S.D. S-li S.D. MEANm3rd 3.18 (.64) 3.02 (.53) 2.90 (.91) 3.034th 5.20 (1.17) 5.30 (.81) 4.50 (1.38) 5.005th 6.57 (.85) 7.16 (.81) 5.92 (1.70) 6.55M6th 8.42 (1.03) 9.82 (1.29) 9.02 (1.37) 9.09MEAN (1.17) (.86) (1.34)MEAN S.D.(.69)(1.12)(1.12)(1.23)(1.04)3C. 5TH AT PULSE RATESPPS S-7 S.D. S-i0 S.D. S-li S.D. MEAN MEAN S.D.81 6.84 (1.27) 8.10 (.91) 7.92 (.85) 7.62 (1.01)163 6.57 (.85) 7.16 (.81) 5.92 (1.70) 6.55 (1.12)326 5.82 (.85) 4.58 (.83) 4.88 (1.40) 5.09 (1.03)MEAN 6.41 (.99) 6.61 (.85) 6.24 (1.32) 6.42 (1.05)*Legend: S= Subject; m3rd and M6th = minor 3rd and Major 6th.7980z600Figure 11. Intonation quality judgements: interval ofa 5th. Percentage of items labelled, “flat”, “in—tune”,and “sharp”. For all stimulus items, the lower noteremained fixed. The ratios between the upper and lowernotes of the interval, indicated along the X—axis,progress (in semitone steps) from a semitone to anoctave above the lower note. The musically correctfrequency ratio is marked with a vertical line acrossthe plot. Apical electrode. Pulse rate of lower note163 pps. S= Subject; E= Electrode.60)_ 40z! ::1001.12 1.19 1.26 1.33 1.41 1.50 1.59 1.RATIO OF UPPER TO I.OWER PULSE RATE- FLAT -v-IN-TUNE-. SHARPza0I.zUaaaa0I—aUaIaa0D0I-.aVawa8010080604020RATIO OF UPPER TO LOWER PULSE RATE-6- FLAT -v- IN-TUNE -S SHARPFigure 12. Intonation quality judgements: minor 3rd.Pulse rate of lower note 137 pps. Legend as in Figure 11.815wCD0UawC0Ua.inIz‘US‘U‘3C=U0Iz‘U1‘Ua-a- FLAT -v- IN-TUNE e- SHARPFigure 13. Intonation quality judgements: interval of a4th. Pulse rate of lower note 145 pps. Legend as inFigure 11.RATIO OF UPPER TO LOWER PUlSE RATE8216080Iz6O‘9aa04oV2010080z60t9CC40201008060CCC040200-a FLAT —V--IN-TUNE-S-SHARPFigure 14. Intonation quality judgements: major 6th. Pulserate of lower note 127 pps. Legend as in Figure 11.83,11inu.Jz0ILUV.)LUNV.)7Figure 15. Intonation quality judgements for four musicalinterval sizes, from a major 6th to a minor 3rd. Triplets ofbars and markers for each interval size represent the results,from left to right, for Subjects 7, 10 and 11, respectively.Markers indicate subjective interval size (point of subjectiveequality). Vertical error bars show ± one standard deviation.Correct interval size is indicated by bold horizontal lines foreach interval size.MAJOR 6TH 5TH 4TH MINOR 3RDMUSICAL INTERVAL84in Figure 15). For example, the interval of a 5th is normallycharacterized by a frequency difference of 7 semitones, sothat tones comprising a 5th are related by a frequencydifference of* 27/12— for a frequency ratio of 1:1.5, where f0 equals the frequencyor pulse rate of the lower note. For S—7 and S—b, thelocation of the maxima of the “in tune” functions (Figure 11)and PSEs (Figure 15) for the 5th were at or near this ratio ornumber of semitones, when the pulse rate assigned to the lowernote was 163 pps. These “in tune” functions showed maxima inthe region of 6-7 semitones (with PSEs of 6.57 and 7.16 for S7 and S-b, respectively), and steep slopes on either side ofthe maxima. The slopes for the “flat” and “sharp” functionswere similarly steep, progressing from near—lOO% to 0% over aspan of two or three semitones. Near the target ratio, therewas a narrow region of overlap between the “flat” and “sharp”labelling functions, reflecting a region of uncertainty. The“in—tune” function for S-il (Figure 11) yielded a broader,less distinct peak, with more gradual slopes, and a PSE of5.92 (Figure 15), as well as a standard deviationapproximately twice that of the two other subjects. This mayreflect the difficulty this subject encountered in making thejudgements. It is, of course, also possible that S-il wouldhave yielded similar results with acoustical signals, prior todeafness.85For the major 6th, the results for the three subjectsvaried (Figures 14 and 15). Subject 7 tended tounderestimate, and S—b to overestimate the size of thisinterval. Results for the latter subject showed a wide “intune” function spanning approximately three semitones, withoccasional “in tune” labelling of octaves, 4ths, and 5ths.This subject, however, claimed not to remember the test tune(the first two notes of “My Bonnie Lies Over the Ocean”) verywell. This uncertainty of S-lO (for whom the standarddeviations for all other intervals tended to be smaller thanfor the other two subjects) appeared to be reflected in arelatively larger standard deviation than that which thissubject obtained for the other intervals. For the interval ofa 4th (Figures 13 and 15), 5-11 underestimated the size of theinterval by approximately one semitone, while S-7 and S-icachieved PSE5 very close to the target interval. For theminor 3rd (Figures 12 and 15), all three subjects yielded PSE5within close proximity of the target interval.Thus, it is evident from Figures 11-14 and the PSEsplotted in Figure 15 that the size of the subjectiveintervals, for each of the musical intervals tested, wasgenerally within one semitone of the target size of theinterval. Subjects were more accurate in estimating the sizeof smaller intervals, such as the minor 3rd and the 4th, thanof the larger intervals, such as the 5th and the major 6th.Thus, subjects showed a smaller amount of constant error (thedifference between the PSE and the target interval size) and86smaller mean standard deviations for the narrower intervals(Table 2).For all intervals, subjects and conditions, shifts in thepreferred size of the interval from one block of trials to thenext were common. Thus, while a pulse rate ratio of 1:1.5might be considered “in tune” for one block of 60 stimuli, thenext block might show a definite preference for an intervalsize one or two semitones larger or smaller.A comparison of the data obtained for the 5th onelectrodes located in the apical, basal, and intermediateregions of the electrode array is shown in Figure 16 assubjective size (PSE) of the interval, plotted against thephysical size of the interval, together with standarddeviations. The pulse rate of the lower note was 100 pps forall three electrodes. The size of the subjective interval ofa 5th for our subjects closely approximated the correctinterval size (7 semitones) for normal—hearing subjectslistening to acoustical stimuli. Subjects 7 and 11 showed aslight trend towards increased standard deviations asstimulation was moved basally, whereas S—b achieved a lowerstandard deviation on the basal, rather than the apicalelectrode. Subject 11 appeared to prefer a physically larger5th as stimulation was moved apically. All subjects commentedthat apical stimulation sounded more “musical” than basalstimulation, and that the tasks were more difficult on basalelectrodes.87‘—.9V.,UIz0IUIUIFigure 16. Intonation quality judgements for three subjects foran interval of a 5th, played on three different electrodes.Triplets of bars and markers for each subject, represent theresults, from left to right, for Electrodes 5, 12 and 18,respectively. Markers indicate the subjective interval size(point of subjective equality). Vertical error bars show ± onestandard deviation. Correct interval size equals 7 semitones.Pulse rate of lower note: 100 pps.SUBJECT 7 SUBJECT 10 SUBJECT 1188The intonation quality judgements for a 5th transposedinto three different octaves, from 81 to 326 pps, are shown inFigure 17. Again, only the calculated PSEs and standarddeviations are plotted. For all three subjects, thesubjective interval of a 5th was within approximately onesemitone of the physical 5th (pulse rate ratio of 1:1.5,equivalent to 7 semitones) when the lower note of the intervalwas 81 or 163 pps. However, when the lower note of theinterval was 326 pps, all subjects showed a preference for aphysically smaller pulse rate ratio. For S-1O and S-li, thesubjective interval of a 5th equalled, in physical terms,approximately 5 semitones, when the lower note of the intervalwas 326 pps. The preference for physically smaller intervalsat higher pulse rates was most marked for S-b and S-il.While subjects complained about the difficulty of assigningintonation quality labels to intervals at both the lowest andhighest pulse rates, there were no systematic shifts instandard deviations between these conditions.The results of the two informal pilot experimentssuggested that subjects were able to assign intonation qualitylabels also to pitch intervals resulting from changing thelocation of stimulation (i.e. by switching from one electrodeto another, at a constant pulse rate). Only two subjects wereassessed. When electrode 19 served as the lower note of theinterval, the electrode yielding an optimal 5th was located1.5-3.0 mm (mean 2.25 mm) in a basal direction. When pulserate and electrode were covaried, intonation quality89LI.Iz0ILULUr%JV•)Figure 17. Intonation quality judgements for three subjects foran interval of a 5th, using three different pulse rates for thelower note of the interval. Triplets of bars and markers foreach subject represent the results, from left to right, forlower note pulse rates of 81, 163, and 326 pps. Markersindicate subjective interval size (point of subjectiveequality). Vertical error bars show ± one standard deviation.Correct interval size equals 7 semitones.SUBJECT7 SUBJECT1O SUBJECTII90judgements appeared to become very difficult, and subjectscomplained that this sounded “like each note was being playedon a different instrument”.DISCUSSIONCategorization of musical interval size is a difficulttask for musically untrained normal—hearing subjects, evenwhen the stimuli are musical tones. Musicians have been shownto be more proficient at making fine intonation qualitydiscriminations than musically inexperienced subjects.Wapnick, Bourassa, and Sampson (1982), for example,demonstrated that musicians were able to categorize intervalswhich were only one—fifth of a semitone flat or sharp. Burnsand Ward (1978) showed that both the absolute values of thedifference limen for interval size and the variability of thedata were much greater for musically inexperienced subjects.In spite of the difficulty of the task, all three of oursubjects were able to categorize electrical pulse rate pitchintervals with a reasonable degree of accuracy andconsistency. The PSEs showed that the subjective intervals,as represented in familiar tunes stored in the long—termmemory of these deaf subjects, corresponded closely to themusical target intervals, suggesting that subjects wereresponding to the ratio properties of the stimuli, and notmerely to their ordinal relations. The fact that the PSEs didnot always coincide precisely with the correct frequency ratio91is not surprising, as even trained musicians judge musicalintervals with a degree of constant error.Shifts in the preferred size of intervals from one blockof trials to the next were common in our data. This also doesnot necessarily imply weakness of pitch. Musically untrainedsubjects have no verbal structure for describing musicalinterval sizes (either by such names as 4th, minor 3rd, or bya sol—fa scale), and are thought to use different and lessefficient strategies than musically trained subjects in makingfrequency ratio discriminations (Siegel and Siegel 1977). Infact, small intrasubject shifts in subjective responsecriteria in musical interval labeling tasks have also beenreported with musically sophisticated listeners (Burns andWard 1978). Because musically untrained listeners judgeinterval size more on the basis of pitch height than on thebasis of cues associated with target frequency ratios, suchlisteners are less precise and less consistent than musicallytrained subjects, when judging musical interval size.Musically untrained subjects can also be expected to bemore easily influenced by context effects. For example, whenthe randomization of stimulus intervals resulted in thepresentation of several intervals which the subjects labelledas obviously too small (“very flat”), subjects appeared to bemore ready to label almost any larger interval as “in tune”.Preliminary results of Eddington et al. (l978a, l978b)with melody recognition via electrical stimulation of deafears suggested that the frequency ratios required to elicit92the correct musical pitch interval might vary with thelocation of the stimulating electrode and hence with thepopulation of nerve fibers excited. The PSE5 for thesubjective interval of a 5th for electrodes in differentcochlear regions, in our experiments, fail to support thisnotion. While our subjects reported that the intervals onbasal electrodes sounded less musical and were more difficultto judge, these difficulties were not reflected in anyconsistent trends in the size of the standard deviations. Fortwo subjects (S—7 and S-il), the standard deviations showed aslight increase as stimulation moved basally, whereas for oneother subject (S-b), the standard deviations showed a slightdecrease. Thus, at low pulse rates, the musical intervalsheard by these subjects appeared to correspond closely to thesubjective musical intervals heard by normal—hearing subjects,regardless of the location of the stimulating electrode.A greater than normal increase in pitch for a givenincrease in pulse rate may be suggested by the performance ofS—b and S-il at higher pulse rates. When the lower note ofthe interval was a pulse rate of 81 pps, all three subjectsdemonstrated PSE5 within one semitone of the target interval(7 semitones). However, when the lower note of the intervalwas set to 326 pps, the subjective interval of a 5th, forthese two subjects, corresponded to a pulse rate ratioequivalent to approximately 5 semitones (in musical terms, aninterval of approximately a 4th). Rapid increases in pitchwith increases in electrical pulse rate or frequency have been93anecdotally reported by other investigators (Shannon 1983).Similarly, Tong et al. (1979) anecdotally reported that pulserates above 250 pps were consistently described as havingpitches above that of the highest note on the piano.Thus, the results for S-b and S-il suggest that, over arange of pulse rates from low to high, equal ratios of pulserates did not yield subjectively equivalent musical intervals.It is possible that, at higher pulse rates, these subjectswere unable to extract the musical pitch and to apply theirmusical interval sense, perhaps because of an unpleasantness,shrill timbre, or a lack of musical quality of stimuli. Largeincreases in pitch for physically small frequency ratios werealso demonstrated in normal—hearing musical subjects byAttneave and Olson (1971), in a musical interval adjustmenttask, when the frequency of the higher tone exceeded the upperlimit for musical pitch (approximately 5000 Hz). Theseauthors suggested that subjects were unsuccessfully attemptingto apply musical standards to nonmusical tones.Musical pitch, or the capacity to identify musicalintervals, in normal—hearing ears, may have an upper limitsimilar to that of neuronal phase—locking. In the normal ear,action potentials in auditory neurons are elicited byunidirectional deflections of the basilar membrane, and occurwithin a restricted time window relative to the stimulatingwaveform. Thus, cochlear nerve fibers in the cat are known topreserve, in their temporal discharge patterns, informationregarding the temporal fine structure of the stimulus waveform94for frequencies below approximately 5000 Hz (Rose et al.1967; Hind, Anderson, Brugge, and Rose 1967; Sachs and Young1980; Greenberg, Geisler, and Deng 1986). In the cat andsquirrel monkey, phase—locking has been shown to remain goodand fairly constant up to 2 kHz, and to decline at higherfrequencies, until it is no longer detectable at 5-6 kHz. Inother species, such as some rodents (Palmer and Russell 1986),phase—locking decreases above 600 Hz, and is not detectableabove 3.5 kHz. The upper limit for phase-locking in humanshas not been established. The saturation of firing rate ofindividual fibers is reported to result from filtering by thehair cell membrane (Palmer and Russell 1986), from hair cell-to—neuron transmission, and from the refractory period of theneuron (Rose et al. 1967). The deterioration of theprecision of neural timing information with increasingfrequency has been attributed to the increasing temporalvariance or standard error of neural phase—locking, relativeto the stimulus period, at higher frequencies (Goldstein1978), or to a decrease in the alternating current (a.c.)component of the hair cell response, relative to the steadydepolarization (Palmer and Russell 1986). With acousticalstimulation, individual fibers do not fire on every cycle ofthe stimulus or at precisely the same point in every effectivehalf—cycle, and the modal values of interspike intervals occurat integral multiples of the stimulus period. Individualfibers do not generally achieve firing rates in excess of 200spikes per second.95With electrical stimulation, maximum firing rates havebeen reported to be much higher than with acousticalstimulation (Moxon 1965, 1971). At low electricalfrequencies, all neurons within the suprathreshold portion ofthe electrical field fire synchronously in response to everycycle of the electrical waveform (Glass 1983; Hartmann, Topp,and Klinke 1984), up to about 500 Hz (van den Honert andStypulkowski l987b). Javel et al. (1987) reported saturationdischarge rates that usually equalled electrical pulse ratesup to at least 800 pps. The alternate depolarizations andhyperpolarizations of the neuronal membrane, resulting fromthe electrical signal, are opposed by accommodation processes,which work to restore the resting transmembrane voltage(Clopton et al. 1983), and place an upper limit on the firingrate of individual neurons. The absolute and relativerefractory periods have been reported to be approximately .3insec and 5 msec, respectively (van den Honert and Stypulkowskil987b).Animal data using electrical stimulation have shown ahigher degree of synchronization than that observed