Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Individual differences in external-ear transfer functions of cats Li Xua) and John C. Middlebrooksb) Kresge Hearing Research Institute, University of Michigan, 1301 East Ann Street, Ann Arbor, Michigan 48109-0506 共Received 6 July 1999; accepted for publication 7 December 1999兲 Knowledge of the direction-dependent filter characteristics of the external ears is useful for the study of spatial hearing in experimental animals. The present study examined individual differences in the directional components of external-ear transfer functions 共directional transfer functions, DTFs兲 among 24 anesthetized cats. Ears were fixed in a frontal position. Inter-cat differences in DTFs were quantified across a mid-frequency range from 8 to 16 kHz and across 30 locations in the horizontal plane and vertical midline. Across cats, DTFs showed similar direction dependence, but tended to differ in regard to the center frequencies of spectral features, such as spectral peaks and notches. Certain mid-frequency notches, for instance, varied in frequency across cats by nearly a factor of 2. Scaling of DTFs in frequency could reduce the overall differences between pairs of cats. Scale factors that minimized inter-cat differences ranged as high as 1.57 and correlated moderately with cats’ body weights. Nevertheless, appreciable individual differences remained after frequency scaling. Inter-cat differences in DTFs were substantially larger than differences that resulted from variability in positioning the ears. The results suggest some guidelines regarding the conditions under which it is acceptable to apply DTF measurements from one cat to another. © 2000 Acoustical Society of America. 关S0001-4966共00兲02903-9兴 PACS numbers: 43.64.Ha, 43.66.Pn, 43.66.Qp 关BLM兴 INTRODUCTION As a sound passes from a free-field source to the ear canal, its spectrum is transformed by interaction with the head and external ear. The transfer function is commonly known as the head-related transfer function 共HRTF兲 共Butler, 1987; Morimoto and Ando, 1980; Wightman and Kistler, 1989a兲. The HRTF varies with the angle of incidence of sound, so the spectrum of the sound in the ear canal carries directional information. Domestic cats are widely used as experimental animals for studies of spatial hearing, both for cortical physiology 共Brugge et al., 1994, 1996, 1998; Imig et al., 1990; Middlebrooks and Pettigrew, 1981; Middlebrooks et al., 1998; Rajan et al., 1990; Xu et al., 1998, 1999兲 and for psychophysics 共Casseday and Neff, 1973; Jenkins and Masterton, 1982; Huang and May, 1996; May and Huang, 1996; Populin and Yin, 1998兲. Knowledge of cats’ HRTFs can guide physiological study of specific spatial cues 共e.g., Rose et al., 1966兲 and can inspire models of spatial sensitivity 共e.g., Middlebrooks, 1987; Rice et al., 1992兲. In several recent studies, cat HRTFs have been used for synthesis of a virtual auditory space 共Brugge et al., 1994, 1996, 1998; Chan et al., 1993; Delgutte et al., 1999; Nelken et al., 1997; Reale et al., 1996兲. Human HRTFs show prominent differences among listeners 共Mehrgardt and Mellert, 1977; Middlebrooks, 1999a; Shaw, 1966; Wightman and Kistler, 1989a兲. The functional significance of such individual differences is demonstrated by studies of virtual sound localization. When human subjects listen through HRTFs recorded from other subjects, a兲 Electronic mail: [email protected] Electronic mail: [email protected] b兲 1451 J. Acoust. Soc. Am. 107 (3), March 2000 they show substantial errors in localization 共Middlebrooks, 1999b; Wenzel et al., 1993兲. This raises concern about the validity of generalizing HRTFs measured from any particular cat to a variety of other cats. The present study characterized the individual differences in cat HRTFs across a sample of 24 cats. Particular attention was paid to the directional components of HRTFs, which are known as directional transfer functions 共DTFs; Middlebrooks and Green, 1990兲. The results showed that the amount of individual differences in cat DTFs is comparable to, if not larger than, that of inter-subject differences in human DTFs. The individual differences could be reduced by frequency scaling, and the scale factors were related to the body weights of the animals. That indicates that there existed a relationship between the physical sizes of the animals and the frequencies of the prominent spectral features in the external-ear transfer functions. I. METHOD A. Animal preparation Twenty-four purpose-bred adult cats 共11 females and 13 males兲 were used in this study. Three of them were supplied by Sinclair Research Center, MO, and the rest were supplied by Harlan Industrials, Inc., IN. Weights ranged from 2.4 to 6.1 kg. For the 21 cats obtained from Harlan Industrials, Inc., three shared a father, six pairs shared fathers, and the remaining six cats were not related directly. Measurements of the head-related transfer function were preceded by single-unit recording in the right auditory cortex 共Furukawa et al., 2000; Middlebrooks et al., 1998; Xu et al., 1998, 1999兲. Each cat was anesthetized for surgery with isoflurane, then was transferred to ␣-chloralose for single-unit recording followed by acoustic measurement. A stainless-steel fixture was attached 0001-4966/2000/107(3)/1451/9/$17.00 © 2000 Acoustical Society of America 1451 to the frontal bones with screws and dental cement. A midline scalp incision was made, the temporalis muscle was retracted on the right side, and a recording chamber was positioned over the right auditory cortex. Portions of scalp and temporalis muscle were removed to accommodate the recording chamber, and the scalp incision was closed to minimize distortion of the ears. The presence of the skull opening and recording chamber almost certainly influenced details of external-ear transfer functions. Nevertheless, essentially the same opening and chamber were present in all animals, so there should have been minimal influence on the inter-cat differences. Also, inter-cat