withacoustical stimulation of the normal ear (Glass 1984; Javel etal. 1987). The narrow phase angle of neuronal responses tolow electrical frequencies (Parkins 1989) results in a smallerdegree of temporal variance of neural responses than thatobtained with acoustical stimulation (Hartmann, Topp, andKlinke 1984; van den Honert and Stypulkowski 1987b). Theupper limit of phase-locking to electrical signals has not96been determined with certainty. While some studies have shownan upper limit of 2-3 kHz (Loeb, White, and Jenkins 1983;Merzenich 1983), others have reported phase-locking up to atleast 12 kHz (Glass 1984). Although phase-locking has beendemonstrated to be statistically significant at highelectrical frequencies, the firing rates at these frequenciesappear to be determined not by the stimulus frequency, but byinteractions between the refractory period of the neurons andthe amount of charge delivered during the excitatory portionof the stimulus waveform. As pointed out by Parkins (1989),these responses at high frequencies are, in effect, thresholdresponses, with poor synchronization to all but the firstpulse of each burst.It is reasonable to assume that the properties of theelectrically elicited auditory nerve fiber responses in ourimplanted subjects would be similar to those observed inanimal experiments. In view of the demonstrated precision oftemporal patterning of neural responses, as shown inphysiological experiments, and the absence of a mechanical,spectral, place analysis with electrical stimulation of deafears, as well as the close relationship between the timing ofelectrical stimuli and the neural responses, our resultssuggest that the auditory nervous system is capable of basingmusical interval identification on interspike intervalinformation. Temporally mediated pitches resulting frompulsatile electrical stimulation thus appear to besufficiently salient to support musical interval perception,97at least at low pulse rates. Salient, in this sense, isroughly defined as “noticeable”, in that salience of aperceptual feature is an important determinant of whichfeatures are selected or emphasized by listeners, even whenother features are perfectly discriminable (Miller andCarterette 1975).Our results, while providing evidence for temporallymediated musical pitch, do not imply that place mechanisms areirrelevant in musical pitch perception. The consistentobservations of our subjects regarding the more pleasant andmusical sound quality of the apical electrodes suggest that acongruence of place and timing information may be important.The informally assessed ability of subjects to assignintonation quality labels to musical intervals which wererepresented solely by changes in cochlear place of stimulationalso suggest a role by place mechanisms.It is recognized that the informal pilot experiment withplace pitch, by means of electrode switching, represents butthe crudest attempt at estimating the optimal cochlear placerepresentation for a musical interval. Firstly, it is notpossible to provide a continuous gradation of place ofstimulation, due to the fixed location and coarse spacing(0.75 mm separation) of the electrodes. Secondly, it is well-known that electrical stimuli activate neurons over aconsiderable range of characteristic frequencies, due tocurrent spread within the cochlea. This current spread may besymmetrical or asymmetrical about the stimulating electrodes,98depending on the electroanatomy of the residual cochlearapparatus. Thirdly, the location of the responding neuralpopulation cannot be determined, and may also be asymmetrical,especially in ears with irregular distributions of nervesurvival.In spite of these limitations and the preliminary,informal nature of the data for these two subjects, the meandistance judged optimal for a 5th (2.25 mm) is in fairagreement with the Bekesy-Skarstein cochlear map, as fitted bythe frequency-position function of Greenwood (1961, 1990).According to this function, the interval from 1 kHz to 1.5 kHz(a 5th), is represented by a distance of 2.58 mm. Similarcalculations show that, in the 1 kHz region of the cochlea,1.5 inn represents a distance corresponding to approximately amajor 3rd, 2.25 mm to approximately one-half octave, and 3.0mm to approximately a minor 6th (Greenwood 1994, personalcommunication)The substantial differences between subjects in theability to extract musical pitch information, especially athigher pulse rates, is not surprising, in view of the probableunnaturalness of the sound quality obtained with electricalstimulation. Large individual differences in pitch perceptionhave also been reported in acoustical experiments with normalhearing subjects. Risset (1978), for example, showed thatsounds which increase in shrillness while decreasing infrequency may be judged as increasing or decreasing in pitch,depending on the relative weights attached to each of thesecues by individual observers.99100EXPERIMENT IV: INTERVAL RECONSTRUCTIONWhile psychologists of music have sometimes beenskeptical about the ability of untrained subjects toparticipate in experiments dealing with musical pitchperception (Siegel and Siegel 1977; Burns and Ward 1978),other investigators have shown that, given appropriate tasks,musically unsophisticated subjects are able to perform withremarkable accuracy and consistency, and to make valuablecontributions to the understanding of musical perception(Sloboda and Parker 1985). Thus, even though such subjectsare not familiar with the vocabulary for the formal naming andprecise categorization of musical interval size, this does notnecessarily mean that they have no appreciation for correctinterval size. Dowling (1978) showed that even for musicallynaive subjects, long—term memory of familiar tunes is based onstable representations of interval size, and not merely ongeneral melodic contour.The importance, for nonmusical subjects, of a familiarmelodic context was demonstrated by Attneave and Olson (1971).They showed, in their second experiment, that musically naivesubjects transposed on a logarithmic or musical frequencyscale when asked to reconstruct a very familiar melodypresented in one octave to other frequency regions. Musicallynaive subjects, in their first experiment, had been shown to101be unable to perform the transposition task with simple,artificially constructed 2—note melodies. In a musicalinterval adjustment task, Elliot, Platt, and Racine (1987)showed that both musically experienced and musicallyinexperienced subjects appeared to have internalized standardsfor the consistent intonation of musical pitch intervals, butthat these standards were better—developed in the experiencedsubjects, who were more accurate and more consistent in theirfrequency settings. The ability of deaf subjects to adjustelectrical pulse rates in the reconstruction of musical pitchintervals abstracted from familiar melodies, stored in long—term memory, has not been previously investigated, and couldprovide additional evidence of temporally mediated musicalpitch.METHODSIn the previous experiment, subjects labelled theintonation quality of a variety of musical intervals. Thesubjective sizes of the musical intervals, which consisted ofratios of pulse rates, appeared to be similar to the ratios offrequencies which characterize musical intervals heardnormally. However, additional evidence for the salience ofmusical pulse rate pitches could be obtained using the methodof adjustment. In Experiment IV, subjects were required, byadjusting a pulse rate on a single electrode, to reconstruct102each of three musical pitch intervals: a 5th, a 4th, and aminor 3rd.The subjects were provided with a fixed reference oranchor “note” which represented either the upper or the lowernote of the interval. When the upper note of the interval wasbeing adjusted, the lower note (93 pps) served as the fixedanchor, and the “correct” targets were 140, 125, 111 pps forthe 5th, 4th, and minor 3rd, respectively. When the lowernote of the interval was being adjusted, the upper note (140,125, lii pps for the 5th, 4th, and minor 3rd, respectively)served as the fixed reference, and the target pulse rate was93 pps. The initial pulse rate of the variable note wasselected at random by the computer, at a pulse rate somewherebetween one octave above and one octave below the target pulserate.The musical intervals under examination were abstractedfrom melodies well-known to individual subjects. Subjectswere instructed regarding the tune and the interval underexamination. For the 5th, the tune for all subjects consistedof the first 4 notes of “Twinkle, Twinkle, Little Star”. Forthe 4th, the tunes consisted of the first 3 or 4 notes of“Away In A Manger” or “0 Christmas Tree”. The minor 3rds wereabstracted from the first 3 notes of “0 Canada”, the first 8notes of the chorus of “Jingle Bells”, or the first 3 notes of“Jesus Loves Me, This I Know”. It will be noted that each ofthese excerpts consisted of several notes, but only twopitches. Subjects were told to rehearse covertly these103excerpts, and to take careful note of the pitch change betweenthe upper and lower notes of the interval. Subjects wereinformed that they would be provided with a fixed note whichwould have the “correct” pitch, and a variable note, of whichthey would be required to adjust the pitch. In one set oftrials, the lower note of each interval was fixed, and theupper note was the variable note. In a second set of trials,the upper note of each interval was fixed, and the lower notewas the variable pulse rate. Thus, the adjustments of theupper note and the lower note of the intervals wererepresented by an equal number of trials. The fixed and thevariable notes were specified by keyboard control. Subjectswere, of course, informed whether the upper or the lower noteof the interval would be placed under their control. Theywere instructed to adjust the pitch of the variable note,using the “+“ and”—” keys of the computer keyboard, until thepitch change, when switching between the fixed and thevariable note, approximately matched their memory of the sizeof the musical interval in question.The subjects accessed the first note of the test intervalby pressing the number “1” on the computer keyboard, and thesecond note by pressing “2”. In the case of ascendingintervals, note “1” represented the lower note of theinterval, and note “2”, the upper note. In the case ofdescending intervals, note “1” was the upper note, and note“2” the lower note. Thus, when the upper note of an ascendinginterval was to be adjusted, the subject first pressed “1”,104and heard a repetitive pulse train (500 insec on! 500 msec of f)fixed at 93 pps, and then pressed “2”. Pressing “2” resultedin the subject hearing a repetitive (500 msec on! 500 msecoff) but variable stimulus, at a pulse rate selected at randomby the computer (anywhere from an octave above to an octavebelow the target pulse rate). This rate was then adjusted bythe subject.Keyboard control, by the examiner, permitted selection ofeither note “1” or note “2” as the fixed pulse rate. Whennote “1” was selected as the fixed pulse rate, note “2” wasthe variable pulse rate, and vice versa. The pulse rate ofthe variable note was adjustable within a range from oneoctave above to one octave below the target pulse rate. Thisrange was computed automatically by the software, uponspecification of the target ratio by the experimenter. Thus,for the 5th, when the lower note was 93 pps, specification ofa ratio of 1:1.5 resulted in a target pulse rate of 139 pps,and a range from 70 pps to 279 pps, when the upper note of theinterval was being adjusted. When the lower note of theinterval was being adjusted, calculation of the target andrange of the variable lower note of the interval wasaccomplished by specification of the inverse of the 1:1.5ratio (i.e., .67:1).Subjects were encouraged to bracket the target pitch, byalternately adjusting the pitch of the variable note slightlytoo high and slightly too low for the interval, beforedeciding on an optimal pulse rate setting for the variable105note. They were permitted, at any time, to review, insequence, the anchor pulse rate and current pulse rate settingof the variable note, and to make further adjustments to thevariable note until an optimal adjustment was achieved. Theywere also encouraged to stop stimulation for at least 5seconds prior to reviewing their final adjustment of thevariable pulse rate. Subjects indicated their satisfactionwith the pitch adjustment by pressing the <Enter> key on thecomputer keyboard. This resulted in the pulse rate valuebeing stored in computer memory, and in the initiation of thenext trial.For each interval tested, the subjects completed aminimum of 20 adjustments of the upper note and 20 adjustmentsof the lower note. A minimum of 10 practice trials precededformal measurements for a given target interval. Todiscourage the use of identity matches as a platform for afixed number of increments or decrements, the magnitude of thechanges in pulse rate effected by pressing the “+“ and “—“keys was randomized between .5 and 1.0 semitones by thecomputer. Pulse rates delivered to the subjects were roundedoff to the nearest integer value. Subjects were unable to seethe display of stimulus parameters on the computer monitor.No exemplars were provided. Subjects were permitted to stopstimulation at any time, by pressing the “P” (Pause) key onthe computer keyboard.A typical session lasted 2-3 hours. Only one intervalwas examined per session. Mean pulse rate settings were106calculated, by converting each pulse rate value to alogarithm, computing the mean logarithm for the adjustment ofthe upper and lower notes of the intervals, and calculatingthe antilogs of these mean logarithms. The ratio between eachadjusted and reference pulse rate was used to compute theinterval size, in semitones, of each adjustment. Thesenumbers were then used to calculate mean interval sizes andstandard deviations. The accuracy of the adjustments wasfurther examined by determining the percentage of adjustmentswhich were 2 semitones or less from the target pulse rate.RESULTSThe mean pulse rate adjustments, interval sizes, andstandard deviations of the adjustments for each subject andpitch interval are presented in Table 3. The reference andtarget designations in Table 3 (Ref/Target) represent,respectively, the fixed reference pulse rate and the targetpulse rate specified by each musical interval. When subjectsadjusted the upper (or lower) note of the interval, the lower(upper) pulse rate served as the reference. The mean pulserates (in pps) and mean interval sizes (in semitones) obtainedby adjustment of the upper and lower notes of the intervalsare shown in the “Upper” and “Lower” rows of Table 3, and thestandard deviations (in semitones) are shown in parentheses.The data for the adjustment of the upper and lower notes ofthe intervals were also combined to yield an overall meanoCl)t1Cl)•CD0CDCD til-’-CDl--CDI))ti-(-i-tiH00H1——CDCDtU)FCDCl)U)U)CD——I-,ci 0 CD U)a,HWWD(,J)JHa,Mwoo0‘Da,0o..HOHWLJ.o-MUi.OOlt’3 HWC-(.JI’JWMl’JO-0Ha,CDHD) ci w H H H Cl) H 0 Cl) H HlOa,‘.DWu,wa,H.t..1-.-)ooa,a,wt’3t’JoCi2Cl)c<(DOCDCD tiF’-CDl--CD0)ri-i-((ttiH00H—i—CDdCD—WCU)C))U)U)CD——C, 0 CD U)JH—_It’3JJwo-H‘.0—_IJ0JHwui0HH0L’S)cJl.