differences that varied with body weight and gender 共Sec. II C兲 would not have been affected by the recording chamber. The animal was transferred to the center of a soundattenuating chamber with the cat’s interaural axis centered in the sound chamber. The body of the cat was supported in a canvas sling, and its head was supported from behind by a bar attached to the skull fixture. Under anesthesia, the relaxed pinna position tended to be lateral and downward in comparison to the pinna position of an alert cat. Care was taken to restore the normal-appearing position of the pinnae. Both ears were supported by a pair of curved wire supports so that the pinnae resumed a symmetrical forward position that resembled the pinna position adopted by a cat attending to a frontal sound. B. Experiment setup and acoustic system Experiments were conducted in a sound-attenuating chamber that was lined with acoustical foam 共Illbruck兲 to suppress reflections of sounds at frequencies ⬎500 Hz. Sound stimuli were presented from loudspeakers 共Pioneer model TS-879 two-way coaxials兲 mounted on two circular hoops, 1.2 m in radius, one in the horizontal plane and one in the vertical midline plane. On the horizontal hoop, 18 loudspeakers spaced by 20 degrees covered 360 degrees. On the vertical hoop, 14 loudspeakers spaced by 20 degrees ranged from 60 degrees below the frontal horizon, up and over the top, to 20 degrees below the rear horizon. Vertical locations were labeled continuously in 20-degree steps from ⫺60 to ⫹200 degrees. Since the vertical and horizontal hoops shared speakers at 0 and 180 degrees, the total number of locations was 30. Experiments were controlled with an Intel-based personal computer. Acoustic stimuli were synthesized digitally, using equipment from Tucker-Davis Technologies. The sampling rate for audio output was 100 kHz, with 16-bit resolution. C. Measurements of transfer functions A 12-in. probe microphone 共Larson-Davis model 2540兲 was used for the acoustic measurements. The stainless-steel probe tube that connected to the 21-in. microphone had an inner diameter of 3.2 mm and an outer diameter of 4.0 mm. The probe tube was inserted approximately 2 mm into the ear canal through a vertical incision at the posterior base of the pinna. The probe stimuli delivered from each of the 30 loudspeakers were pairs of Golay codes 共Foster, 1986; Golay, 1961; Zhou et al., 1992兲 that were 81.92 ms in duration. 1452 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 Recordings from the microphone were amplified, then were digitized at a rate of 100 kHz, yielding a spectral resolution of 12.2 Hz from 0 to 50 kHz. Before each experiment, the frequency responses of the loudspeakers were measured using the Golay stimuli and the 1 2-in. microphone positioned in the center of the sound chamber in the absence of the animal. The influence of the frequency responses of the loudspeakers was eliminated by dividing them from the ear-canal spectra. The resulting transfer functions were head-related transfer functions 共HRTFs兲. A similar procedure has been used in previous studies of cats’ external-ear transfer functions 共Musicant et al., 1990; Rice et al., 1992兲. In the present study, our analysis was based on the direction-specific components of the HRTFs, referred to as directional transfer functions 共DTFs兲. The DTF for each sound source location was obtained by dividing the HRTF for each location by the root-mean-squared HRTF averaged across all 30 locations 共Middlebrooks and Green, 1990; Middlebrooks, 1999a兲. In control experiments performed in three cats, recordings from the probe microphone in the ear canal were made simultaneously with recordings from a Knowles miniature microphone 共model 1934兲 that was placed at a deeper position in the ear canal. Consistent with such dual-microphone recordings in humans 共Middlebrooks et al., 1989; Middlebrooks and Green, 1990兲, HRTFs measured at the two sites were quite different, whereas DTFs computed for the two recording sites were nearly identical. D. Data analysis We convolved each DTF in the linear frequency scale with a bank of bandpass filters to obtain samples spaced equally on a logarithmic scale 共Middlebrooks, 1999a兲. The purposes of this procedure were 共1兲 to roughly equalize the number of samples per perceptual critical band and 共2兲 to facilitate scaling of DTFs in frequency. The filter bank consisted of 118 triangular filters. The 3-dB bandwidth of the filters was 0.0571 oct, filter slopes were 105 dB per oct, and the center frequencies were spaced in equal intervals of 0.0286 oct from 3 to 30 kHz. The interval of 0.0286 oct was chosen to give intervals of 2% in frequency. As a consequence of bandpass filtering, the very sharp peaks or notches in the DTFs were blunted, which resembled the perceptual consequence of filtering by the cochlea 共Glasberg and Moore, 1990兲. Individual differences between DTFs were quantified by a metric that will be referred to as the inter-subject spectral difference. That metric was computed across the 36 filterbank components from 8 to 16 kHz. That mid-frequency band was chosen because it contains prominent directiondependent spectral features 共Musicant et al., 1990; Neti et al., 1992; Rice et al., 1992兲 and because our preliminary tests indicated that it captured the most systematic differences in the DTFs among cats. Behavioral studies have demonstrated that cats can accurately localize noise that is limited to mid-frequencies 共Huang and May, 1996兲, and we have shown that mid frequency sounds have the strongest influence on cat neuronal signaling of sound locations 共Xu et al., 1999兲. The spectral difference was computed as follows. For each pair of cats, the dB amplitudes of the 36 L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1452 components of DTFs were subtracted, component by component, to form a difference spectrum for each source direction. Then, the variance across the 36 components of the difference spectrum was computed. Finally, the variance was averaged