-JW30l-—.IU1(311.0(DZ—txjCD ri-clxitxi‘.0 wjt txjWHH‘-31:11 z ‘-3Cl)‘-3 LxiI)Lxi C) 0 z Cl)H ;oCot’CoC<(DOCDCD tiI-’-CDHCD)ri-liri-tiI-00Hu——CDCD—U)1CU)1Cl)U)U)CD Cl 0 CD U)H‘.0Ha,U1HH0lU1a,l-j:-HH 0H MOk)H0W‘.0a,coa-’HHH wawH(31CDZ03<CD ri-0 lxj‘.0 H‘-3T Cl) H 0 Cl) H H Cl)HH0000-’F-.)1.0? —o-co‘.0wCl)‘-3 Lxi w-‘i—’Ht’)F’) -Ja,-, (31(3!0”a,F-.)a,C,,HH(‘3a,C)0a,a,T H H Cl)H0H(‘3a,01.01.0C —3108interval size and overall standard deviation for eachinterval, in semitones (“Overall” row, in Table 3).The results shown in Table 3 are plotted graphically inFigure 18. The intervals under examination, (a 5th, 4th, andminor 3rd) are characterized by interval sizes of 7, 5, and 3semitones, respectively. The solid horizontal lines whichintersect the vertical axes of the plot at 7, 5, and 3semitones indicate the target size of intervals of a 5th, 4th,and minor 3rd, respectively. For S-7 and S-ic, the magnitudeof the subjective pulse rate intervals obtained by the methodof adjustment corresponded closely to both the physical,equal—tempered intervals and to the subjective interval sizes(PSEs) obtained by the intonation quality judgements inExperiment III. While Subject 11 tended to underestimate thesize of the 5th (in both Experiment III and Experiment IV),the sizes of the 4th and the minor 3rd remained relativelywell—preserved.It will be recalled that subjects were required to adjustboth (although in separate sessions) the upper and the lowernotes of the intervals. In the case of ascending intervals,the upper note represents the second note of the intervals,and in the case of descending intervals, the lower noterepresents the second note. In the memory of the subjects fora melody, the first note of any interval always temporallyprecedes the second note. Thus, it is not surprising that all3 subjects reported greater difficulty when they were requiredto adjust the first note of an interval (i.e., when the second10910—‘9LULULlz60LUZ4LUz0LUv,10.......5TH 4TH MINOR 3RDMUSICAL INTERVALFigure 18. Interval reconstruction: mean size (insemitones) of ascending intervals of a 5th, 4th, andminor 3rd, as adjusted by 3 subjects. The vertical axisshows the response intervals in semitone steps. The leftand right markers for each subject indicate the meaninterval size achieved with adjustment of the upper andlower notes of the interval, respectively. The left andright vertical error bars for each subject show ± 1standard deviation from the mean interval size with theadjustment of the upper and lower notes of the interval,respectively. The bold horizontal lines intersecting thevertical axes at 7, 5, and 3 semitones represent themusically correct intervals for the 5th, 4th, and minor3rd, respectively.110note of the interval was provided as the fixed reference).For example, when S—b was required to reconstruct anascending 4th, using the first three notes of “Away in aManger”, he reported that this was not a difficult task whenhe was provided with the first note. However, when he wasprovided with the second stimulus (i.e., the upper two notesof the interval) as the fixed pulse rate, and was required toadjust the first (lower) note, he complained that this wasmore difficult because this required him “to think backwards”.For this interval, the relatively greater difficulty S-bencountered in this condition appeared to be reflected in alarger standard deviation (1.96 sexaitones), compared to thatwhich he obtained with adjustment of the upper note of thesame interval (0.81 semitones). Similarly, Subject 10demonstrated slightly greater intertrial variability when headjusted the upper (first) note of the descending minor 3rd in“Jesus Loves Me, This I Know” (0.84 and 0.63 semitones foradjustment of the first and the second notes of the melody,respectively). This effect, however, was not evident in hisdata for the ascending 5th (perhaps because this interval wasthe last to be tested with S-b), or in the data for the othersubjects. For S—li, the standard deviations were consistentlylarger when she adjusted the upper note of the intervals (mean1.75 and 1.30 semitones for adjustment of the upper and lowernotes, respectively), possibly due to a deterioration of pitchstrength for this subject at pulse rates above 100 pps. ForS-7 and S—b, standard deviations tended to be smaller for the111minor 3rd than for the larger intervals (mean .88 semitonesfor the minor 3rd, 1.36 for the 4th, and 1.15 for the 5th).In general, S-7 and S-b tended to adjust intervals somewhatsmaller when they adjusted the upper note than when theyadjusted the lower note.The percentage of pulse rate adjustments which werewithin close proximity of the target (Figure 19) was greaterfor the minor 3rd than for the two larger intervals. For allinterval sizes, S—il achieved a smaller percentage of closeapproximations than the other subjects.DISCUSSIONThese results show that some subjects with cochlearimplants are able to reconstruct musical intervals frommelodies stored in long-term memory, by adjusting electricalpulse rates. The ratio properties of the mean pulse rateadjustments of S-7 and S-b for all three intervalscorresponded closely to the frequency ratios whichcharacterize musical intervals for musical subjects withnormal hearing. These ratio properties, in general, alsocorresponded closely to the subjective interval sizesdemonstrated in the intonation quality judgements ofExperiment III. Thus, the results of the intonation qualityexperiments and of the production experiments are in generalagreement.112w0zwC)Ui0MUSICAL INTERVALL1SEMITO1E 2SEMITONESFigure 19. Interval reconstruction: percentage of pulserate adjustments 1—2 semitones or less from the target, forthree intervals. Subjects 7, 10, and 11.5TH 4TH MINOR 3RD113While the adjustments of the intervals of a 4th and minor3rd by S-il also corresponded closely to the target intervalsizes, this subject’s subjective 5th was approximately twosemitones smaller than its conventional size. For theinterval of a 5th in Experiment III, this same subject yieldeda subjective interval size approximately one semitone smallerthan the physical interval. The contraction of the 5th bythis subject could be due to a deterioration of pitch strengthabove approximately 100-200 pps, or to a confusion between a5th and a 4th. Thus, it is equally possible that S—li mighthave yielded similar results with acoustical stimuli, prior todeafness. Both the 5th and 4th are highly consonant intervalswhich bear a close relationship to each other. For example,for a 5th, an octave downward displacement of its upper note,or an octave upward displacement of its lower note results inan interval of a 4th. Comparable inversions of the notes of a4th result in an interval of a 5th. Furthermore, in somecontexts, these two intervals serve as functional equivalents(as in the “Subject” and “Answer” of a fugue). Intervals of a5th and 4th are easily confused by untrained, normal—hearinglisteners, and under some circumstances, even by somemusically trained subjects (Wapnick, Bourassa, and Sampson1982).In the presumed absence of place coding of electricalsignals in deaf ears, this pitch or interval informationalmost certainly has to be dependent upon the timing ofinterpulse and interspike intervals. Dobie and Dillier (1985)114have shown, using electrical pulse trains from which apercentage of pulses were stochastically omitted (i.e.,interpulse intervals were equal to either the pulse period orto multiples thereof), that cochlear implant subjects rankpitches on the basis of interpulse interval information,rather than on the basis of the total number of pulsesdelivered per unit time. These data, like ours, suggest thatthe human brain is able to make perceptual decisions aboutmusical pitch or musical intervals on the basis of interspikeinterval information.These findings are in agreement with the results ofacoustical experiments which have utilized stimuli designed toprovide listeners with exclusively or predominantly temporalinformation. Thus, musically trained subjects have been shownto be able to perform musical interval identification anddictation tasks, as well as interval adjustment tasks, withstimuli which consist of closely spaced, unresolvableharmonics or of sinusoidally amplitude—modulated noise (Burnsand Viemeister 1976, 1981; Houtsina and Smurzynski 1990; Pierce1991). While the ratio information conveyed by pitchesresulting from such stimuli has been questioned, and thepitches reported to be weaker than those of pure or complextones with resolved low frequency components (Houtsma 1984),others have shown that even musically naive subjects are ableto utilize this temporal information in open—set or closed—setmelody recognition tasks (Burns and Viemeister 1976, 1981;Moore and Rosen 1979).115The difficulty experienced by one presumably relativelymusical subject (S-ll) in adjusting pulse rate pitches isconsistent with the greater constant error (i.e., differencebetween the subjective and target intervals) and largerstandard deviations shown for this subject in the intonationquality judgements of Experiment III, especially for intervalslarger than the minor 3rd. This subject also evidenced apoorer performance on the closed—set melody recognitionexperiments in Experiment II. The standard deviation of thepulse rate adjustments for this subject in Experiment IV,averaged across the three test intervals, was 1.54 semitones,or 37% greater than the 1.11 to 1.14 semitone standarddeviations for S-7 and S-b. Even simple pulse rate pitchmatching during informal preliminary investigations proved tobe extraordinarily difficult for this subject, who appeared tobecome increasingly confused during prolonged efforts ateffecting subjectively satisfactory adjustments. Therelatively poorer results for this subject may reflect poornerve survival, relative to the other subjects. Ears deafenedby bacterial meningitis or labyrinthitis have been shown tohave lower mean spiral ganglion cell counts than ears deafenedby other pathologies (Nadol and Hsu 1991). A number ofinvestigators have demonstrated a positive correlation betweenneuronal integrity and performance on psychophysical tasks(Pfingst and Sutton 1983). It is also conceivable that thepoorer results for S—li reflect insult to the central auditorytracts or pitch processing mechanism. Major neurological116sequelae have been reported in 15 to 71 percent ofpostmeningitic patients (Vernon 1967).In general, the standard deviations for the minor 3rdwere smaller than those for the two larger intervals,especially for S-7 and S—iC. Smaller deviations for narrowerpitch intervals have also been reported in acousticalexperiments. Attneave and Olson (1971), for example, examinedthe ability of normal—hearing, nonmusical subjects toreconstruct the musical intervals of a familiar tune, usingpure tones, and found a smaller amount of variability for amajor 3rd (equivalent to 4 semitones) than for a major 6th(equivalent to 9 semitones). These authors attributed thisphenomenon to a possible “keynote” effect, in that the major3rd in their test melody ended on the keynote (tonic) of themelody. They suggested that intervals incorporating thekeynote might somehow be easier to adjust. Burns andViemeister (1976, 1981) tested musically trained subjects withpure tones and a variety of sinusoidally amplitude-modulatednoises, and also reported smaller deviations for major andminor 3rds than for larger intervals.The “keynote effect” cannot explain the smallerdeviations for the minor 3rd in our data, as the notes of theminor 3rd in our melodies consisted of the interval betweenthe third and fifth steps (in musical terminology, the mediantand the dominant, respectively) of the diatonic scale. It ispossible that the greater consistency observed for smallerintervals, at least in our data, resulted from the strategies117used by the subjects in the adjustment task, especially withsmall pitch intervals. Subjects were frequently observed tobring the reference and the variable pulse rates into closeproximity before effecting an optimal adjustment. While therewas no evidence that subjects were using these near—matches asa platform for a fixed number of increments or decrements, theability to match the reference and variable pulse rates couldbe expected to result in a substantial reduction in theprobability of large errors, especially for smaller intervals.It is possible that the normal—hearing subjects in acousticalexperiments utilizing the method of adjustment (Burns andViemeister 1981; Attneave and Olson 1971) employed a strategysimilar to that of our subjects.Data of Elliot, Platt, and Racine (1987), however,suggest that even when musically untrained subjects wereunable to use pitch matching as a platform for musicalinterval adjustment, smaller pitch intervals were stillcharacterized by smaller degrees of variability than widerintervals. In their protocol, identity matches were precludedby permitting subjects an adjustment range of only onesemitone. These data and ours suggest that for musicallyinexperienced subjects, small intervals may be easier toadjust than large intervals. The smaller amount ofvariability noted for smaller intervals in Experiment IV is,furthermore, consistent with the results of Experiment III,which showed smaller standard deviations for narrow intervalsthan for wider intervals.118No obvious relationship was observed between the self—reported musical history of the subjects prior to deafness andthe accuracy or consistency of their pulse rate adjustments.Subject 7 reported little interest in music prior to deafness,but yielded fairly consistent measurements on the adjustmenttask. Subject 11, on the other hand, reported a considerableinterest in music prior to deafness from meningitis, andreported having been able to play tunes on a piano keyboard,with a fair degree of accuracy, without the benefit of printedmusic. The pulse rate adjustment task appeared particularlydifficult for this subject, and this appeared to be reflectedin a larger constant error and a larger degree of variability.Subject 10 indicated a considerable appreciation for musicprior to the onset of total deafness, but had no musicaltraining and had never played a musical instrument. Thissubject described his strategy during this experiment as“getting the two notes to harmonize”. In the intonationquality judgements of Experiment III, this subject had shownconsistently smaller standard deviations than the other twosubjects. In the results of the interval reconstructionexperiment, there were no consistent differences between thestandard deviations of S-b and S-7.It is interesting to compare the postoperative emergenceof speech perception abilities in these three implantrecipients. Following 6 weeks of cochlear implant experience,S-7 and S—b scored a mean 88.3-90.5% on the auditory-onlyspeech perception test battery, consisting of CID Everyday119Sentences, Iowa Sentences, arid Bamford—Kowal—Bench (BKB)Sentences in quiet. In contrast, following six weeks ofcochlear implant experience, S—il scored a mean 58%. However,this subject subsequently showed considerable improvement, andachieved scores of 80-90% within two years of device fitting.While the delayed improvement observed for S—il is common incochlear implant recipients (von Wallenberg and Battmer 1991),the mechanisms underlying these differences between patientsare not understood. These differences may reflect differencesin neuronal survival, processing abilities, or differences inplasticity of the central auditory system. Theneurophysiological limitations of individual patients maycertainly be expected to be reflected in difficultpsychophysical tasks (Tyler, Moore, and Kuk 1989).While it could be argued that, from a musicalperspective, the data of Experiment IV lack precise ratioproperties, because of the substantial amount of variability,it is important to consider the musical inexperience of thesesubjects (Elliot, Platt, and Racine 1987). It must, inaddition, be considered that musically trained subjects alsoperform musical interval adjustment tasks with a degree ofconstant error. Furthermore, even following training andpractice, some normal—hearing, musically naive subjects havebeen shown to remain unable to perform musical intervaladjustment tasks, with pure tone stimuli (Attneave and Olson1971)120It is instructive to compare the standard deviationsobtained in Experiment IV with those reported by otherinvestigators, utilizing the method of adjustment withacoustical stimuli and normal-hearing listeners. Burns andViemeister (1981) reported a standard deviation ofapproximately one—third of a semitone for musically trainedsubjects adjusting musical intervals with pure tone stimuli.In contrast, adjustment of modulation frequencies (in theregion of 100 Hz) of sinusoidally amplitude-modulated 10 kHzpure tones or noise by the same subjects yielded standarddeviations of approximately one semitone. These findings, inaddition to those of Houtsma (1984), suggest that the pitch ofpure tones and complex tones with resolved frequencycomponents may be more salient than the pitch of stimuli whichprovide predominantly temporal information. The standarddeviations in our data for Experiment IV were somewhat greater(mean 1.27 seinitones, when averaged across intervals andsubjects) than those reported by Burns and Viemeister (1981)for musically trained subjects adjusting modulationfrequencies of amplitude—modulated noise. These differences,however, appear to be relatively small, and can certainly beexplained by the musical inexperience of our subjects, none ofwhom had a verbal structure for labelling musical intervals.Furthermore, it is well—known that the temporalcharacteristics of neuronal response patterns to electricalstimulation differ substantially from those generated in thenormal ear with acoustic stimulation (van den Honert and121Stypulkowski 1987b). In the normal ear, information regardingthe stimulus period is preserved over an array of neurons.The responses of these neurons, over a range of characteristicfrequencies, are staggered in time, due to the progressivephase lag along the basilar membrane. The responses of singleunits in a normal ear are phase—locked to the stimulus, andthese units fire stochastically and independently of eachother. Thus, even for frequencies greater than 1000 Hz,interval histograms could be expected to contain someintervals as small as the stimulus period, even though singleunits are not able to sustain firing at these frequencies. Itis believed, by some, that the central auditory nervous systemanalyzes this aggregate output of an array of stochasticallyindependent neurons. However, with electrical stimulation,all neurons within the electrical field are known to firesynchronously to every cycle of the stimulating waveform, upto repetition rates of 600-900 spikes per second (Kiang andMoxon 1972; Hartmann, Topp, and Klinke 1984). In addition,the precision of phase-locking is known to be greater withelectrical than with acoustical stimulation. Thesedifferences can be assumed to result in relatively unnaturalfiring patterns with electrical stimulation (Dobie and Dillier1985), and may be expected to result in pitch percepts whichare qualitatively different from those elicited withacoustical stimulation of the normal ear.122EXPERIMENT V: MUSICAL TRANSPOSITIONAppreciation of musical pitch, in the form of melody orharmony, requires appreciation of musical intervals. In thefrequency range over which musical pitch is perceived,musicians (and laypersons who are not tone deaf) interprettones in equal frequency ratios as equal musical pitchintervals. Thus, as long as the frequency ratio between thetones of a musical interval is preserved, the interval will bereadily recognized as a transposed musical equivalent,regardless of its location in the musical pitch range. Toexamine the changes in musical pitch resulting from electricalstimulation over a variety of pulse rates, the subjects wererequired to transpose the intervals used in Experiment IV tolower and higher pulse rates.METHODSThe methodology for this experiment was similar to thatof Experiment IV, in that subjects heard a note with a fixedpulse rate and were required to adjust a second note with avariable pulse rate. The variable note was either the upperor the lower note of a number of musical pitch intervalsabstracted from melodies familiar to the subjects. InExperiment V, subjects