across the 30 tested directions for each pair of cats. We quantified systematic frequency differences between the DTFs of pairs of cats by using a frequency-scaling procedure 共Middlebrooks, 1999a兲. That procedure served as a tool to quantify frequency disparities between DTFs and to evaluate the relationship between the physical sizes of the animals and the spectral features of their DTFs. For each pair of cats, the set of 30 DTFs from one cat was scaled upward by 2%, alternating with downward 2% scaling of the other cat’s set of DTFs. Amplitudes of the DTFs were not adjusted. After each scaling, the spectral difference was computed between 8 and 16 kHz. This procedure was repeated until the scale factor was found that produced the minimum spectral difference between any given pair of cats. That was termed the optimal scale factor. The percent reduction in the inter-subject spectral differences as a result of optimal frequency scaling was also computed. II. RESULTS A. General properties of cat DTFs For the 24 cats used in the present study, DTFs were measured from all 24 left ears and 22 of the right ears. For the 22 cats in which both ears were measured, the spectral differences in DTFs between left and right ears averaged 11.3 dB2. The right ear was somewhat more distorted than the left by the right-sided skull opening and recording chamber 共see Sec. I兲. That asymmetry presumably contributed to the left/right spectral differences. Our subsequent data analysis was based on recordings from the 24 left ears. Figure 1共a兲 shows 14 DTFs of one cat 共cat 19兲 before they were processed with the bandpass filter bank. Filtering with the bandpass filter bank resulted in significant smoothing of the spectra and yet preserved the major spectral features 关thick line in Fig. 1共b兲兴. One of the prominent features in the cat DTFs was the mid-frequency notches, as has been emphasized in previous studies of cat HRTFs 共Musicant et al., 1990; Rice et al., 1992兲. Generally, the notch center frequencies tended to increase systematically as source elevation increased. The trend often was violated, however, for DTFs at lowest elevations. Also, the DTFs at the highest locations 共⫹80 and ⫹100 degrees兲 in about half of the cats failed to show any obvious spectral features such as notches or peaks. Changes in the position of a cat’s ears can produce systematic changes in HRTFs 共Middlebrooks and Knudsen, 1987; Young et al., 1996兲. When we compared DTFs among different cats, we attempted to position the ears in a ‘‘normal-appearing’’ forward position that was consistent across all subjects. In six cats, we tested the reliability with which a consistent ear position could be attained. In each of the six cats, the ears and the probe microphone were positioned twice, once by each of two investigators, and two sets of DTF measurements were made. Figure 1共b兲 共thin and thick lines兲 shows examples of DTFs measured by two different experimenters. The two measurements matched each 1453 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 FIG. 1. Directional transfer functions 共DTFs兲 from the left ear of one cat 共cat 19兲 for the 14 elevation locations in the vertical midline plane. 共a兲 Unfiltered DTFs. 共b兲 Two sets of measurements processed by a bank of bandpass filters 共see Sec. I for details兲. Thick lines are from the same set of measurements as shown in 共a兲. Thin lines are after a second experimenter repositioned the probe tube microphone and the pinnae. other fairly well at all locations and across frequencies. The spectral difference between the two measurements computed across all 30 locations was 3.25 dB2. The spectral differences of the two replicated measurements in the other five cats were 0.56, 2.83, 3.04, 5.80, and 7.10 dB2. These spectral differences are considerably smaller than the spectral differences between pairs of cats 共detailed later in Fig. 4兲. In another subset of three cats, we deliberately changed the pinna to extreme forward and backward positions beyond those that could normally be obtained in a behaving cat. This was done by pulling a thin string attached to the tip of the pinna and by pushing or retracting the wire support of the pinna. We found that the spectral differences between those extreme pinna positions averaged 14.19 dB2 (n⫽9, i.e., three cats times three possible combinations of pinna positions兲, which was smaller than the mean spectral differences between pairs of cats 共detailed later in Fig. 4兲. Young et al. 共1996兲 tested the influence of more moderate change of ear position on spectral notches of cat HRTFs. Data illustrated in their figure 6 showed that inter-cat differences were substantially larger than differences resulting from changes in ear position. Therefore, we were confident that the individual differences in cat DTFs were not solely due to the differences in ear position. L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1453 FIG. 2. The DTFs from five cats for selected locations. The upper left panel represents DTFs for a location in the horizontal plane 40 degrees ipsilateral to the ear that contained the microphone, and the other five panels represent DTFs for ⫺40- to 120degree elevations in 40-degree steps in the vertical midline plane as labeled above the DTFs. Arrows indicate the spectral notches defined in the text. In all panels, DTFs from the five cats are ordered based on the spectral notch frequencies in the DTFs for ⫹40degree elevation. The same order is given by a rank of body weights of the cats; that is, DTFs for the lightest cat are placed on the top and those for the heaviest cat at the bottom. Cats 8, 12, 15, 1, and 11 had body weights of 2.5, 2.9, 3.6, 4.0, and 6.1 kg, respectively. B. Individual differences Despite some similarity in the general features of the DTFs from different cats, we observed substantial individual differences in the DTFs. Figure 2 plots DTFs from five cats at six locations. The upper left panel represents DTFs at ipsilateral 40 degrees on the horizontal plane. The rest of the panels