were required to transpose threeintervals (a 5th, a 4th, and a minor 3rd) to different pulse123rate regions, when provided with different pulse rates asreference (anchor) notes. To preclude the use of identitymatches as a basis for a fixed number of increments ordecrements, the changes in pulse rate effected by pressing the+“ and “—“ keys was randomized between 1.0, 0.5, and 0.25semitones. Step size was calculated by the formulaincrement or decrement stepsize = 2n/12where n equals 1.0, 0.5, or 0.25.Subjects were given a minimum of 10 practice trials at astandard low pulse rate. The reference and target pulse ratesof the standard were 105/125 pps for the minor 3rd, 99/132 ppsfor the 4th, and 93/140 pps for the 5th. It will be notedthat while the reference and target pulse rates differed foreach test interval, the 3 intervals nevertheless shared acommon geometric mean (114 pps, for the standard or controlpulse rate intervals). It should also be noted that thereference and targets for the control pulse rate (geometricmean of 114 pps) were either in close proximity or identicalto those used in Experiment IV. In other words, while thesubjects received practice at reconstructing the targetintervals at the control pulse rate, they received no practicein transposition. Since the sessions for Experiment IVpreceded those for Experiment V by 1—2 weeks, the practiceitems in Experiment V served to ensure that the subjects wereusing the correct musical interval for the transposition task.During the practice items at the control pulse rate, feedback124regarding the correct interval size was provided whenadjustments deviated from the target by more than twochromatic steps. Subject 11 was unable to produce consistentadjustments for the interval of a 5th at the control pulserate, and did not participate in the remainder of theexperiment.Following this re—familiarization, subjects were requiredto transpose each interval to both higher and lower pulserates, over approximately a two and one—half octave range.Subjects were tested at six geometric mean pulse rates: 81,114 (the standard), 162, 231, 326, and 466 pps, each separatedby one—half octave (a tritone or 6 semitones). These meanpulse rates were used to define the reference and target pulserates for each of the three intervals. For example, when thegeometric mean of the 3 test intervals was 231 pps, thereference and target pulse rates were 212/252 pps for theminor 3rd, 200/268 pps for the 4th, and 188/284 pps for the5th.The transposition task required subjects to reconstructthe intervals at the standard pulse rate (as during thepractice trials), and to transpose the intervals to one pulserate region below the standard, and 4 regions above thestandard. The subjects were informed that, for the experimentproper, the pulse rate of the fixed anchor note would bevaried from trial to trial, and that the transposition taskmight be compared to having the same interval sung or playedby a voice or an instrument with a different pitch range. At125each reference pulse rate, 20 trials involved the adjustmentof the upper note, and 20 trials the adjustment of the lowernote of the intervals. No limit was imposed on the amount oftime or the number of pulse rate changes required by thesubjects to complete a trial. No feedback was provided duringthis portion of the experiment. Each test session involvedthe adjustment of either the upper or lower note of one testinterval, and lasted approximately 2-3 hours, with 15-minuterest periods between each 10-20 trials.The presentation order of reference pulse rates wasrandomized from trial to trial, with the restriction that nopulse rate served consecutively more than once as the anchor.The initial pulse rate value of the variable note was randomlyselected by the computer, with maxima and minima from anoctave above to an octave below the target pulse rate. Arange restriction of the variable note was necessitated at thelowest, and for one subject, at the highest reference pulserates. The lower pulse rate limit of the computer program waspps. Thus, for example, when testing for an ascending 5thwith an upper note reference pulse rate of 99 pps, the lowerlimits of the variable note were 54 pps, rather than 33 pps(an octave below the target pulse rate of 66 pps).Observation of the subjects by the experimenter suggested thatthis range restriction did not significantly affect theadjustments of the subjects, as they tended to avoid the lowerportion of the range. At higher pulse rates, the upper limitfor S-10 was set at 900 pps, to preclude the utilization of126idiosyncratic attributes of the sounds anecdotally reported bythis subject (such as “double notes”), which could potentiallybe used to identify a region of high pulse rates. No suchrestriction was placed on the upper limits for S-7. The upperlimit permitted by the software was 1339 pps. All otherprocedural details were identical to those of Experiment IV.The treatment of the data was comparable to that ofExperiment IV, and consisted of calculation of the mean pulserate values for the adjustment of the upper and lower note ofeach interval at each reference pulse rate. This wasaccomplished by converting individual pulse rate values tologarithms, and then computing a mean logarithm for theadjustment of the upper and lower notes of each interval ateach reference pulse rate. The antilogs of the meanlogarithms then provided the mean pulse rate values. Theinterval size of each adjustment, in semitones, was computedfrom the ratios between the adjusted and the reference pulserates by the formula2n1l2f/f0orn= 12 * log(.f/.f0) / log 2where f and f0 represent the upper and lower notes of theinterval, respectively, and n represents the number ofsemitones in the interval. These interval size conversionswere then used to calculate the mean interval sizes andstandard deviations. The computations for the adjustment of127the upper note were performed separately from those for thelower note, in order to examine potential differences betweenthese two sets of data. The data for the adjustment of theupper and lower notes of each interval in each pulse rateregion were subsequently combined to yield an overall meaninterval size and an overall standard deviation. Theseinterval sizes were then compared to those of musicalintervals heard acoustically, and plotted as a function of thegeometric means of the reference and target pulse rates.For the data obtained with adjustment of the upper (orlower) note of the interval, the standard deviations reflectthe intertrial consistency of the adjustments. The standarddeviations from the overall mean interval size (i.e., thecombined data for the adjustment of the upper and lower noteof the interval), were affected not only by intertrialconsistency, but also by the agreement between the two sets ofdata. The accuracy of the pulse rate adjustments was furtherexamined by calculating the percentage of adjustments whichfell within reasonable proximity of the target rate (2semitones or less from the target).Most of the melodies used in this experiment were thesame as those used in Experiment IV. For both subjects, the5th was abstracted from “Twinkle, Twinkle, Little Star”. Theinterval of a 4th was taken from the first 3 or 4 notes of“Away In A Manger” or “0 Christmas Tree”. The minor 3rd forS-7 consisted of the first B notes of the chorus of “JingleBells”. It should be noted that all of the above intervals128were ascending intervals (i.e., the progression from the firstto the second note of the intervals was associated with anincrease in pitch). Subject 10 did not know any suitablemelodies incorporating an ascending minor 3rd, and for thisreason, was tested with a descending minor 3rd, abstractedfrom the first three notes of “Jesus Loves Me, This I Know”.In view of the large differences in the results shown bySubject 10 for the ascending 4th vis-a-vis the descendingminor 3rd, this subject was further tested with a descending4th, abstracted from the first three notes of “0 Come, All YeFaithful”.RESULTSThe mean pulse rate adjustments and standard deviationsobtained by the subjects at each reference pulse rate, and thecorresponding interval sizes, relative to the reference, areshown in Tables 4—6. These tables detail, for each referencepulse rate, the mean of the adjustments (in pps) and the meaninterval sizes in musical terms (“Semitones” column) for boththe adjustment of the upper (“Upper” rows) and the lower notes(“Lower” rows). The standard deviations (S.D.) are shown inparentheses, in semitones. In the Tables, the reference andtarget pulse rates are shown by the “Ref/Target” values. Whenthe upper (lower) note was adjusted, the lower (upper) noteserved as the reference. The “Overall” rows in Tables 4—6show the combined data (mean interval size and standard129deviations, in semitones) for the adjustment of the upper andlower notes of the intervals.The data of Tables 4-6 are shown graphically in Figures20 to 23. Because each interval in a given pulse rate regionhad different reference and target pulse rates but shared acommon geometric mean with the other intervals, theperformance of each subject was plotted as a function of thegeometric mean of the reference and target pulse rates of theintervals. In Figures 20 to 23, perfect performance, such aswould be approximated closely by musically trained subjectswith normal hearing, when listening to musical tones, would berepresented by horizontal straight lines intersecting thevertical axes of the plots at interval sizes of 7, 5, and 3semitones, representing the 5th, 4th and minor 3rd,respectively. The standard deviations obtained with theadjustment of the upper note of the intervals are plotted inFigures 20 to 23 as the upper vertical error bars, and thoseobtained with the adjustment of the lower note as the lowererror bars.Clearly, the transposition behaviour of the two subjectswas very different. Subject 7 transposed the intervals on alogarithmic frequency scale, in which equal musical pitchintervals were represented by approximately equal ratios ofpulse rates (Table 4; Figures 20 and 21). While small localdeviations from the “correct” interval sizes were evident forall three intervals, both the relative sizes and the ratioproperties of all three intervals remained well—preserved.130- Cl Oio Cl CN‘-4 rr-N orNO No‘.0(1 Nfl0 (NCI)Lfl0 CI’ 0•.00•’ 00 •Ca’In .Lfl C’) (NIirCN.U) 01 ‘.0(N (N In‘-4 rClfl 0(1 (ILnclr- ‘.0(I..Or O(1 (NU)CN U)O•i.(I (N(NI.’,0NElC)I.)U)U)0fririU)c1:: ElEl0HElHU)0El11.1ElU)I-,fri000El0U)ClInIn‘.0Cl0U)CN0U)‘.0(N0U)(NU)(NU)U)‘-I0U)U)0-t(N0U)0iCl‘-ICl0”0U)0’Elf.Z4’.OfrIicx:frIQ4)a)>CarEl4-iG)0 In N0i ‘.0.(IN ‘(N 0 U)N (NO i’ (I. . • CI)U)N 1N NN N ClIn Cl 0.It) 0’ (N-1 In 1’(1(1Lfl O’ In0L- 0O Cl N. . . . (Y)‘‘.0 (IN N0 InCl CN U)(NN (N (NCl N Cl,- (N (NIt)’j’ rN If) NClU) O’.0 N ‘.0. . . • (NClLfl ON ‘.0‘.0 U) 0(N 0(N‘.0 U) Cl‘.0 Cl -4r-1 (N (NCl0 Clr-1 in NU)0 i.0 N U)U)N -4U) N0 (N 0-I ‘Ca’-IO N -tN N ‘0 ‘-4(IN (l(N 0’Cl0 U)(N CI)NN r-4N N. o‘-4.-I Cl (N0’ 0 ‘.00 (N -IN0 ‘.0 (N(NU) ,-40‘ El0’. . .U)’.0 OLn ‘.00’ Nfri—. friCi) 111 Q4JP4CQ P4 a)PIG) P4W—:: ••— ‘-40 0 —44i 44-) caG)-- W•r4 -i fr.P4 a) El4-IPIG) 00) >U) U) 0-4 U) (N,-l U) In‘—N0Cl. -1 in(N (N (N,-; -1 ‘-IN 0’U) In‘.0(N U)U) In(IL( OInU) ‘.0Cl (N0 (1 0’‘.0 ‘.0 Lf)‘- - -N(N CNU) (NN .-1O’i If).. (NClc3’ (NClIn .4 (N(N (N(N-I ‘Ca’ ‘In1 ,4 r4‘.o-t o’.o ‘.00 (N(N N. ..U)L() 0’.0U) Cl U)t r4‘-40 (N NCl (N (Nr ,1 1InO U,0’-1 (N..(NInCl In,-1 0‘-ICl Cl ClN ‘.0,—t 1 ,—t‘.00 ‘.0 U)‘-4(N N U)0’.0 In0 N‘-4P40) P4W HG)P4W PIG)‘-‘- r40 0 — Ca-i.p Ca0)—I 0)—I ti O0) 4-iPIG) 00) > HG)U) U) 0U)‘.0InLfl’.0InLf0‘.0ClIn0’ Cl Clr1 U)r1 c rU)In (N(N .Cl0 00’ ‘ClCI) 0(N (N(N r4 Cl0’ ‘.0 U)O 0d‘.ON (NO 0’Nr-I CIa) ‘Ca’Nfl -1Cl ClN ‘Ca’-I -I0’ 0) If)U) in Nd d0‘-t CI’.0 ‘00 (0 0’Cl Lf)(N (N(N 0‘-IIn ‘ N(N U) 0d(NO (Na) ‘-t’.0 U)(N ‘rCl (Nfl ClN940) 940)940) 940)—c ‘—t:: —i0 0 r1t44J 4J Ca0)—I (D-.-I 4P4 GE 0)940) OW >U) i-lU) 0O(NN Cl‘.0131121110zwN-J<532I121110zLlLLIN-J32IFigure 20. Method of adjustment: intervaltransposition, Subject 7. Intervals of a 5th and minor3rd. Upper and lower error bars show standarddeviations for the adjustment of the upper and lowernotes of the interval, respectively. Refer to text for81 114 162 231 328 466GEOMETRIC MEAN OF REF/TARGET81 114 162 231 328 466GEOMETRIC MEAN OF REF/TARGETfurther explanation.121110Iii4TH28LUinz7N6:±fa—321Figure 21.Subject 7.Method of adjustment: interval transposition,Interval of a 4th. Refer to text for explanation.13281 114 162 231 328 466GEOMETRIC MEAN OF REF/TARGET133In contrast, the results for the ascending 5th and 4th ofS—lO (Table 5; Figure 22) showed a compression of intervalsize at pulse rates above those of the standard. For example,the subjective 5th, which is normally equivalent to 7semitones, was adjusted to a mean physical interval of 2.5 to3.5 semitones, at pulse rates above the standard. Theseintervals correspond to approximately a minor 3rd, for normal—hearing musical subjects. This tendency to compress largeascending intervals at higher pulse rates was even greaterthan that observed for the same subject in Experiment III (cf.Figure 17). At pulse rates above the standard, S-b failed topreserve even the rank order of interval sizes. In otherwords, at some pulse rates, 4ths were adjusted to be largerthan 5ths. This apparent compression of interval size wasless consistent and less marked when the subject adjusted thelower note of the intervals than when he adjusted the uppernote. At low pulse rates, however, S—b transposedapproximately on a logarithmic scale, with roughly equal pulserate ratios for equal musical intervals. The interval sizecompression did not appear to be electrode—specific, asvirtually identical results for the ascending 5th and 4th wereobtained on electrodes situated in the middle of the array, inadditional informal experiments.Subject 10 did not know any melodies incorporating asuitable ascending minor 3rd. For this reason, a descendingminor 3rd was substituted. The results showed that for thedescending minor 3rd, this subject, like S-7 for ascending134U) ‘.D 0 Q NlD N Lfl Lfl N O- c H -cc’1H C)H C\1 U)’DD O 0LflNU) (‘)C’ ‘ CU)H C)N 00 It).. . . Lfl . . .H aH Qc’) C1 OH OC’ CNOO It)El OC.)riL() H C’ H U).,• C) C’.) N H C’.) CoCo 0 H C’. C’.) H H.—0CNLflLfl cOCo C) OU)C’4 C)Q0ND OD 0 NC’IH OLfl. • . • • . C’, •-1 CN’.O ‘ O LflLC)U)a N N.DC’q CN U)C’, C’.)riEl . .. .Z co - .-i o’ r-i!ElU)Lt)0 N NC’.)D 0 cZ OoO CoC’.) It) U)OO ‘.DIt) N0 H. • • • . H • •H ‘.OC’4 c (‘, ‘.OC’) C’.)Cl C’,El cLfl it) ONH c)H H ‘H HU) H H0:l4 • —Cl) fl C’.) H 1’, C’, ‘.0 0• H C’) C’.) 0 C’, C’.)g: C,)H H H H H HElOLC)Lfl 0) c’)H C’.)> C’) N Ii) N C’.) CO C’) 0 C’) C’)H.. •. • H..Lfl’.0 .-.)N ‘.0 Lfl it)El C’)C’) 0)z 0)H 0i.-IH;•r.i • C’) C’) ‘3’ C’) 0 ‘Ci) • • .El H H H H H HU)IDO)HC’) CoO H C’,OH ‘.DH Hl O)C• ra i Eloac ou ci’‘.D’.OCO ‘.D’.D N O’.DN Nit) ‘.00 H’.0 ‘.0 0N0 ‘.0rx4 H HCl) 4JW 1%)W4W 4O) P4CI) P4Wi-i 4W G) PO) 1)4’- —a H 4’— •— H>ci 0 0 H 0 0 HE-i44-l 44J (d 4)-) (fri (1)-d a’—i 4 fri 0)—I W-IElq-P4 E (1) E--14 E Ii)0)4w 00) > ZWP4C) 00) >HIDCl) U) 0 HIDU) -iU) 0Figure 22.Subject 10.to text for121110UIzUIuz7LUMethod of adjustment: interval transposition,Ascending intervals of a 5th and a 4th. Referexplanation.13581 114 162 231 328 466GEOMETRIC MEAN OF REF/TARGET12 -111098UIz0ILUzUI4Th76543281 114 162 231 328 466GEOMETRIC MEAN OF REF/TARGET136intervals, transposed on a logarithmic scale. This suggestedthat the apparent pitch compression of the intervals might bedue either to interval size (i.e., in that perhaps smallerintervals were easier to transpose than larger intervals), orto the directional properties of the intervals (i.e.,ascending versus descending intervals). Thus, a furtherassessment of this apparent pitch compression was undertakenby requiring this subject to transpose a descending 4th. Theexperimental paradigm, reference and target pulse rates, andthe treatment of the data were identical to those used withthe ascending 4th. The results for the descending 4th (Table6; Figure 23) contrasted sharply with those for the ascending4th (Figure 22), and showed an essentially logarithmictransposition, with nearly constant pulse rate ratiosthroughout the range of pulse rates tested.The adjustment task for S-b, at high pulse rates, wascomplicated by the frequent perception of multiple pitches.Above 500 pps, S-ic frequently observed hearing “doublenotes”, consisting of both high-pitched (“like a pigsquealing”) and low—pitched (“like a cow mooing”) components,and became confused as to which of the pitches to use fortuning the interval. Both S-7 and S-b complained of aninconsistency of pitches at high pulse rates, so that when twoidentical pairs of pulse trains (i.e., exemplifying identicalratios) were presented within a few seconds of each other,subjects reported hearing two very different musicalintervals.1370 Cl U It) Cl Clr1 N C’Cl)Cl Cl 0a)N c’40 a) OCl a)C\ClOc’J OCl Lfl OOC OD CNIf).. • It). •CN N It) O C’4 CN Cl ClEl Cl-l I.)) a)0OLt) Cl CU)I-)• —% — — — —a)t) Q Cl 0 CN 0a) • a) a) .-I N .-ICl) • • •:i. -40 .—a) oO)It) 00 1K) a)C’1 Ca) .-4NU)LK) Ca) LX) Lfl0O) N NicE: Cl . Cla)Cl rlLf) a)Cl CCl ClN rN 0)a)Cl OCl C4El C’4 Cl•_H fl Cl Cl 1.0 N CX)a) 0 0) N a)0 . . .0 .-4 0 0 0H0N1.00) Cl It) 00) N ClC) occi a) Cl Lfl00’l Lfl Na) • . • . . c’i • •1.D 1.0 c’4C’4 a)CN C’10 01K) 0 C’4LK) c-iOc’4 C’4 (N(N (N (N1.0 —fri N a) N 0 00-401.00 1.0Lfl (NN .-Ia)a)Cl 1.0(N NLflO NIt) Clo . . . • c-I • •H Q l 1.0Cl 0El ON a)N 1.0H I—Ia) .-I c-I0— .-. —Cl) 0 N (N Cl Cl 0) CNCl) • • •• 0) 1.0 0) rI 1.0 c-I0 0 0 c-I 0 c-IEl ‘—(NOIt) I4)(N 0) fl(N tn 0)Clc-I(N O)Cl N (Nc-IN 0c-I • • • • • c-I • •Lfl Ci OCl a)(N ClEl 0)Cl 0 Lf)Cl 0z O),-i c-I 0 c-IH.. 0 0 0) Cl Cl 0) (N• (N Cd Cl Cl 1.0 Nz 0 ‘-1Ela) ClO(N Oa) It) a) 0)I-) 0)0 LU If) 0Cl 01.0 0)CI 0O QCN Cl Lfl 0 (NNO) 0) 0) a)rLloElCIo — —04-3m CO COEl rzOjP4Co pCO HGJQ4(O P4COP4W P4W P4W P40)El4 .