represent DTFs from five midline elevation locations at ⫺40 to ⫹120 degrees in 40-degree steps. Arrows in those panels indicate mid-frequency notches in the DTFs. A spectral notch was defined here as a spectral minimum in the frequency range between 5 and 20 kHz, bounded on either side by slopes that were ⭓5 dB high. The DTFs plotted in each panel are ordered based on the spectral notch frequen- cies defined at ⫹40 degrees elevation, and the order of cats is the same across all panels. The inter-subject differences in the DTFs included differences in the detailed shape of each DTF and in the frequencies of the prominent spectral features such as spectral notches and peaks. The prominent mid-frequency spectral notches highlight the individual differences among cat DTFs. Figure 3 summarizes the inter-subject differences in the spectral notch center frequencies in the DTFs of the 14 elevation locations. Figure 3共a兲 plots the center frequencies of notches for the five cats whose DTFs are shown in Fig. 2. Each line represents data from one of the cats. The lines are broken at locations at which no spectral notches could be identified. In frontal lo- FIG. 3. Individual differences in the spectral notch center frequencies of the DTFs. 共a兲 Each line plots for one of the five cats shown in Fig. 2 the spectral notch frequencies in the DTFs for the 14 elevation locations. 共b兲 The box plot summarizes the distribution of spectral notch frequencies in the 14 elevation DTFs of all 24 cats. At each elevation, the mean is represented by the small square and the minimum and maximum are represented by the asterisks. The three vertical bars represent the 25, 50 and 75 percentiles and the ends of the two horizontal bars represent the 5 and 95 percentiles of the distribution. The number to the right indicates the number of cats at each elevation at which mid-frequency notches could be identified. 1454 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1454 FIG. 4. Distributions of inter-subject spectral differences in DTFs before 共a兲 and after 共b兲 frequency scaling. Values were computed for all 276 possible pairwise combination of 24 cats. cations, notch frequencies increased systematically as source elevation increased from ⫺20 to ⫹80 degrees. In rear locations, notch frequencies also increased as sources moved up from the rear horizon. Individual differences in the notch frequencies of cat DTFs were apparent. The lines were roughly parallel to each other, at least for frontal locations, suggesting that the set of all DTFs from one cat could be aligned with that of another cat by scaling in frequency. The distributions of the individual differences in spectral notch frequency among the sample of 24 cats are illustrated in Fig. 3共b兲 in a box-plot format. In this format, the mean notch frequency at each location is represented by the small square and the minimum and maximum values are represented by the asterisks. The three vertical lines represent 25, 50, and 75 percentiles and the ends of the two horizontal bars represent 5 and 95 percentiles of the distribution, respectively. The number of cats in which mid-frequency notches could be identified varied from elevation to elevation. Those numbers are given in a column on the right side of Fig. 3共b兲. There was a clear trend that the spectral notch center frequencies varied systematically with the frontal midline elevation. Nevertheless, the inter-subject differences in the spectral notch frequencies could be as large as 1 oct for most locations in elevation. The variation was especially large in the rear locations. The smallest range of notch frequencies was found at 0 degrees elevation at which they varied over only 0.60 oct in our sample of 24 cats. The extent of the individual differences in the spectral notch frequencies can also be appreciated from another perspective. If one makes a vertical line in Fig. 3共b兲 anywhere between 10 and 12 kHz, it passes within the distribution of notch frequencies at all el1455 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 FIG. 5. Examples of frequency scaling between a pair of cats 共cat 11 and cat 21兲. 共a兲 The DTFs from cat 11 共thin lines兲 and cat 21 共thick lines兲 for the eight frontal elevation locations before frequency scaling. 共b兲 At optimal frequency scaling, DTFs from cat 11 共thin lines兲 were scaled upward by 1.219 and those from cat 21 共thick lines兲 were scaled downward by 1.195, a combined relative scaling of 1.457. The vertical grid lines indicate the frequency band 共8–16 kHz兲 over which quantitative comparisons were made. evations. That means that a notch at, say, 11 kHz could signal nearly any location among one or more of the cats in our sample. We computed the spectral differences in the DTFs averaged across all 30 locations for each pair of cats. These spectral differences were computed across 36 frequency components from 8 to 16 kHz 共see Sec. I for details兲. The distribution of the spectral differences across all of the 276 pairwise combination of cats is shown in Fig. 4共a兲. The mean of the distribution was 17.02 dB2 and the spectral differences of 9.5% of pairs were ⭓25 dB2. C. Frequency scaling The notch data in Fig. 3共a兲 showed that inter-cat differences in notch frequencies tended to be fairly constant across most locations. That suggests that inter-cat alignment of spectral features could be improved, and spectral differences decreased, by scaling the sets of DTFs for each pair of cats by a single scale factor. That procedure was followed as described in Sec. I. An example is shown for one pair of cats in Fig. 5. Panels 共a兲 and 共b兲 show the eight frontal-midline DTFs of the two cats 共cat 11 and cat 21兲 before and after frequency scaling. At optimal scaling, the DTFs from cat 11 were scaled upward by 2% ten times and the DTFs from cat 21 were scaled downward nine times, resulting in a total relative scale factor of 1.457 (1.0210/1.02⫺9 ⫽1.457). Note that the prominent spectral features 共e.g., notches兲 in the two sets of DTFs matched well after frequency scaling. The spectral difference of this pair of cats was reduced by 51.3% from L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1455 FIG. 6. Spectral difference as a function of frequency scale factor for the pair of cats represented in Fig. 5 共cat 11 versus cat 21兲. The optimal scale factor is indicated by the filled symbol. 34.5 dB2 before scaling to 16.5 dB2 after. Figure 6 shows the spectral differences for this pair of cats as a function of relative scale factor. The distribution of spectral differences for the 276 pairs of cats after scaling is shown in Fig. 4共b兲. The mean was 14.01 dB2. Only one pair showed a spectral difference ⭓25 dB after scaling. Figure 7共a兲 and 共b兲 shows the distributions of the optimal frequency scale factor and the percent reduction of spectral differences by frequency scaling. For this illustration, all pairs were ordered such that the optimal scale factor was ⭓1. The optimal scale factors ranged from 1 to 1.57 with a median of 1.10. The percent reduction in the inter-subject spectral differences accomplished by frequency scaling had a median of 11.5%, ranging from 0% to 60.7%. A substantial amount of spectral difference remained even after frequency scaling due to nonsystematic differences in the DTFs among cats that could not be removed by frequency scaling. The spectral details of a DTF result from the physical FIG. 7. Distributions of optimal scale factor 共a兲 and percent reduction of spectral difference by frequency scaling 共b兲. Values were computed for all 276 possible pairwise combination of 24 cats. 1456 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 FIG. 8. Relationship between cat body weight and mean frequency scale factor. Eleven female and 13 male cats are represented by the circles and squares, respectively. Solid and dashed lines are linear fits to the data from female and male cats, respectively. interaction of an incident sound wave with the head and pinnae. For that reason, one would expect that the frequencies of spectral features depend on the sizes of the head and pinnae as well as the shapes of the pinnae. We used the body weight of each cat as a single number to represent its overall size. The body weights correlated somewhat with the frequencies of spectral features in the DTFs. For example, for the DTFs of the five cats shown in Fig. 2, which were sorted by the notch frequency in the DTF for ⫹40 degrees elevation, the body weights fell in order such that the lightest cat was represented by the top line and the heaviest cat by the bottom line. In general, spectral features in the DTFs of a smaller cat tended to fall in a higher frequency region than those of a bigger cat 共Fig. 2兲. We summarized the frequencies of spectral features in each cat’s DTFs as the geometric mean of the optimal scale factor between that cat and every other cat. Figure 8 plots the mean optimal scale factor for each cat as a function of its body weight. The circles represent female cats and squares represent male cats. The solid and dashed lines are linear 共least-square兲 fits for each set of data. Within each gender, mean scale factor correlated with body weight. Correlation coefficients 共r兲 were 0.609 for females and 0.806 for males. The correlation was reduced to 0.547 when females and males were combined. The gender differences were interesting. We measured the pinna sizes in a few cats and found that female cats tended to have bigger pinnae compared to male cats of similar body weights. Conversely, given males and females having heads and ears of similar sizes, the males tended to be more muscular and, thus, heavier. We speculate that the gender differences would disappear if we could use the pinna or head size to correlate with the scale factors. Unfortunately, we did not have ear and head measurements from every cat. In a recent study of individual differences in human DTFs, we found no indication L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1456 of a gender effect that could not be accounted for by the tendency of females to be smaller than males 共Middlebrooks, 1999a兲. The correlation between the body weights and the optimal scale factor is further evidence that the inter-cat differences that we observed were not entirely due to different ear positions among cats. III. DISCUSSION The general properties of the directional components of HRTFs 共i.e., DTFs兲 that we studied were similar to previous studies of cat HRTFs, but we noted considerable individual differences in the HRTFs among cats. In the following, we compare our results to previous studies of cat HRTFs and discuss the implications of individual differences in HRTFs for studies of spatial hearing. A. Comparison with previous studies Previous descriptions of cat external-ear transfer functions have emphasized characteristics of the direction dependence of the transfer functions that are consistent among cats. Musicant and colleagues 共1990兲 measured the HRTFs from six cats for sounds at numerous horizontal and vertical locations. They observed prominent spectral notches in the 8–18 kHz frequency region. Individual differences were considered small. Rice and colleagues 共1992兲 confirmed and extended the observations of the spectral notches in the HRTFs. They focused on a slightly larger frequency range 共5–18 kHz兲 and proposed a sound-localization scheme based on the mid-frequency notch frequencies in the two ears. Again, they emphasized the similarities rather than differences in the HRTFs among the five cats used in that study. Our measurements of individual DTFs were largely consistent with previous studies in regard to the general location dependence of spectral features. In general, we found that spectral features in the mid-frequency region shifted to higher frequency as sound source moved from lower elevation to higher elevation in the front midline 共Fig. 1兲. The same was true for rear midline locations, but variability there was much greater. We sampled lower elevations than in previous studies, i.e., ⫺60 degrees rather than ⫺30 degrees 共Musicant et al., 1990; Rice et al., 1992兲. At ⫺40- and ⫺60degree elevations, mid-frequency notches tended to be wide and blunt 共Figs. 