-I El34 - r1QCd 0 0 -4 cd 0 0 .-IICdE-43-I4-) 3.44-) Cd E-l44.) 3.44) Cdr’ W-.-I W—4 3 0 W.-4 0)—I 14q-4P4E E W t4IP4 E 0)frWP4W OW > HWP4G) OWfl0) a) 0 a) 0138121110V..z9wVIz7wUIz32Figure 23. Method of adjustment: interval transposition,Subject 10. Descending intervals of a 4th and a minor 3rd.Refer to text for explanation.4Th81 114 162 231 328 466GEOMETRIC MEAN OF REF/TARGETMINOR 3RDHft1211UI10z3281 114 162 231 328GEOMETRIC MEAN OF REF/TARGET466139The results at the control pulse rate (geometric mean of114 pps) compared favourably with those obtained in ExperimentIV. Subject 7 adjusted the size of the 5th and 4th to besomewhat larger in Experiment V than in Experiment IV (byapproximately 1.7 semitones), whereas he adjusted the minor3rds to be somewhat smaller than those in Experiment IV. Atthe control pulse rate, the adjustments of S—b were alsosimilar to those obtained in Experiment IV. For this subject,the mean of the adjustments for all three intervals inExperiment V at the control pulse rate differed by only 0.23semitones from that in Experiment IV, and no differencesexceeded 0.47 semitones. It should be remembered that theadjustments of Experiment IV were obtained without feedback,and the close correspondence between these two sets of resultsfor S-b more closely approximates the replicability whichwould be expected of musically trained subjects adjustingmusical tones. The standard deviations obtained at thecontrol pulse rate for both subjects in Experiment V were alsosimilar to those obtained in Experiment IV.Some systematic differences in interval size wereobserved which appeared to be related to whether the subjectadjusted the upper or the lower note of the intervals (Figure24). At the lowest reference pulse rates, both subjectstended to create larger intervals when adjusting the uppernote of the intervals, resulting in a slight overestimation ofinterval size. This trend was similar for ascending anddescending intervals. These discrepancies in interval size14032ujiz0uJLI-1z-2zLU-3LULI-5-68- ASC. 5TH -- ASC. 4TH-- ASC.MINOR 332v1LUz0i-0LUu -1zUI‘-IzLULU -ULI-5-6-B- ASC. 5TH -- A5C. 4TH -G- DESC. 4TH -a-- DESC. MINOR 3RDFigure 24. Method of adjustment: mean interval sizewith adjustment of the lower note minus mean intervalsize with adjustment of the upper note, for Subjects 7and 10, for a variety of intervals. Positive valuessignify larger intervals when the subjects adjusted theupper note of the intervals. Negative values signifylarger intervals when subjects adjusted the lower noteof the intervals. Because each interval had differentreference and target pulse rates but a common geometricmean pulse rate, results were plotted as a function ofthe geometric mean of the target and reference pulserates.114 162 231GEOMETRIC MEAN OF REF/TARGET141may have resulted from attempts by the subjects to effect amore musical sound quality. For example, when adjusting thelower note at very low pulse rates, subjects tended to avoidrates much below 100 pps, which were described by bothsubjects as “rough”, “intermittent”, “like a motorbike” or“like a pig grunting”. This may have resulted in anunderestimation of interval size at very low pulse rates, whenthe lower note was adjusted, and an overestimation when theupper note was adjusted. At higher reference pulse rates,both subjects tended to create larger intervals when theyadjusted the lower note. The magnitude of this effect waslarger for S-b, especially with wider ascending pitchintervals. These discrepancies may have resulted from atendency to avoid higher pulse rates, which both subjectsdescribed as “unpleasant”, and “shrill”.The variability in the adjustment data of each subjectwas examined by computing the standard deviation, insemitones, from each subject’s mean pulse rate setting foreach interval, at each reference pulse rate. Recall that thestandard deviations obtained with the adjustment of the uppernote of the intervals are plotted in Figures 20 to 23 as theupper vertical error bars, and those obtained with theadjustment of the lower note as the lower error bars. Thesestandard deviations were typically less than 2 semitones. ForS-7, the standard deviations for the 5th showed a markedincrease at the highest pulse rates tested. The standarddeviations obtained with the adjustment of the upper notes of142the intervals did not vary systematically from those obtainedwith the adjustment of the lower notes of the intervals. Incontrast, variability tended to decrease at the higher pulserates for the ascending interval adjustments of S-lO. Thefair replicability of adjustments for some intervals (e.g.minor 3rd for both subjects; 4th for S-7), even at high pulserates, was unexpected, in view of frequent comments by bothsubjects regarding the difficulty of the adjustment task athigh pulse rates.Variability also tended to increase with interval size(Tables 4-6). For S-7, the standard deviation, at the 4lowest pulse rates, averaged 1.87 semitones for the 5th, 1.48for the 4th, and .95 for the minor 3rd. When measured as apercentage of target interval size (7, 5, and 3 semitones forthe 5th, 4th, and minor 3rd, respectively) these standarddeviations represented 27% of a 5th, 30% of a 4th, and 32% ofthe minor 3rd. Thus the amount of variability appeared torepresent approximately a constant proportion of the intervalsize. For S—b, there appeared to be no relationship betweeninterval size and magnitude of the standard deviation. Itshould also be noted that smaller standard deviations were notnecessarily associated with a greater proximity to the targetpulse rates. Some of the most consistent adjustments of S—b,in fact, occurred at higher pulse rates, where mean intervalsize showed the greatest deviation from the target pulse rateand the results showed the poorest preservation of intervalrank size. Thus, our data suggest that measures of143variability may, at least in some circumstances, not providean accurate index of pitch salience.The proximity of individual adjustments to the targetpulse rate was also examined by calculating, for the ascendingintervals, the percentage of trials which yielded responseswithin a reasonable proximity (2 semitones or less) of thetargets. This analysis (Figure 25) further confirmed that theadjustments of S-7 in all but the highest pulse rate regionsfell within a narrow range of the target, and that accuracywas greater for the minor 3rd than for the 4th and 5th. Onlyfor the 5th was there a trend towards decreased accuracy athigher pulse rates. For S—b, the compression of largerascending intervals at higher pulse rates was reflected in amarked decrease in the percentage of adjustments within closeproximity of the target. For the minor 3rd, performance wascomparable for both subjects.DISCUSSIONMusical intervals are ratio—specified. Thus, formusicians, and under some circumstances for nonmusicians(Attneave and Olson 1971), within a musical context, in therange over which musical pitches are perceived, acousticaltones in equal frequency ratios produce approximately equalmusical pitch intervals. Subject 7 transposed the pitchintervals to different pulse rates with a reasonable degree ofaccuracy, and did so in a manner similar to that observed when‘.1a-1008060402010080604020ASCENDING 5THASCENDING 4THMINOR 3RD74/88 105/125 148/176 212/iSa 298/354 4251505PULSE RATES OF ANCHOR AND TARGETI SEMITONE 2 SEMITONESFigure 25. Interval transposition experiment: Subjects 7and 10. Percentage of pulse rate adjustments 1—2 semitonesor less from the target, for three intervals, over a rangeof reference and target pulse rates.144145normal—hearing, nonmusical subjects are required to transposea well—known melody into different frequency ranges (Attneaveand Olson 1971). Thus, for this subject, equal musical pitchintervals were characterized by equal ratios of pulse rates.This appears to be clear evidence of pitch percepts whichpermit ratio—governed musical interval recognition. Pitchpercepts can be ranked and scaled ordinally on a high—to—lowcontinuum, but this does not necessarily imply that they varywith the stimulus in such a manner that stimulus ratios can beadjusted with accuracy to the target intervals. In otherwords, while musical interval recognition or productionimplies the ability of subjects to make ordinal pitchjudgements, the reverse is not necessarily true (Houtsma1984). Subject 10 also transposed on a ratio scale, but didso consistently only for descending intervals. For ascendingintervals at high pulse rates, this subject appeared to beable to make ordinal pitch judgements, in that the ordinalproperties of the intervals he adjusted were generally correct(i.e., ascending intervals were generally adjusted so that thepulse rate of the upper note was higher than that of the lowernote of the interval). However, at high pulse rates, heappeared to be unable to adjust pulse rates to the targetinterval values when the intervals were large and ascending.Perception of small pulse rate ratios as large subjectiveintervals, or compression of interval size, as observed withS—b (for ascending intervals), could be interpreted tosuggest a rapid increase in pitch with increases in pulse146rate, such as that anecdotally reported by Shannon (1983).However, the data for descending intervals do not support thisnotion, and suggest that the apparent increase in pitch may bedue to other factors such as, for example, marked changes intimbre (e.g., an increase in shrillness, sharpness, orunpleasantness), which may covary with pitch, as pulse rate isvaried. Changes in timbre or other attributes of a sound havebeen shown to complicate the pitch adjustment process,especially with wide, ascending intervals (Burns andViemeister 1981). Certainly, from the perspective of musictheory, differential processing of ascending and descendingintervals, at identical frequencies, would appear unlikely(Deutsch 1969).Compression of musical interval size has also beenobserved in normal—hearing musical subjects. Attneave andOlson (1971), for example, reported that the two musicalsubjects in their first experiment transposed simple 2—notepure tone synthetic melodies on a logarithmic scale over thefirst 5 octaves. However, when the upper frequencies exceeded5 kHz, both subjects showed a consistent compression of someof the larger intervals, to the extent that these intervalswere adjusted to less than half their original size. Smallerintervals that were wholly below 5 kHz were not significantlyaffected. Their data for ascending and descending intervalswere pooled, precluding assessment of effects of intervaldirection. The authors hypothesized that the subjects mayhave become confused when one tone of a pair was within the147musical pitch range, and the other above it. These authorsnoted that this compressive aberration recovered partially inthe octave from approximately 6 to 12 kHz, when both toneslacked musical quality.The standard deviations observed in our data for S—7 andS—b (descending intervals) appeared, on average, larger thanthe deviations reported in the second experiment of Attneaveand Olson (1971), in which normal—hearing nonmusical subjectsadjusted pure tones in the transposition of a familiar,overlearned melody. However, direct comparison of thevariability in our data with theirs is difficult, as theseinvestigators reported only the mean deviations from themedian frequency adjustment, and provided no informationregarding the statistical distribution of their data. Themean deviations reported by Attneave and Olson (1971) averagedapproximately 0.73 semitones for the major 6th, and 0.45semitones for the major 3rd, in the frequency range from 92.5Hz to 4186 Hz. In comparison, the standard deviations for S7, up to approximately 300 pps, averaged 1.87 semitones forthe 5th, 1.48 semitones for the 4th, and 0.95 semitones forthe minor 3rd. The variability in the data for S-b, withdescending intervals, was in the order of 1 to 1.2 semitones.The differences between our results and those of acousticalexperiments should not be surprising, and could be due to anumber of factors, including musical aptitude and musicalexperience of our subjects prior to deafness, or the accuracyof their auditory memory for interval sizes following148prolonged auditory deprivation. Differences between neuronalresponse patterns to electrical and acoustical stimulation mayalso be expected to be associated with qualitative differencesbetween pitch percepts in normal—hearing subjects andelectrically stimulated subjects.The standard deviations obtained for our electricallystimulated subjects were also greater than those obtained byBurns and Viemeister (1976, 1981) with musically trainedsubjects and sinusoidally amplitude—modulated noise, even atlow modulation frequencies. These authors reported that,while their subjects performed well on a musical intervalidentification test when the intervals were separated by aminor 3rd (i.e., when the response set consisted of a minor3rd, a tritone, a major 6th and an octave), they scored muchlower when the intervals in the response set were separated byonly a semitone (i.e., semitone, major 2nd, minor 3rd, and amajor 3rd). Burns and Viemeister (1976) concluded thatmusical interval identification with amplitude-modulatednoise, which provides predominantly or exclusively temporalinformation, was accurate to within approximately onesemitone. Comparable deviations, in the order of onesemitone, were subsequently obtained by the same authors,using the method of adjustment, with highly trained musicians(Burns and Viemeister 1981). The most obvious potentialexplanation for the larger standard deviations in our data isthe difference in musical experience of the subjects. Subject10 may have had a musical aptitude, but had received no149musical training, and S—7 reported no particular interest inmusic prior to deafness. Additional explanations for thesedifferences could be sought in either the quality of thesounds heard by the electrically stimulated subjects, or inthe accuracy of their memory for interval size, followingprolonged auditory deprivation.The trend towards increased variability of the data athigher pulse rates has some parallels in the results of bothacoustical and physiological experiments. Attneave and Olson(1971) showed a sharp increase in the mean deviation from thecorrect frequency adjustments when acoustical frequenciesexceeded 5000 Hz. Burns and Viemeister (1976) reported thatmusically trained subjects listening to amplitude-modulatednoise at low frequencies (with a first—note modulationfrequency of 84 Hz), were able to identify musical intervalsseparated by minor 3rds, with an accuracy of 84%. While theupper limits of these temporally mediated pitches for somesubjects were reported to be as high as 800 to 1000 Hz, markedperformance decrements were generally observed for modulationrates above 300 Hz.Physiological data suggest that, for both acoustical andelectrical stimulation, the upper limit of the range overwhich musical interval recognition is accurate correspondsapproximately to the upper limit at which the stimulus periodsare accurately preserved in the temporal patterning of nervefiber responses. Thus, with acoustical stimuli, 5000 Hz isapproximately the upper limit for the phase-locking of150auditory neurons to the stimulus frequency (Rose et al.1967). Data from electrically stimulated auditory neuronshave shown cycle—for—cycle firing patterns, up to rates of600—900 spikes per second (Kiang and Moxon 1972; Hartmann,Topp, and Klinke 1984). At higher pulse rates, the temporalcharacteristics of the neuronal responses to electricalstimuli are known to become increasingly dominated byinteractions between the refractory period of the neurons andthe amount of charge delivered during the excitatory portionof the stimulus cycle (Parkins 1989). The increasingrepresentation, at higher pulse rates, of interspike intervalsrelated to the refractory status of the neurons rather than tothe stimulus period, or multiples thereof, may account for theweakening of pitch at higher pulse rates.The upper limits of pitch perception with electricalstimulation have been shown to be highly variable betweensubjects. While some studies have reported rate—based pitchdiscrimination to be restricted to frequencies below 400-500Hz (Sachs 1983; Simmons 1983), others have shown higherlimits. Tong and Clark (1985), for example, reported that 2out of 6 implanted listeners accurately labelled pulse ratesup to at least 600-1000 pps, while two other listenerssaturated at 200-400 pps. Furthermore, the upper limit ofpitch labelling ability was not found to correlate with themusical history of the subjects. Others have reportedfrequency discrimination in implanted listeners up to 1000 Hz(Spillmann, Dillier, and Guentensperger 1982; Hochinair-Desoyer151et al. 1983), or even 2000 Hz (Bilger 1977). While it is notpossible to assess the adequacy of the loudness equalizationprocedures used in these experiments, our results areconsistent with data that show ordinal ranking of pitch to bepossible, at least in some implanted subjects, up to at least600 to 1000 pps. The increased amount of variability in ourtransposition data at pulse rates above approximately 300-400pps suggest that the ratio relations of the stimulus pulserates became increasingly difficult to extract. The resultsalso suggest that judgements of stimulus ratios were possiblein only a portion of the range of pulse rates over whichordinal pitch judgements were possible.152GENERAL DISCUSSIONWhile physiological evidence has shown that thedischarges of nerve fibers in the peripheral and lower centralportions of the auditory system are highly synchronized toacoustical frequencies below 4-5 kHz (Rose et al. 1967; Kiangand Moxon 1972; Javel 1980; Greenberg and Rhode 