1 and 2兲. The center frequencies of those notches tended to shift to higher frequency with decreasing elevation, which was counter to the trend at elevations ⭓⫺20 degrees. In a behavioral study, May and Huang 共1996兲 tested cat’s head orientation to sounds presented as low as ⫺45 degrees elevation and found that response errors increased dramatically compared to those at more centric locations. They attributed these increased errors to the poorer quality of spectral cues at lower elevations. We note that most single neurons in the cat superior colliculus show spatial tuning centered above ⫺30 degrees 共Middlebrooks and Knudsen, 1984兲. Also, coding of low elevations by neurons in the cat’s auditory cortex tends to be relatively inaccurate 共Xu et al., 1998兲. In the present study, optimal scale factors served as global measures of inter-cat differences in the frequencies of spectral features. In the sample of 24 cats, optimal scale fac1457 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 tors ranged as high as 1.57 with a median value of 1.10. Individual spectral notches varied across a larger range, up to a factor of 2 in some cases 共Fig. 3兲. Although previous studies have measured smaller samples of cats, one can see instances of variation in the frequencies of spectral features. In the study by Wiener and colleagues 共1966兲, the center frequencies of mid-frequency notches in five cats ranged from approximately 10 to 13 kHz for sound sources at 0 degrees azimuth and 0 degrees elevation 共a factor of 1.3; measured from fig. 6 in Wiener et al., 1966兲. However, in that study, the authors focused on the low-frequency features of the external-ear transfer functions and largely ignored the individual differences in frequencies higher than 10 kHz. In the data illustrated by Musicant and colleagues 共1990兲, three cats showed similar directional dependence of HRTFs, and it is difficult to see any differences in frequencies of spectral features. Rice and colleagues 共1992; Fig. 11兲 showed HRTFs from four cats that appeared very similar. A fifth animal showed a mid-frequency notch that was displaced upward by a factor of ⬃1.15, but the authors discussed several experimental factors that might have produced that difference. An additional cat measured in the same laboratory 共Young et al., 1996, Fig. 3兲 showed a HRTF for location at 0 degrees azimuth, 0 degrees elevation that was quite similar in shape to the HRTFs shown in the 1992 report but was displaced upward in frequency by a factor of ⬃1.3. Considering the smaller sample sizes of previous studies, the magnitude of inter-cat variation in published data is roughly in line with the present observations. We consider in the next section whether that magnitude of variation is likely to be significant for studies of spatial hearing. The magnitudes of individual differences in cats are roughly comparable to those observed in a similar study in humans 共Middlebrooks, 1999a兲. Optimal scale factors for DTFs from pairs of humans were somewhat smaller than our results in cats: humans exhibited a median of 1.06 and maximum of 1.38 compared to median of 1.10 and maximum of 1.57 in cats. The reduction of spectral differences by frequency scaling was more efficient in humans than in cats: median and maximum percent reductions were 15.5% and ⬃80%, respectively, in humans compared to 11.5% and 60.7% in cats. The lower percentage reduction in cats indicates that efforts to tailor DTFs from one subject for use with another might be less satisfactory in cats than in humans 共Middlebrooks, 1999b兲. We speculate that inter-cat differences in ear position might contribute spectral differences that remain after optimal frequency scaling. B. Implications of individual differences for physiological studies of spatial hearing When human subjects listen through headphones to sounds that have been filtered by HRTFs measured from their own ears, they report a well localized ‘‘virtual target’’ that appears to originate from the location of the original free-field source 共Wightman and Kistler, 1989b兲. Virtual localization technology has been applied recently to physiological studies in cats 共Brugge et al., 1994, 1996, 1998; Chan et al., 1993; Delgutte et al., 1999; Nelken et al., 1997; Reale et al., 1996兲 and other species 共Hartung and Sterbing, L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1457 1997; Keller et al., 1998兲. This approach offers several advantages, including the potential for independent manipulation of various acoustic parameters and the ability to deliver spatial cues under conditions in which it would be impossible to achieve high-quality free-field conditions 共i.e., due to interference from experimental apparatus or due to necessary surgical distortion of the external ear兲. When human subjects attempt to localize virtual sources that have been synthesized using HRTFs measured from other listeners’ ears 共‘‘nonindividualized’’ HRTFs兲 they show an increase in localization error, particularly in the vertical and front/back dimensions 共Wenzel et al., 1993兲. Systematic trends in errors can be related to the optimal frequency scale factors between subjects’ own DTFs and the DTFs from which virtual targets are synthesized 共Middlebrooks, 1999b兲. When small subjects localize targets that are synthesized from larger subjects’ ears, the most common errors are front/back confusions. When large subjects localize targets that are synthesized from smaller subjects’ ears, the most common error is an upward bias in responses. This increase in localization error resulting from use of nonindividualized DTFs raises concern about the use of nonindividualized DTFs in physiological experiments in cats, since systematic frequency differences as small as the median value in our sample of cats 共i.e., 1.10兲 produce conspicuous localization errors in humans. Based on the observations in the present study, we can speculate on some of the likely consequences of using nonindividualized DTFs in cat physiological experiments. We observed that the general shapes of DTFs and the general spatial dependence of DTFs are quite similar across cats. This indicates that basic studies of sensitivity to particular spectral features and, perhaps, studies of the general characteristics of spatial receptive fields would not suffer from use of nonindividualized DTFs. Indeed, Reale and colleagues 共1996兲 examined the receptive fields of cortical neurons in a cat using stimuli synthesized with HRTFs from three other cats. Receptive fields measured in the three conditions differed at near-threshold stimulus levels but were quite similar at higher levels. We note, however, that the HRTFs used in that study were themselves quite similar. In a study in the guinea-pig inferior colliculus, Hartung and Sterbing 共1997兲 found that receptive fields of units changed substantially when stimuli synthesized from another guinea pig’s HRTFs were exchanged for those synthesized from an animal’s own HRTFs. Spectral features 共i.e., peaks and notches兲 at particular frequencies corresponded to different locations in different cats. Thus, one might expect to encounter errors in experiments in which spatial sensitivity of neurons would be compared to an independent spatial reference, such as when auditory and visual modalities are compared in the superior colliculus 共e.g., Middlebrooks and Knudsen, 1984; Wallace et al., 1998兲. Similarly, one might expect to find distortions in any systematic progression of spatial tuning in such a midbrain structure, particularly since 共at least in humans兲 use of nonindividualized DTFs in psychophysical procedures can result in a high rate of front/back confusions that are greater than 90 degrees in magnitude. 1458 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 We recently have presented results of a study that examined the locations that are signaled by cortical neurons when presented with narrow bandpass stimuli 共Xu et al., 1999兲. By using narrow-band sounds, that study emphasized fine details of DTFs. Stimuli were presented in a free field, so cats were exposed to stimuli filtered by their own DTFs. Physiological results showed a close correspondence with the output of a model of spectral recognition that was based on DTFs measured from each cat’s own ears. That correspondence would have been largely obscured had we used nonindividualized DTFs. Indeed, two sets of DTFs that were used as examples in the report showed optimal scale factors, one cat to the other, of only 1.025, yet inter-cat differences in details of the DTFs produced markedly different results in the acoustical model and in the physiology. We regard our measurements of inter-cat differences in DTFs as conservative. All the measurements in the present study were from domestic short hair cats supplied from one of two breeding colonies. Similarly, previous studies appear to have used purpose-bred domestic short hair cats. We have by no means explored the great variability that can be found among other breeds of cats. A visit to any cat show reveals a wide variety of shapes and sizes of heads and ears. The potential for error derived from a mismatch between cats and DTFs is large. Nevertheless, the present results indicate that under certain conditions DTFs are generally similar among cats. We urge caution in interpreting results obtained with nonindividualized DTFs, but affirm that such procedures are acceptable for studies of spatial hearing when suitable care is taken to match animals by breed, size, and gender. ACKNOWLEDGMENTS We acknowledge the expert technical assistance by Zekiye Onsan. Shigeto Furukawa helped in DTF measurements. Shigeto Furukawa, Ewan Macpherson, Brian Mickey, and David Moody made valuable comments on an early draft of the manuscript. This research was supported by NIH Grant Nos. RO1 DC00420 and T32 DC00011. Brugge, J. F., Reale, R. A., and Hind, J. E. 共1996兲. ‘‘The structure of spatial receptive fields of neurons in primary auditory cortex of the cat,’’ J. Neurosci. 16, 4420–4437. Brugge, J. F., Reale, R. A., and Hind, J. E. 共1998兲. ‘‘Spatial receptive fields of primary auditory cortical neurons in quiet and in the presence of continuous background noise,’’ J. Neurophysiol. 80, 2417–2432. Brugge, J. F., Reale, R. A., Hind, J. E., Chan, J. C. K., Musicant, A. D., and Poon, P. W. F. 共1994兲. ‘‘Simulation of free-field sound sources and its application to studies of cortical mechanisms of sound localization in the cat,’’ Hear. Res. 73, 67–84. Butler, R. A. 共1987兲. ‘‘An analysis of the monaural displacement of sound in space,’’ Percept. Psychophys. 41, 1–7. Casseday, J. H., and Neff, W. D. 共1973兲. ‘‘Localization of pure tones,’’ J. Acoust. Soc. Am. 54, 365–372. Chan, J. C. K., Musicant, A. D., and Hind, J. E. 共1993兲. ‘‘An insert earphone system for delivery of spectrally shaped signals for physiological studies,’’ J. Acoust. Soc. Am. 93, 1496–1501. Delgutte, B., Joris, P. X., Litovsky, R. Y., and Yin, T. C. T. 共1999兲. ‘‘Receptive fields and binaural interactions for virtual-space stimuli in the cat inferior-colliculus neurons,’’ J. Neurophysiol. 81, 2833–2851. Foster, S. 共1986兲. ‘‘Impulse response measurement using Golay codes,’’ in I.E.E.E. 1986 Conference on Acoustics, Speech, and Signal Processing, Vol. 2 共I.E.E.E., New York兲, pp. 929–932. L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1458 Furukawa, S., Xu, L., and Middlebrooks, J. C. 共2000兲. ‘‘Coding of soundsource location by ensembles of cortical neurons,’’ J. Neurosci. 20, 1216– 1228. Glasberg, B. R., and Moore, B. C. J. 共1990兲. ‘‘Derivation of auditory filter shapes from notched noise data,’’ Hear. Res. 47, 103–138. Golay, M. J. E. 共1961兲. ‘‘Complementary series,’’ IRE Trans. Inf. Theory 7, 82–87. Hartung, K., and Sterbing, S. J. 共1997兲. ‘‘Generation of virtual sound sources for electrophysiological characterization of auditory spatial tuning in the guinea pig,’’ in Acoustical Signal Processing in the Central Auditory System, edited by J. Syka 共Plenum, New York兲, pp. 407–412. Huang, A. Y., and May, B. J. 