1990), thereremains uncertainty whether the human brain actually utilizesthis temporal information in the analysis of pitch (Javel andMott 1988). Thus, at least for some types of stimuli whichyield musical pitch sensations, recordings from singleauditory nerve fibers have shown that the temporal dischargepatterns contain sufficient information to account for manypitch—related phenomena. Evans (1978) reported that, in thecat, interspike interval histograms obtained in response toinharmonic stimuli, which in humans are known to evoke pitchshifts, were distinguishable from those obtained with theirharmonic counterparts. In other words, the interspikeintervals measured were consistent with the period of thepitches reported by humans. Javel (1980, 1988) has presenteddata showing, in the interspike interval histograms of singlecochlear nerve fibers of the chinchilla, a representation ofintervals corresponding not only to the period of theperceived pitch of both harmonic and inharmonic stimuli, butalso, in the predicted proportions, to other temporal153intervals present in the fine structure of the waveform.These data suggest that the discharge patterns of auditoryneurons may contain sufficient information for the analysis ofpitch on a temporal basis.Greenberg (1980) and Greenberg, Marsh, Brown, and Smith(1987) have presented electrophysiological evidence, in thescalp—recorded frequency—following—responses (FFR) of humansubjects, for a temporal processing of pitch. The spectralanalyses of the FFR recordings obtained in response to a rangeof harmonic stimuli, with and without fundamental frequencycomponent, as well as of inharmonic stimuli were found to besimilar to FFR5 generated in response to pure tones equal infrequency to the perceived pitch. These physiological datasuggest that the pitch of complex tones may be encoded in thetemporal discharge patterns of neurons not only at the levelof the cochlear nerve, but also at higher levels within theauditory brainstem.Psychophysical data from acoustical experiments supportthe conclusion that, under certain experimental conditions,temporal information alone can convey pitch, even afterappropriate precautions have been taken to eliminate thepossible role of distortion products and switching transients.Stimuli consisting solely of high, unresolvable harmonics havebeen found to evoke residue pitches corresponding to theabsent fundamental frequency (Schouten 1940; de Boer 1956;Ritsma 1962; Moore 1973, 1977; Hoekstra and Ritsma 1977;Houtsma and Smurzyns]ci 1990; Pierce 1991). Stimuli consisting154of amplitude—modulated white noise have also been shown toconvey pitch information (Miller and Taylor 1948; Small 1955;Harris 1963; Pollack 1969; Patterson and Johnson-Davies 1977)even under conditions of bandpass filtering and band—rejectmasking. Further evidence for temporal mechanisms in pitchperception can be found in time separation and phase—shiftphenomena. Small and McClellan (1963), for example,demonstrated that when a pulse train is added to a delayedreplica of itself, a pitch is heard which corresponds to thereciprocal of the time delay (t) between the two pulse trains,even though there is no energy in the frequency spectrum atl/t. Cramer and Huggins (1958) showed that dichoticpresentation of white noise of which a small frequency regionpresented to one ear had been phase—shifted, resulted in theperception of a faint pitch corresponding to the frequency ofthe phase transition.Few studies have addressed musical salience of temporallybased pitches. Burns and Viemeister (1976) demonstrated thatthe pitches evoked by sinusoidally amplitude-modulated noisewere sufficiently salient to enable musically naive subjectsto recognize simple melodies. They also showed that musicallytrained subjects were able to identify musical intervals, whenthe constituent “notes” corresponded to low modulationfrequencies. Recognition with amplitude—modulated noise,however, was found to be consistently inferior to thatachieved with pure tones. The weakness of the pitchesassociated with such stimuli has been hypothesized to result155from random fluctuations in the input waveform of the noise.These fluctuations, in turn, could be expected to result inirregularities in interspike intervals, greater than thoseobtained with tonal stimuli (Moore and Glasberg 1986).Similar conclusions, confirming that pitches which presumablyarise solely from the temporal aspects of the waveform aresufficiently salient to convey musical interval informationhave been reported by others (Houtsma and Smurzynski 1990;Pierce 1991). It appears generally agreed, however, thatthese pitches are weaker than those resulting from tonalstimulation (Burns and Viemeister 1976, 1981; Houtsma 1984;Houtsma and Smurzynski 1990).However, with acoustical experiments, it is difficult orimpossible to eliminate, unequivocally, the contributions ofspectral or place mechanisms in the analysis of pitch. It hasbeen argued, for example, that with amplitude—modulated noise,short—term spectral fluctuations related to the modulationrate may contribute to pitch perception (Pierce, Lipes, andCheetham 1977). While these alterations in the statisticalproperties of the noise are thought to be small and of limitedsignificance in demonstrations of temporally based pitch,particularly with sinusoidal modulation (Moore and Glasberg1986), these possibilities introduce elements of uncertaintyregarding direct utilization of temporal information. Furtherdoubts arise from the potential contributions of distortionproducts generated within the ear. Thus, spectral peaks not156present in the stimulus may be generated by cochlearnonlinearities (Horst, Javel, and Farley 1990).In electrically stimulated cochleas, there is believed tobe no mechanical, place—related frequency analysis of thestimulus waveform. These ears, therefore, represent uniqueopportunities for the study of temporal mechanisms inaudition. Electrical pulse trains in deaf cochleas are knownto precipitate highly synchronized temporal spike dischargepatterns in the target neural population (Hartmann, Topp, andKlinke 1984; Javel et al. 1987; van den Honert andStypulkowski 1987b). While many psychophysical studies havedocumented basic relationships between electrical frequency orpulse rate and pitch (Eddington et al. 1978a, l978b; Simmonset al. 1979, 1981; Tong et al. 1982; Hochmair-Desoyer et al.1983; Shannon 1983; Pfingst 1985; Tong and Clark 1983, 1985;Townshend et al. 1987; Shallop et al. 1990), thesetemporally based pitch percepts have not generally beenassessed within a meaningful (i.e., musical) context. For thepurpose of our experiments, pitch was narrowly defined as thatattribute of auditory sensation which conveys musical intervalinformation. Such a definition of pitch is considered to beboth conservative and realistic (Burns and Viemeister 1976,1981; Houtsma 1984).Our results support the findings of acoustical studieswhich have demonstrated temporally based pitches, and suggestthat pulsatile electrical stimulation of totally deaf earsmay, at least in some subjects, result in pitches sufficiently157strong to support musical interval perception. It is expectedthat, in view of the well-documented relationships between thetemporal characteristics of the electrical stimulus and thetemporal characteristics of the rieuronal responses in suchears (Hartmann, Topp, and Klinke 1984; Javel et al. 1987; vanden Honert and Stypulkowski 1987b), that these pitch effectsmust be based on an analysis of interspike intervalinformation. Additional evidence for the ability of implantedsubjects to utilize interspike intervals in the analysis ofpitch has been obtained from experiments utilizing stochasticpulse trains. Dobie and Dillier (1985) showed that theability of deaf subjects to discriminate pitches remainedintact when the probability of pulse delivery at eachprescribed time in the pulse train remained greater than 0.5.These data suggest that the auditory nervous system is capableof making perceptual decisions about pitch on the basis ofinterspike intervals. While the possibility of excitation ofneurons by means of electrophonic vibrations, such as might behypothesized to result from electrical stimulation of intactouter hair cells, or by means of direct depolarization ofresidual inner hair cells cannot be entirely ruled out, thesemechanisms appear to be highly unlikely in ears with noresidual acoustic sensitivity (van den Honert and Stypulkowski1984, 1987b)These conclusions are not intended to imply that temporalinformation is the sole mediator of musical pitch or musicalinterval recognition, or even that the rate—based subjective158musical intervals reported in this paper are immediatelyobvious to all implanted listeners. During some of ourinformal pilot experiments, some listeners commented thatthere was little or no noticeable change in pitch as pulserate was varied. Others commented that a slow apical—to—basalelectrode sweep (at one electrode per second, using a constantpulse rate of 125 pps) produced a stronger, more noticeablerise in pitch than a slow pulse rate sweep on a singleelectrode (one pulse rate per second, using pulse rates whichincreased from 54 pps to 1096 pps, by a factor of 1.22).Furthermore, informal pilot studies demonstrated that subjectsappeared to be able to make intonation quality judgements alsoon the basis of variation in the place of stimulation (i.e.,by switching from one electrode to another, at a constantpulse rate). All subjects, including those who hadconsiderable experience listening to pulse rate melodies onbasal electrodes, agreed that melodies sounded more “musical”and “pleasant” on apical electrodes. This suggests that acongruence between place and rate information may be importantin achieving a musical sound quality (Evans 1978). In Nucleusimplant recipients with complete electrode insertions, thelocations of the most apical and basal electrodes have beenestimated to correspond, respectively, to the 600 and 8000 Hzregions of the normal cochlea (Greenwood 1961, 1990; Blamey,Dowell, Brown, Clark, and Seligman 1987).Experiment I showed that many familiar tunes were readilyrecognized and identified by most subjects, especially when159these tunes were played on a single intracochlear electrode,at low pulse rates, and with intact rhythmical patterns. Thiscontrasts with the poor tune recognition results reported forNucleus subjects by Gfeller and Lansing (1991). Thesesubjects, however, were tested in the soundfield, andstimulation was delivered by the body—worn processor, whichperformed a feature extraction upon the incoming acousticsignals (Koch et al. 1990). This processing strategyextracts from the acoustic signal the spectral peaks with thehighest amplitude, and quasi-simultaneously activateselectrodes selected on the basis of a place code, at a pulserate equal to the fundamental frequency. It is conceivablethat such a processing strategy could result in a confoundingof pitch information based on temporal features and pitchinformation based on place of stimulation. Our resultssuggest that the application of a systematic series of pulserates to a single intracochlear electrode may yield morerobust rate pitches.Experiment II demonstrated that systematic variation ofthe electrical pulse rate resulted in pitches sufficientlystrong to support tune recognition, even in the absence ofrhythmical information. While variation of pulse ratesappeared to permit pure melody (i.e., musical intervalsequence) recognition, the contribution of melodic contourremained undetermined. Performance was best at low pulserates for all subjects. Although subjects scoredsignificantly above chance at high pulse rates, they reported160that the melodies “sounded like music” only at the lowestpulse rates. The results of an ancillary informal experiment,furthermore, suggested that at least one of the subjectsprocessed the melodies in a musical way only at low pulserates. This subject appeared to remain oblivious to thepresence of an unfamiliar tune in the stimulus set until thistune was played at a low pulse rate. This suggests thepossibility that, at least at high pulse rates, subjects mayhave been attending to the melodic contour, or applyingcomplex, nonmusical cognitive strategies, by listening forsystematic changes in nonpitch attributes.Large individual differences between the subjects in theability to utilize temporal information, such as thosereported here, are not unique to electrically stimulatedsubjects. Large intersubject differences in performance onacoustical pitch perception tasks are common, even withnormal-hearing subjects (Burns and Viemeister 1976, 1981).Similarly, Tong and Clark (1985) showed large differences inthe ability of implanted patients to perform pulse rateidentification tasks. These authors concluded that the uppercutoff rates, beyond which no change in pitch could beperceived, did not bear any relation to the degree of previousmusical training. For S-il, the task appeared to be extremelydifficult, even at low pulse rates. This subject appeared torequire more time per trial than the other two subjects, inall the experiments, and evidenced a significantly greaterperformance decrement with increasing pulse rate. It is161possible that this subject had poor nerve survival relative tothe other two subjects, as a result of deafness secondary tomeningitis (Nadol and Hsu 1991). It is of interest tospeculate whether there are any relationships between theability to utilize temporal information, as in tasks such asthose reported here, and speech perception results.Longitudinal comparison of speech perception scores showedthat S—7 and S-b achieved high scores on open-set speechmaterials within 6 weeks of being fitted with theirprocessors, whereas S—li required an extended learning periodto reach an asymptotic, although comparable score. A numberof reports in the literature have supported the possibility ofa correlation between temporal measures such as gap detectionthresholds and the performance on speech perception tests(Tyler, Moore, and Kuk 1989; Cazals, Pelizzone, Kasper, andMontandon 1991; Blarney et al. 1992).The results of Experiment III indicated that musicalinterval information was available in the memory of theimplanted listeners, and that the ratio relationshipsappropriate for acoustical musical tones were at least grosslycorrect for electrical repetition rates. This means that theneural responses to different pulse rates permit both ordinalranking of pitches as well as ratio information. At low pulserates, the intonation quality judgements demonstrated that allsubjects preferred interval sizes which consisted of pulserates in the same ratios which characterize musical intervalsheard acoustically. At higher pulse rates, subjects tended,162to varying degrees, to compress the size of large intervals.It is difficult to determine, on the basis of these results,whether this apparent compression of interval size was amanifestation of a rapid increase in pitch, such as thatanecdotally reported by Shannon (1983). It is alsoconceivable that this phenomenon was attributable to markedchanges in timbre, or to other attributes of the sound, whichmight serve to “exaggerate” the subjective size of thephysical interval. Compression of interval size at highacoustic frequencies has been observed in normal—hearingmusical subjects performing a musical interval adjustmenttask, by Attneave and Olson (1971), when one of the two puretone frequencies exceeded 5000 Hz. This compression was muchless marked when both stimulus frequencies were above 5000 Hz.The results of Experiment III also showed that,regardless of the cochlear location of the stimulatingelectrode and (presumably) the responding neural population,equal ratios of pulse rates appeared to yield approximatelyequal musical pitch intervals, at least for the interval andrepetition rates examined. Psychophysical and physiologicaldata support the notion that electrical stimulation viaclosely spaced bipolar electrode pairs in the scala tympanican result in the excitation of relatively restricted groupsof neurons (Shannon 1983; van den Honert and Stypulkowskil987a). The spread of excitation, in patients with good nervesurvival, has been estimated at 2—4 mm, at a comfortablelistening level (Black, Clark, and Patrick 1981; Merzenich1631983; Tong and Clark 1986). Data regarding spatiallocalization of current within implanted cochleas have beenextensively corroborated by anecdotal clinical observationsregarding gradations of pitchlike effects which are related tothe place of electrical stimulation (Muller 1983). Theseeffects are generally in accordance with the tonotopicorganization of the cochlea (Tong and Clark 1985).Even though equal pulse rate ratios applied to differentelectrodes appeared to yield approximately equal subjectivemusical intervals, subjects reported the pitch changes to bemore difficult to hear and to sound less musical on basal thanon apical electrodes. In spite of the anecdotal observationsof the subjects regarding the greater difficulty of theintonation quality labeling task with stimulation on basalelectrodes, there were no consistent differences in standarddeviations, between these electrode locations. These findingsare in accord with results of acoustical experiments, whichsuggest that stimulation of a given region of the basilarmembrane may give rise to different pitches, depending on thetemporal patterning nerve impulses (Schouten 1940; Davis,Silverman, and McAuliffe 1951; Ritsma 1962; Ritsma and Engel1964). These phenomena have been extensively documented inthe psychoacoustical literature.The difficulties encountered in assigning pitches tostimuli which consist of conflicting temporal and placeinformation, such as those reported here with basalelectrodes, have also been reported in studies with normal—164hearing listeners. For example, Davis et al. (1951)demonstrated that when subjects were required to match thepitch of a pure tone to that of 2 kHz rectangularly gatedpulses, presented at rates of 90 to 150 times per second andbandpassed through a 2 kHz filter, the matches of somesubjects corresponded to the center frequency of the filter(i.e., 2 kHz), while the matches of others corresponded to therepetition rate. Similarly, Burns and Viemeister (1981)demonstrated relatively poorer performance with a sinusoidallyamplitude-modulated 10 kHz acoustic sine wave than with avariety of broadband and bandpass amplitude—modulated noises,on tasks of melody identification, dictation, and musicalinterval adjustment. These differences were large for somelisteners. It is possible that the strength of temporallybased pitches declines in proportion to the mismatch betweenplace and timing information. The difficulty of discerningpitch changes on basal electrodes, as anecdotally observed byour subjects, may explain the findings of Shallop et al.(1990), who reported proportionately smaller pitch changeswith variation of electrical pulse rates on basal than onapical electrodes. Marked qualitative differences in thesound evoked by electrodes in different cochlear regions mayalso explain the findings of Eddington et al. (1978a), whoreported that melodies which were identifiable when played onsome intracochlear electrodes were not identifiable on others.Our limited data do not support the hypothesis of Eddington etal. (l978a) that different electrical pulse rate or frequency165ratios might be required to produce recognizable tunes indifferent regions of the cochlea.The interval adjustment task of Experiment IV showed thatat least two of the three subjects were able to adjust lowelectrical pulse rates in the reconstruction of common musicalintervals, when these intervals were abstracted from familiartunes. The ratio properties of these intervals closelyapproximated those of the corresponding acoustical musicalintervals.The fact that the precision with which our subjectsperformed the musical interval adjustment task did notcorrelate with the musical history of the subjects is notentirely surprising. Clinical studies have amply documentedlarge individual differences in the speech perceptionabilities of patients with cochlear implants, unrelated toindividual linguistic skills. Thus, individual differences inmusical pitch perception, unrelated to musical ability, couldalso be anticipated. Furthermore, musical ability or aptitudeis difficult to define and measure, and may fail to correlatewith musical achievement, interest, or experience. Forexample, tonal memory has been reported to be possessed bydifferent subjects in widely varying degrees, which are notnecessarily related to length of musical experience (Shuter—Dyson 1982). Musical achievement and interest are believed torelate in a complex fashion to genetic differences betweenindividuals in their capacities for building up appropriateneural and mental schemata, and the stimulation offered within166the environment. The latter, in turn, is strongly influencedby cultural conditions and individual leisure pursuits(Shuter—Dyson 1982). Conversely, it must be remembered thatsome individuals who do possess a level of musical achievement(Table 1) have neither musical interest nor aptitude.The results of the interval transposition task inExperiment V were particularly interesting. It must berecognized that some normal—hearing, musically untrainedsubjects have been shown to be unable to perform this type oftask with musical tones (Attneave and Olson 1971). Subject 7transposed all three intervals on an essentially logarithmicscale, where equal musical intervals were represented by equalratios of pulse rates. These results were comparable to thoseof the second experiment of Attneave and Olson (1971), whosenonmusician subjects were required to reconstruct a familiarmelody using pure tones. Our results suggest that lowelectrical pulse rate pitches, when judged in a musicalcontext, possess ratio properties similar to those ofacoustical musical tones. The apparently larger variabilityin our data, compared to that of the subjects of Attneave andOlson (1971), may originate from a variety of sources,including differences in the musical ability of the subjects,the accuracy of their memory for melodic pitch intervals, thedifficulty of the transposition task, the effects of prolongedauditory deprivation, and the quality of the sounds perceivedwith pulsatile electrical stimulation.167In the transposition behaviour of Subject 10, thedirectional properties of the intervals appeared to becritical. Thus, ascending intervals were adjusted differentlyfrom descending intervals of the same magnitude, even whenreference and target pulse rates of the ascending anddescending intervals were identical. Ascending intervals wereadjusted on an essentially logarithmic scale for low pulserates. At higher pulse rates, larger ascending intervalsbecame severely compressed, and interval sizesundifferentiated, especially when the subject adjusted theupper note of the intervals. In marked contrast, descendingintervals were adjusted on an essentially logarithmic scale,with fairly constant pulse rate ratios over the range of pulserates tested (similar to S—7). These marked discrepanciesbetween the results for ascending and descending intervals aredifficult to explain. It is possible that marked changes inattributes such as shrillness or unpleasantness, which mayhave accompanied changes in pulse rate, may have obscured thepitch percepts and complicated selectively the adjustment ofascending pitch intervals, and that these qualitative changeswere less obtrusive with descending pitch intervals, or whenthe subject adjusted the lower note of ascending pitchintervals.While our experiments did not specifically address thelimits of the existence region of electrical pulse rate pitch,significant performance decrements were observed at pulserates above 200-800 pps, in Experiments II, III and V.168Similar upper limits for the processing of temporalinformation have been demonstrated in acoustical experiments(Burns and Viemeister 1976; Javel and Mott 1988) and in otherpsychophysical experiments with electrically stimulatedsubjects (Shannon 1983; Tong and Clark 1985; Shannon 1992).On a physiological level, the ability of individual neurons torespond with temporally patterned firing at high electricpulse rates is limited by the relative refractory period,which has been reported to begin at 300 sec following spikeonset, and to extend to at least 5 msec (Moxon 1971; Hartmann,Topp, and Klinke 1984; Stypulkowski and van den Honert 1984).Single fiber recordings have demonstrated strongly unimodalinterval histograms at electrical frequencies up to 200 Hz,indicating a firing on every cycle of the stimulus, especiallyat levels well above threshold (van den Honert andStypulkowski l987b). In these studies, neural responses atfrequencies above 200—500 Hz were characterized by anincreasing representation of interspike intervals at multiplesof the stimulus period.Thus, at the relatively high pulse rates used in some ofour experiments, it is unlikely that individual nerve fiberswere responding on a cycle—for—cycle basis to the electricstimulus. It is more likely that at these rates, there was anincreasing representation of interspike intervals at multiplesof the stimulus period. Psychophysical studies involvingstochastic electrical pulse trains have shown that frequencydiscrimination can remain essentially unimpaired, even when169every period of the test frequency is not represented in thestimulus (Dobie and Dillier 1985). This is consistent withour data at high pulse rates, suggesting that the centralauditory nervous system remains capable of extractinginterpulse intervals, even when it is highly unlikely thatevery stimulus cycle results in a neural response, in a givensingle fiber. It is possible that the increasingrepresentation of a range of interspike intervals at multiplesof the stimulus period, such as might be expected at highpulse rates, results in a weakening of the pitch percept.Pitóh extraction at high pulse rates could also be based on ameasurement of multiples of the interpulse intervals, or on avolleying mechanism (Wever 1949) between electricallystimulated neurons. This would allow temporal processes toextend to pulse rates higher than the limit imposed by therefractoriness of individual neurons. Little is known aboutindividual differences in the spatio—temporal responsepatterns in a population of electrically stimulated neurons.Parkins (1989) has pointed out that the gradation of stimulusintensity within electrical fields may result in differencesin the temporal patterning of responses of neurons situatednear the center of the field than for those at the periphery,and that the effects of a range of interspike intervals onpitch perception, particularly with electrical stimulation, isnot known.In conclusion, it appears that variation in pulse rate onsingle intracochlear electrodes can, at least in some cochlear170implant recipients, result in pitch changes which aresufficiently salient to support musical interval perceptionand tune recognition. In the absence of spectral analysis insuch ears, these results are almost certainly due to thetemporal firing patterns of residual auditory neurons.171REFERENCESAmerican National Standards Institute: Acoustical Terminology,(ANSI S3.20-1973) New York.Attneave, F. and R.K. Olson. 1971. Pitch as a medium: a newapproach to psychophysical scaling. Am. J. Psychol. 84:147—166.Bilger, R.C. 1977. Psychoacoustic evaluation of presentprostheses. Ann. Otol. Rhinol. Laryngol. 86 (Suppl.38),92—140.Black, R.C., G.M. Clark, and J.F. Patrick. 1981. Currentdistribution measurements within the human cochlea. IEEETrans. Biomed. Eng. 28: 721-725.Blarney, P.J., R.C. Dowell, A.M. Brown, G.M. Clark, and P.M.Seligman. 1987. Vowel and consonant recognition ofcochlear implant patients using formant—estimating speechprocessors. J. Acoust. Soc. Am. 82: 48—57.Blarney, P.J., R.C. Dowell, G.M. Clark, and P.M. Seligman.1987. Acoustic parameters measured by a formant—estimating speech processor for a multiple—channelcochlear implant. J. Acoust. Soc. Am. 82: 38—47.Blarney, P.J., B.C. Pyman, B.M. Clark, R.C. Dowell, M. Gordon,A.M. Brown, and R.D. Hollow. 1992. Factors predictingpostoperative sentence scores in postlinguistically deafadult cochlear implant patients. Ann. Otol. Rhinol.Laryngol. 101: 342—348.Burns, E.M., and N.F. Viemeister. 1976. Nonspectral pitch.J. Acoust. Soc. Am. 60: 863-869.Burns, E.M., and N.F. Viemeister. 1981. Played-again SAN:Further observations on the pitch of amplitude-modulatednoise. J. Acoust. Soc. Am. 70: 1655—1660.Burns, E.M., and W.D. Ward. 1978. Categorical perception-phenomenon or epiphenomenon: Evidence from experiments inthe perception of melodic musical intervals. J. Acoust.Soc. Am. 63: 456—467.172Cazals, Y., M. Pelizzone, A. Rasper, and P. Montandon. 1991.Indication of a relation between speech perception andtemporal resolution for cochlear implantees. Ann. Otol.Rhinol. Laryngol. 100: 893—895.Chouard, C.H. 1978. Multiple intracochlear electrodes forrehabilitation in total deafness. OtolaryngologicClinics of North America 11: 217-233.Clark, G.M., R. Black, D.J. Dewhurst, IC. Forster, J.F.Patrick, and Y.C. Tong. 1977. A multiple-electrodehearing prosthesis for cochlear implantation in deafpatients. Med. Progr. Technol. 5: 127-140.Clopton, B.M., F.A. Spelman, I. Glass, B.E. Pfingst, J.M.Miller, P.D. Lawrence, and D.P. Dean. 1983. Neuralencoding of electrical signals. In Cochlear Prostheses:An International Symposium, ed. C.W. Parkins and J.W.Anderson. Ann. N.Y. Acad. Sci. 405: 146-158.Cramer, E.M., and W.H. Huggins. 1958. Creation of pitchthrough binaural interaction. J. Acoust. Soc. Am. 30:413—417.Cross, I., P. Howell, and R. West. 1985. Structuralrelationships in the perception of musical pitch. InMusical Structure and Cognition, ed. I. Cross, P. Howell,R. West, 121-142. London: Academic Press Inc.Davis, H., T. Morgan, J.E. Hawkins, R. Galambos, and F.W.Smith. 1950. Temporary deafness following exposure toloud tones and noise. Acta. Otolaryng. (Stockh).(Suppl.8): 1—57.Davis, H., S.R. Silverman, and D.R. McAuliffe. 1951. Someobservations on pitch and frequency. J. Acoust. Soc. Am.23: 40—42.de Boer, E. 1956. On the Residue in Hearing. AcademicThesis, University of Amsterdam.Deutsch, D. 1969. Music recognition. Psychological Review76: 300—307.Deutsch, D. 1972. Octave generalization and tunerecognition. Perception & Psychophysics 11: 411—412.173Dillier, N., T. Spillmann, and J. Guentensperger. 1983.Computerized testing of signal-encoding strategies withround-window implants. Ann. N.Y. Acad. Sci. 405: 360-369.Dobie, R.A., and N. Dillier. 1985. Some aspects of temporalcoding for single-channel electrical stimulation of thecochlea. Hearing Res. 18: 41-55.Dowling, W.J. 1978. Scale and Contour: Two components of atheory of memory for melodies. Psychological Review 85:341—354.Dowling, W.J., and J.C. Bartlett. 1981. The importance ofinterval information in long—term memory for melodies.Psychomusicology 30-49.Dowling, W.J., and D.S. Fujitani. 1971. Contour, interval,and pitch recognition in memory for melodies. J. Acoust.Soc. Am. 49: 524—531.Eddington, D.K., W.H. Dobelle, D.W. Brackmann, M.G.Mladejovsky, and J.L. Parkin. l978a. Auditoryprosthesis research with multiple channel intracochlearstimulation in man. Ann. Otol. Rhinol. Laryrigol.87(Suppl.53): 5—39.Eddington, D.K., W.H. Dobelle, D.E. Brackmann, M.G.Mladejovsky, and J. Parkin. 1978b. Place andperiodicity pitch by stimulation of multiple scalatympani electrodes in deaf volunteers. Trans. Am. Soc.Artif. Intern. Organs 24: 1-5.Eddington, D.K., and J.L. Orth. 1985, Speech recognition ina deaf subject with a portable, multichannel cochlearimplant system. In Cochlear Implants, ed. R.A. Schindlerand M.M. Merzenich, 277-282. New York: Raven Press.Elliot, J., J.R. Platt, and R.J. Racine. 1987. Adjustment ofsuccessive and simultaneous intervals by musicallyexperienced and inexperienced subjects. Perception &Psychophysics 42: 594—598.Evans, E.F. 1978. Place and time coding of frequency in theperipheral auditory system: some physiological pros andcons. Audiology 17: 369-420.Flanagan, J.L., and N. Guttman. 1960. On the pitch ofperiodic pulses. J. Acoust. Soc. Am. 32: 1308-1319.Fletcher, H. 1928. The physical criterion for determiningthe pitch of a musical tone. Phys. Rev. 23: 427-437.174Gfeller, K., and C.R. Lansing. 1991. Melodic, rhythmic, andtiinbral perception of adult cochlear implant users. J.Speech Hear. Res. 34: 916—920.Glass, I. 1983. Tuning characteristics of cochlear nucleusunits in response to electrical stimulation of thecochlea. Hearing Res. 12: 223—237.Glass, I. 1984. Phase—locked responses of cochlear nucleusunits to electrical stimulation through a cochlearimplant. Exp. Brain Res. 55: 386-390.Goldstein, J.L. 1978. Mechanisms of signal analysis andpattern perception in periodicity pitch. Audiology 17:421—445.Greenberg, S. 1980. Neural temporal coding of pitch andtimbre. UCLA Working Papers in Phonetics (Number 52),Los Angeles CA.Greenberg, S., C.D. Geisler, and L. Deng. 1986. Frequencyselectivity of single cochlear—nerve fibers based on thetemporal response pattern to two—tone signals. J. Acoust.Soc. Am. 79: 1010—1019.Greenberg, S., J.T. Marsh, W.S. Brown, and J.C. Smith. 1987.Neural temporal coding of low pitch. I: Human frequency—following responses to complex tones. Hearing Res. 25:91—114.Greenberg, S., and W.S. Rhode. 1990. Periodicity coding incochlear nerve and ventral cochlear nucleus. In AuditoryProcessing of Complex Sounds, ed. W.A. Yost and C.S.Watson, 225-236. Hillside, N.J.: Lawrence Erlbaum.Greenwood, D.D. 1961. Critical bandwidth and the frequencycoordinates of the basilar membrane. J. Acoust. Soc. Am.33: 1344—1356.Greenwood, D.D. 1990. A cochlear frequency—position functionfor several species——29 years later. J. Acoust. Soc. Am.87: 2592—2605.Harris, G.G. 1963. Periodicity perception by using gatednoise. J. Acoust. Soc. Am. 35: 1229—1233.Hartmann, R., G. Topp, and R. Klinke. 1984. Dischargepatterns of cat primary auditory fibers with electricalstimulation of the cochlea. Hearing Res. 13: 47—62.Hind, J.E., D.J. Anderson, J.F. Brugge, and J.E. Rose. 1967.Coding of information pertaining to paired low-frequencytones in single auditory nerve fibers of the squirrelmonkey. J. Neurophysiol. 30: 794-816.175Hochmair—Desoyer, I.J., E.S. Hochmair, K. Burian, and H.K.Stigibrunner. 1983. Percepts from the Vienna cochlearprosthesis. In Cochlear Prostheses: An InternationalSymposium, ed. C.W. Parkins and J.W. Anderson. Ann. N.Y.Acad. Sci. 405: 295—306.Hoekstra, A., and R.J. Ritsma. 1977. Perceptive hearing lossand frequency selectivity. In Psychophysics andPhysiology of Hearing, ed. E.F. Evans and JP. Wilson,263-271. London, New York: Academic Press.Horst, J.W., E. Javel, and G.R. Farley. 1990. coding ofspectral fine structure in the auditory nerve. II: Level-dependent nonlinear responses. J. Acoust. Soc. Am. 88:2656—2681.Houtsma, A.J.M. 1984. Pitch salience of various complexsounds. Music Perception 1: 296-307.Houtsma, A.J.M., and J. Smurzynski. 1990. Pitchidentification and discrimination for complex tones withmany harmonics. J. Acoust. Soc. Am. 87: 304-310.Javel, E. 1980. Coding of AN tones in the chinchillaauditory nerve: Implications for the pitch of complextones. J. Acoust. Soc. Am. 68: 133-146.Javel, E., J.A. McGee, J.W. Horst, and G.R. Farley. 1988.Temporal mechanisms in auditory stimulus coding. InAuditory Function: Neurobiological Bases of Hearing, ed.G.M. Edelman, W.E. Gall, and W.M. Cowan, 515-558. NewYork: John Wiley and Sons.Javel, E., and J.B. Mott. 1988. Physiological andpsychophysical correlates of temporal processes inhearing. Hearing Res. 34: 275-294.Javel, E., Y.C. Tong, R.K. Shepherd, and G.M. Clark. 