共1996兲. ‘‘Sound Orientation behavior in cats. II. Mid-frequency spectral cues for sound localization,’’ J. Acoust. Soc. Am. 100, 1070–1080. Imig, T. J., Irons, W. A., and Samson, F. R. 共1990兲. ‘‘Single-unit selectivity to azimuthal direction and sound pressure level of noise bursts in cat high-frequency primary auditory cortex,’’ J. Neurophysiol. 63, 1448– 1466. Jenkins, W. M., and Masterton, R. B. 共1982兲. ‘‘Sound localization: Effects of unilateral lesions in central auditory system,’’ J. Neurophysiol. 47, 987–1016. Keller, C. H., Hartung, K., and Takahashi, T. T. 共1998兲. ‘‘Head-related transfer functions of the barn owl: Measurement and neural responses,’’ Hear. Res. 118, 35–46. May, B. J., and Huang, A. Y. 共1996兲. ‘‘Sound orientation behavior in cats. I. Localization of broadband noise,’’ J. Acoust. Soc. Am. 100, 1059–1069. Mehrgardt, S., and Mellert, V. 共1977兲. ‘‘Transformation characteristics of the external human ear,’’ J. Acoust. Soc. Am. 61, 1567–1576. Middlebrooks, J. C. 共1987兲. ‘‘Binaural mechanisms of spatial tuning in the cat’s superior colliculus,’’ J. Neurophysiol. 57, 688–701. Middlebrooks, J. C. 共1999a兲. ‘‘Individual differences in external-ear transfer functions reduced by scaling in frequency,’’ J. Acoust. Soc. Am. 106, 1480–1492. Middlebrooks, J. C. 共1999b兲. ‘‘Virtual localization improved by scaling nonindividualized external-ear transfer functions in frequency,’’ J. Acoust. Soc. Am. 106, 1493–1510. Middlebrooks, J. C., and Green, D. M. 共1990兲. ‘‘Directional dependence of interaural envelope delays,’’ J. Acoust. Soc. Am. 87, 2149–2162. Middlebrooks, J. C., and Knudsen, E. I. 共1987兲. ‘‘Changes in external ear position modify the spatial tuning of auditory units in the cat’s superior colliculus,’’ J. Neurophysiol. 57, 672–687. Middlebrooks, J. C., and Knudsen, E. I. 共1984兲. ‘‘A neural code for auditory space in the cat’s superior colliculus,’’ J. Neurosci. 4, 2621–2634. Middlebrooks, J. C., Makous, J. C., and Green, D. M. 共1989兲. ‘‘Directional sensitivity of sound-pressure levels in the human ear canal,’’ J. Acoust. Soc. Am. 86, 89–108. Middlebrooks, J. C., and Pettigrew, J. D. 共1981兲. ‘‘Functional classes of neurons in primary auditory cortex of the cat distinguished by sensitivity to sound location,’’ J. Neurosci. 1, 107–120. Middlebrooks, J. C., Xu, L., Eddins, A. C., and Green, D. M. 共1998兲. ‘‘Codes for sound-source location in non-tonotopic auditory cortex,’’ J. Neurophysiol. 80, 863–881. Morimoto, M., and Ando, Y. 共1980兲. ‘‘On the simulation of sound localization,’’ J. Acoust. Soc. Jpn. 共E兲 1, 167–174. 1459 J. Acoust. Soc. Am., Vol. 107, No. 3, March 2000 Musicant, A. D., Chan, J. C. K., and Hind, J. E. 共1990兲. ‘‘Directiondependent spectral properties of cat external ear: New data and crossspecies comparisons,’’ J. Acoust. Soc. Am. 87, 757–781. Nelken, I., Bar Yosef, O., and Young, E. D. 共1997兲. ‘‘Response of Field AES neurons to virtual-space stimuli,’’ in Psychophysical and Physiological Advances in Hearing, edited by A. R. Palmer, A. Rees, A. Q. Summerfield, and R. Meddis 共Whurr, London兲, pp. 504–512. Neti, C., Young, E. D., and Schneider, M. H. 共1992兲. ‘‘Neural network models of sound localization based on directional filtering by the pinna,’’ J. Acoust. Soc. Am. 92, 3140–3156. Populin, L. C., and Yin, T. C. 共1998兲. ‘‘Behavioral studies of sound localization in cat,’’ J. Neurosci. 18, 2147–2160. Rajan, R., Aitkin, L. M., Irvine, D. R. F., and McKay, J. 共1990兲. ‘‘Azimuthal sensitivity of neurons in primary auditory cortex of cats. I. Types of sensitivity and the effects of variations in stimulus parameters,’’ J. Neurophysiol. 64, 872–887. Reale, R. A., Chen, J., Hind, J. E., and Brugge, J. F. 共1996兲. ‘‘An implementation of virtual acoustic space for neurophysiological studies of directional hearing,’’ in Virtual Auditory Space, edited by S. Carlile 共Springer-Verlag, Heidelberg兲, pp. 153–183. Rice, J. J., May, B. J., Spirou, G. A., and Young, E. D. 共1992兲. ‘‘Pinnabased spectral cues for sound localization in cat,’’ Hear. Res. 58, 132– 152. Rose, J. E., Gross, N. B., Geisler, C. D., and Hind, J. E. 共1966兲. ‘‘Some neural mechanism in the inferior colliculus of the cat which may be relevant to localization of a sound source,’’ J. Neurophysiol. 29, 288–314. Shaw, E. A. G. 共1966兲. ‘‘Earcanal pressure generated by a free sound field,’’ J. Acoust. Soc. Am. 39, 465–470. Wallace, M. T., Meredith, M. A., and Stein, B. E. 共1998兲. ‘‘Multisensory integration in the superior colliculus of the alert cat,’’ J. Neurophysiol. 80, 1006–1010. Wenzel, E. M., Arruda, M., Kistler, D. J., and Wightman, F. L. 共1993兲. ‘‘Localization using nonindividualized head-related transfer functions,’’ J. Acoust. Soc. Am. 94, 111–123. Wiener, F. M., Pfeifer, R. R., and Backus, A. S. N. 共1966兲. ‘‘On the sound pressure transformation by the head and auditory meatus of the cat,’’ Acta Oto-Laryngol. 61, 255–269. Wightman, F. L., and Kistler, D. J. 共1989a兲. ‘‘Headphone simulation of free field listening. I. Stimulus synthesis,’’ J. Acoust. Soc. Am. 85, 858–867. Wightman, F. L., and Kistler, D. J. 共1989b兲. ‘‘Headphone simulation of free-field listening. II: Psychophysical validation,’’ J. Acoust. Soc. Am. 85, 868–878. Xu, L., Furukawa, S., and Middlebrooks, J. C. 共1998兲. ‘‘Sensitivity to sound-source elevation in non-tonotopic auditory cortex,’’ J. Neurophysiol. 80, 882–894. Xu, L., Furukawa, S., and Middlebrooks, J. C. 共1999兲. ‘‘Auditory cortical responses in the cat to sounds that produce spatial illusions,’’ Nature 共London兲 399, 688–691. Young, E. D., Rice, J. J., and Tong, S. C. 共1996兲. ‘‘Effects of pinna position on head-related transfer functions in the cat,’’ J. Acoust. Soc. Am. 99, 3064–3076. Zhou, B., Green, D. M., and Middlebrooks, J. C. 共1992兲. ‘‘Characterization of external ear impulse responses using Golay codes,’’ J. Acoust. Soc. Am. 92, 1169–1171. L. Xu and J. C. Middlebrooks: Individual differences in cat DTFs 1459