1987.Responses of cat auditory nerve fibers to biphasicelectrical current pulses. Ann. Otol. Rhinol. Laryngol.96 (Suppl.128): 26—30.Jones, M.R., L. Summerell, and E. Marshburn. 1987.Recognizing melodies: A dynamic interpretation. J. Exp.Psychol. 39A: 89—121.Kiang, N.Y.S., and E.C. Moxon. 1972. Physiologicalconsiderations in artificial stimulation of the innerear. Ann. Otol. Rhinol. Laryngol. 81: 714-731.Koch, D.B., P. Seligman, C. Daly, and L. Whitford. 1990. Amultipeak feature extraction coding strategy for amultichannel cochlear implant. Hear. Instrum. 41: 28-32.176Krumhansl, C.L., and E.J. Kessler. 1982. Tracing the dynamicchanges in perceived tonal organization in a spatialrepresentation of musical keys. Psychological Review 89:334—368.Krumhansl, C.L., and R.N. Shepard. 1979. Quantification ofthe hierarchy of tonal functions within a diatoniccontext. J. Exp. Psychol. 5: 579-594.Langner, C. 1983. Evidence for neuronal periodicitydetection in the auditory system of the guinea fowl:implications for pitch analysis in the time domain. Exp.Brain Res. 52: 333—355.Langner, G., and C.E. Schreiner. 1988. Periodicity coding inthe inferior colliculus of the cat: I. Neuronalmechanisms. J. Neurophysiol. 60: 1799-1822.Licklider, JC.R. 1954. “Periodicity” pitch and “place”pitch. J. Acoust. Soc. Am. 26: 945.Loeb, G.E., M.W. White, and W.M. Jenkins. 1983. Biophysicalconsiderations in electrical stimulation of the auditorynervous system. In Cochlear Prostheses: An InternationalSymposium, ed. C.W. Parkins and J.W. Anderson. Ann. N.Y.Acad. Sci. 405: 123—136.Massaro, D.W., H.J. Kailman, and J.L. Kelly. 1980. The roleof tone height, melodic contour, and tone chroma inmelody recognition. J. Exp. Psychol.: Human Learning andMemory 6: 77-90.Merzenich, N.M. 1983. coding of sound in a cochlearprosthesis: some theoretical and practicalconsiderations. In Cochlear Prostheses: AnInternational Symposium, ed. c.w. Parkins and J.W.Anderson. Ann. N.Y. Acad. Sci. 405: 502-508.Merzenich, N.M., R.P. Michelson, C.R. Pettit, R.A. Schindler,and M. Reid. 1973. Neural encoding of sound sensationevoked by electrical stimulation of the acoustic nerve.Ann. Otol. Rhinol. Laryngol. 82: 486-504.Miller, J.R., and E.C. carterette. 1975. Perceptual spacefor musical structures. J. Acoust. Soc. Am. 58: 711-720.Miller, G.A., and W.G. Taylor. 1948. The perception ofrepeated bursts of noise. J. Acoust. Soc. Am. 20: 171-182.177Mladejovsky, M.G., D.K. Eddington, W.H. Dobelle, and D.E.Brackmann. 1975. Artificial hearing for the deaf bycochlear stimulation: pitch modulation and someparametric thresholds. Trans. Amer. Soc. Artif. Intern.Organs 21: 1-7.Monahan, C.B., R.A. Kendall, and E.C. Carterette. 1987. Theeffect of melodic and temporal contour on recognitionmemory for pitch change. Perception & Psychophysics 41:576—600.Moore, B.C.J. 1973. Some experiments relating to theperception of complex tones. Q. J. Exp. Psychol. 25:451—475.Moore, B.C.J. 1977. Effects of relative phase of thecomponents on the pitch of three—component complex tones.In Psychophysics and Physiology of Hearing, ed. E.F.Evans and J.P. Wilson, 349-362. London, New York:Academic Press.Moore, B.C.J, and B.R. Glasberg. 1986. The role of frequencyselectivity in the perception of loudness, pitch andtime. In Frequency Selectivity in Hearing, ed. B.C.J.Moore and R.D. Patterson, 251-308. London: AcademicPress Inc.Moore, B.C.J., and S.M. Rosen. 1979. Tune recognition withreduced pitch and interval information. Q. J. Exp.Psychol. 31: 229—240.Moxon, E.C. 1965. Electrical stimulation of the cat’scochlea: a study of discharge rates in single auditorynerve fibers. Masterts Thesis, MIT, Cambridge, MA.Moxon, E.C. 1971. Neural and mechanical responses toelectric stimulation of the cat’s inner ear. Doctoraldissertation, MIT, Cambridge, MA.Muller, C.G. 1983. Comparison of percepts found withcochlear implant devices. Ann. N.Y. Acad. Sci. 405: 412-420.Nadol, J.B., and W. Hsu. 1991. Histopathologic correlationof spiral ganglion cell count and new bone formation inthe cochlea following meningogenic labyrinthitis anddeafness. Ann. Otol. Rhinol. Laryngol. 100: 712-716.Palmer, A.R., and I.J. Russell. 1986. Phase-locking in thecochlear nerve of the guinea—pig and its relation to thereceptor potential of inner hair cells. Hearing Res. 24:1—15.178Parkins, C.W. 1989. Temporal response patterns of auditorynerve fibers to electrical stimulation in deafenedsquirrel monkeys. Hearing Res. 41: 137-168.Patterson, R.D., and D. Johnson—Davies. 1977. Detection of achange in the pitch of AN noise. In Psychophysics andPhysiology of Hearing, ed. E.F. Evans and J.P. Wilson,363-372. London, New York: Academic Press.Pfingst, B.E. 1984. operating ranges and intensitypsychophysics for cochlear implants. Arch. Otolaryngol.110: 140—144.Pfingst, B.E. 1985. Psychophysical data from cochlearimplants: relevance to strategies for rehabilitation.Seminars in Hearing 6: 7-21.Pfingst, B.E. 1988. comparisons of psychophysical andneurophysiological studies of cochlear implants. HearingRes. 34: 243—252.Pfingst, B.E., D.R. DeHaan, and L.A. Holloway. 1991.Stimulus features affecting psychophysical detectionthresholds for electrical stimulation of the cochlea. I:Phase duration and stimulus duration. J. Acoust. Soc.Am. 90, Pt.1.: 1857—1866.Pfingst, B.E., J.A. Donaldson, J.N. Miller, and F.A. Spelman.1979. Psychophysical evaluation of cochlear prosthesesin a monkey model. Ann. Otol. 88: 613-625.Pfingst, B.E., and D. Sutton. 1983. Relation of cochlearimplant function to histopathology in monkeys. InCochlear Prostheses: An International Symposium, ed. C.W.Parkins and J.W. Anderson. Ann. N.Y. Acad. Sci. 405:224—239.Pfingst, B.E., D. Sutton, J.M. Miller, and B.A. Bohne. 1981.Cochlear implants: Relation of psychophysical measures tohistological findings. Acta. Otolaryngol. 92:1.Pierce, J.R. 1991. Periodicity and pitch perception. J.Acoust. Soc. Am. 90: 1889—1893.Pierce, J.R., R. Lipes, and c. cheetham. 1977. Uncertaintyconcerning the direct use of time information in hearing:place cues in white spectra stimuli. J. Acoust. Soc. Am.61: 1609—1621.Pijl, S. and F. Noel. 1992. The Nucleus multichannelcochlear implant: comparison of scala tympani vs. scalavestibuli electrode placement in a single patient.Otolaryngol. Head Neck Surg. 107: 472-474.179Plomp, R. 1976. Aspects of Tone Sensation. London and NewYork: Academic Press.Pollack, I. 1969. Periodicity pitch for interrupted whitenoise - fact or artifact. J. Acoust. Soc. Am. 45: 237-238(L).Risset, J-C. 1978. Musical Acoustics. In Handbook ofPerception Vol.4, ed. Carterette, 521-564. New York:Academic Press Inc.Ritsma, R.J. 1962. Existence region of the tonal residue: I.J. Acoust. Soc. Am. 34: 1224—1229.Ritsma, R.J. 1967. Frequencies dominant in the perception ofpitch of complex sounds. J. Acoust. Soc. Am. 42:191-199.Ritsma, R.J., and F.L. Engel. 1964. Pitch of frequency-modulated signals. J. Acoust. Soc. Am. 36: 1637-1644.Rose, J.E., J.L. Brugge, D.J. Anderson, and J. Hind. 1967.Phase—locked response to low—frequency tones in singleauditory nerve fibers of the squirrel monkey. J.Neurophysiol. 30: 769—793.Sachs, M.B. 1983. Round table discussion. In CochlearProstheses: An International Symposium, ed. C.W. Parkinsand J.W. Anderson. Ann. N.Y. Acad. Sci. 405: 510-511.Sachs, M.B., and M.I. Miller. 1985. Pitch coding in theauditory nerve: possible mechanisms of pitch sensationwith cochlear implants. In Cochlear Implants, ed. R.A.Schindler and N.M. Merzenich, 185-194. New York: RavenPress.Sachs, M.B., and E.D. Young. 1980. Effects of nonlinearitieson speech encoding in the auditory nerve. J. Acoust.Soc. Am. 68: 858—875.Schouten, J.F. 1938. The perception of subjective tones.Proc. Kon. Ned. Acad. Wetensch. 41: 1086—1093.Schouten, J.F. 1940. The residue and the mechanism ofhearing. Proc. ICon. Ned. Acad. Wetensch. 43: 991-999.Schreiner, C.E., and C. Langner. 1988. Periodicity coding inthe inferior colliculus of the cat: II. Topographicalorganization. J. Neurophysiol. 60: 1823-1840.Schubert, E.D. 1983. Use of the pitch response in assessingcochlear implants. Abstracts of the Sixth MidwinterResearch Meeting, Association for Research inOtolaryngology, 127.180Shallop, J.K., A.L. Beiter, D.W. Goin, and R.E. Mischke.1990. Electrically evoked auditory brainstem responses(EABR) and middle latency responses (EMLR) obtained frompatients with the Nucleus Multichannel Cochlear Implant.Ear Hear 11: 5-15.Shannon, R.V. 1983. Multichannel electrical stimulation ofthe auditory nerve in man. I. Basic psychophysics.Hearing Res. 11: 157—189.Shannon. R.V. 1989. Detection of gaps in sinusoids and pulsetrains by patients with cochlear implants. J. Acoust.Soc. Am. 85: 2587—2592.Shannon, R.V. 1992. Temporal modulation transfer functionsin patients with cochlear implants. J. Acoust. Soc. Am.91: 2156—2164.Shannon, R.V., D.D. Adams, R.L. Ferrel, R.L. Palumbo, and M.Grandgenett. 1990. A computer interface forpsychophysical and speech research with the Nucleuscochlear implant. J. Acoust. Soc. Am. 87: 905-907.Shuter-Dyson, R. 1982. Musical ability. In The Psychologyof Music, ed. D. Deutsch, 391-412. New York, London:Academic Press.Siegel, J.A., and W. Siegel. 1977. Absolute identificationof notes and intervals by musicians. Perception &Psychophysics 21: 143—152.Simmons, F.B. 1966. Electrical stimulation of the auditorynerve in man. Arch. Otolaryngol. 84: 24-76.Simmons, F.B. 1983. Percepts from modiolar (eighth nerve)stimulation. In Cochlear Prostheses: An InternationalSymposium, ed. C.W. Parkins and J.W. Anderson. Ann. N.Y.Acad. Sci. 405: 259—263.Simmons, F.B., R.G. Mathews, M.G. Walker, and R.L. White.1979. A functioning multichannel auditory nervestimulator. Acta. Otolaryngol. 87: 170-175.Simmons, F.B., M.G. Walker, R.L. White, and R.G. Mathews.1981. Pitch correlates of direct auditory nerveelectrical stimulation. Ann. Otol. Rhinol. Laryngol.90(Suppl 82): 15—18.Sloboda, J.A., and D.H.H. Parker. 1985. Immediate recall ofmelodies. In Musical Structure and Cognition, ed. I.Cross, P. Howell, and R. West, 143-167. London:Academic Press Inc.181Small, A.M. 1955. Some parameters influencing the pitch ofamplitude modulated signals. J. Acoust. Soc. Am. 27:751—760.Small, A.M., and M.E. McClellan. 1963. Pitch associated withtime delay between two pulse trains. J. Acoust. Soc. Am.35: 1246—1255.Spillmann, T., N. Dillier, and J. Guentensperger. 1982.Electrical stimulation of hearing by implanted cochlearelectrodes in humans. Appi. Neurophysiol. 45: 32-37.Stevens, S.S., and J. Volkmann. 1940. The relation of pitchto frequency: a revised scale. Am. J. Psychol. 53: 329-353.Stypulkowski, P.H., and C. van den Honert. 1984.Physiological properties of the electrically stimulatedauditory nerve. I. compound action potential recordings.Hearing Res. 14: 205—223.Sullivan, W.E., and M. Konishi. 1986. Neural map ofinteraural phase difference in the owl’s brainstem.Proc. Nati. Acad. Sci. USA 83: 8400-8404.Takahashi, T., A. Moiseff, and M. Konishi. 1984. Time andintensity cues are processed independently in theauditory system of the owl. J. Neurosci. 4: 1781-1786.Tong, Y.C., R.C. Black, G.M. Clark, I.C. Forster, J.B. Millar,B.J. O’Loughlin, and J.F. Patrick. 1979. A preliminaryreport on a multiple—channel cochlear implant operation.J. Laryngol. Otol. 93: 679—695.Tong, Y.C., P.J. Blarney, R.C. Dowell, and G.M. Clark. 1983.Psychophysical studies evaluating the feasibility of aspeech processing strategy for a multiple—channelcochlear implant. J. Acoust. Soc. Am. 74: 73-80.Tong, Y.C., and G.M. Clark. 1983. Percepts from scalatympani stimulation. In Cochlear Prostheses: AnInternational Symposium, ed. C.W. Parkins and J.W.Anderson. Ann. N.Y. Acad. Sci. 405: 264—267.Tong, Y.C., and G.M. Clark. 1985. Absolute identification ofelectric pulse rates and electrode positions by cochlearimplant patients. J. Acoust. Soc. Am. 77: 1881—1888.Tong, Y.C., and G.M. Clark. 1986. Loudness summation,masking, and temporal interaction for sensations producedby electric stimulation of two sites in the humancochlea. J. Acoustic. Soc. Am. 79: 1958—1966.182Tong, Y.C., G.M. Clark, P.J. Blamey, P.A. Busby, and R.C.Dowell. 1982. Psychophysical studies for two multiple-channel cochlear implant patients. J. Acoust. Soc. Am.71: 153—160.Tong, Y.C., J.B. Millar, G.M. Clark, L.F. Martin, P.A. Busby,and J.F. Patrick. 1980. Psychophysical and speechperception studies on two multiple channel cochlearimplant patients. J. Laryngol. Otol. 94: 1241-1256.Townshend, B., N. Cotter, D. Van Compernolle, and R.L. White.1987. Pitch perception by cochlear implant subjects. J.Acoust. Soc. Am. 82: 106-115.Tyler, R.S., B.C.J. Moore, F.K. Kuk. 1989. Performance ofsome of the better cochlear-implant patients. J. SpeechHear. Res. 32: 887—911.van den Honert, C., and P.H. Stypulkowski. 1984.Physiological properties of the electrically stimulatedauditory nerve. II. Single fiber recordings. HearingRes. 14: 225—243.van den Honert, C., and P.H. Stypulkowski. 1987a. Singlefiber mapping of spatial excitation patterns in theelectrically stimulated auditory nerve. Hearing Res. 29:195—206.van den Honert, C., and P.H. Stypulkowski. 1987b. Temporalresponse patterns of single auditory nerve fiberselicited by periodic electrical stimuli. Hearing Res.29: 207—222.Vernon, M. 1967. Meningitis and deafness: the problem, itsphysical, audiological psychological and educationalmanifestations in deaf children. Laryngoscope 77: 1856-1874.von Wallenberg, E.L., and R.D. Battmer. 1991. Comparativespeech recognition results in eight subjects using twodifferent coding strategies with the Nucleus 22 channelcochlear implant. British J. Audiology 25: 371-380.Wapnick, J., G. Bourassa, and J. Sampson. 1982. Theperception of tonal intervals in isolation and in melodiccontext. Psychomusicology 2: 21-37.Ward, W.D. 1970. Musical Perception. In Foundations ofModern Auditory Theory, ed. J.V. Tobias, 405-447. NewYork: Academic Press.Warren, R.M., and J.A. Bashford. 1988. Broadband repetitionpitch: spectral dominance or pitch averaging? J. Acoust.Soc. Am. 84: 2058—2062.183Watkins, A.J., and M.C. Dyson. 1985. On the perceptualorganization of tone sequences and melodies. In MusicalStructure and Cognition, ed. I. Cross, P. Howell, and R.West, 71-119. London: Academic Press Inc.Wever, E.G. 1949. Theory of Hearing. New York, Dover.Wever, E.G., and C.W. Bray. 1930. Action currents in theauditory nerve in response to acoustical stimulation.Proc. Nat. Acad. Sci. (U.S.) 16: 344-50.Wing, H. 1948. Tests of musical ability and appreciation.Brit. J. Psychol. Monogr. (Suppl.27).Woodworth, R.S. 1938. Experimental Psychology. New York:Henry Holt and Company.184APPENDIX 1OPEN-SET TUNES*1. TWINKLE TWINKLE LITTLE STAR (14)ccggaagffeeddc222222422222242. JINGLE BELLS (11)eeeeeeegcde224224222243. LONDON BRIDGE IS FALLING DOWN (13)gagfefgdefefg31222242242244. 0 CANADA (10)eggcdefgad43162222245. 0 COME ALL YE FAITHFUL (12)f f c f g c a g a a g2422442222 446. SILENT NIGHT (14)f g f d f g f d C C a bb bb f312631264264267. MARY HAD A LITTLE LAMB (13)edcdeeedddegg31222242242248. HAPPY BIRTHDAY (12)ccdcfeccdcgf1122241122249. GOOD KING WENCESLAS (13)fffgffcdcdeff222222422224410. AULD LANG SYNE (14)cfefagfgaffaCD2312231223122411. CLEMENTINE (15)fffcaaaffaCCbbag1122112211311 1412. HARK THE HERALD ANGELS SING (15)c f f e f a a g C C C a g a223122222231224185Appendix 1 - Continued13. DECK THE HALLS (17Cb agfgafgabbgag f ef3122222211113122414. FRERE JACQUES (14)cdeccdecefgefg2222222222422415. 0 SUZANNA (14)ceggagecdeedcd2223122312222416. OLD MACDONALD HAD A FARM (12)fffcddcaaqgf22222242222417. HOME ON THE RANGE (11)c c f g a f e d bb bb bb222241122 2 418. YANKEE DOODLE (14)ffgafagcffgafe2222222222224219. STAR SPANGLED BANNER (12)g e C e g C E D C e f# g11222411222 420. GOD SAVE THE QUEEN (16kddec def# f#gf#edeac#d222312222312222421. BICYCLE BUILT FOR TWO (10)Cafcdefdfc333311131422. AWAY IN A MANGER (13)c f f g a f f a C C D222112211 222423. JESUS LOVES ME (14)geedeggaaCaagg2222224222222424. WE WISH YOU A MERRY CHRISTMAS (16)cffgfedddggagfec221111222211112225. 0 WHEN THE SAINTS (16)cefgcefgcefgeced2226222622244446186Appendix 1 - Continued26. WALTZING MATILDA (20)gggge CCCbagggagggf ed2112221122211211211227. GOD REST YOU MERRY GENTLEMEN (14)ddaagfedcdefga2222222222222428. ON TOP OF OLD SMOKEY (11)ccegCaafgag2222642222629. FOR HE’S A JOLLY GOOD FELLOW (16)ceeedefeedddcdec242222642422226430. ON THE FIRST DAY OF CHRISTMAS (13)c c c f f f e f g a bb g a22422422222 24*Numbers preceding the tunes correspond to those of thehorizontal axis in Figure 7. Numbers in parentheses representthe number of notes in tune segment. Pulse rate and note durationassignments:c = 100 pps f = 133 pps a#= 178 ppsc= 106 f#= 141 b = 189d = 112 g = 150 C 200d#= 119 g#= 159 D = 224e = 126 a = 168 E = 2521=250msec 3=75omsec 6=l500msec2 = 500 msec 4 = 1000 msec187APPENDIX 2PULSE RATES FOR 7-NOTE MELODIES; BASE PULSE RATE 100 PPS.LARGEST INTMEL.* RANGE (pps) NOTES (pps) INTERVAL EXTENT**1 119—150 150 133 1l 133 150 150 150 M2nd 3G F E F G G G2 100—168 100 126 150 150 168 150 126 M3rd 9C E G G A G E3 100—168 100 100 150 150 168 168 150 5th 9C C G G A A G4 119—150 iig iig 133 150 iig 150 133 M3rd 5E E F G E G F5 112—150 133 150 133 iig 112 iig 133 M2nd 5F G F E D E F6 112—150 150 150 150 112 126 126 112 4th 5G G C D E E D7 112—150 150 126 126 112 126 150 150 m3rd 5G E E D E G G8 106—159 141 141 141 15 141 141 10 4th 5F F F F C*MELODIES:1. Mary Had a Little Lamb 5. London Bridge2. 0 Suzanna 6. old MacDonald3. Twinkle, Twinkle, Little Star 7. Jesus Loves Me4. Yankee Doodle 8. Good King Wenceslas** Range in semitones between highest and lowest notes of themelodies. Largest interval in melody (M = Major;m = Minor)

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0099227/manifest

Comment

Related Items