THE JOURNAL OF COMPARATIVE NEUROLOGY 381:188202 (1997)

Intracellular Labeling of Auditory Nerve

Fibers in Guinea Pig: Central
and Peripheral Projections
J. TSUJI1,2,3 AND M.C. LIBERMAN1,2,4*
1Eaton-Peabody Laboratory, Massachusetts Eye and Ear Infirmary,
Boston, Massachusetts 02114
2Department of Otology and Laryngology, Harvard Medical School,

Boston, Massachusetts 02115

3Department of Otolaryngology, Kyoto University, Kyoto 606, Japan
4Division of Health Sciences and Technology, Massachusetts Institute of Technology,

Cambridge, Massachusetts 02139

ABSTRACT
Auditory-nerve fibers (ANFs) in the cat have been subdivided according to spontaneous
rate (SR), with high-SR fibers showing the lowest thresholds. Cochlear terminals of the three
SR groups differ in caliber and synaptic position around the inner hair cell (Liberman [1982b]
Science 216:12391241); central terminals differ in degree of branching and in which
subregions of the cochlear nucleus (CN) are targeted (Liberman [1991] J. Comp. Neurol.
313:240258). The present study investigates whether these SR-based differences in ANF
connections are unique to the cat. Thirty ANFs from 15 guinea pigs were intracellularly
labeled after measuring characteristic frequency, threshold, and SR. Labeled cochlear
projections showed significant SR-based differences in axonal caliber, with low- and me-
dium-SR fibers 2040% thinner than those of high-SR fibers for both peripheral and central
(modiolar) axons. Spatial segregation in the inner hair cell area could not be assessed;
however, the peripheral axons in the osseous spiral lamina showed the same SR-based
organization reported for the cat (Kawase and Liberman [1992] J. Comp. Neurol. 319:312
318). Labeled central projections also showed significant SR-based differences. Low- and
medium-SR fibers: 1) were more highly branched, 2) sent significantly more terminals to the
small-cell cap region of the CN, and 3) produced endbulb terminals (on spherical cells) that
were significantly more complex than high-SR fibers. All of these SR-based trends for both
central and peripheral projections are analogous to those reported in the cat, and, thus, may
represent a fundamental organizational principle of the mammalian ear. J. Comp. Neurol.
381:188202, 1997. r 1997 Wiley-Liss, Inc.

Indexing terms: cochlea; cochlear nucleus; spontaneous rate; sensory systems

The afferent innervation of the inner hair cells (IHCs) in
the mammalian cochlea constitutes 9095% of the primary
sensory fibers of the auditory nerve (Spoendlin, 1969;
Kiang et al., 1984). Each of these radial fibers (RFs) makes
synaptic contact peripherally with a small number of
IHCs, usually one or two (Liberman, 1980; Liberman et al.,
1990), and sends central projections via ascending and
descending branches to the anteroventral (AV), posteroven-
tral (PV), and dorsal (DCN) divisions of the cochlear
nucleus (CN; Osen, 1970; Lorente de No, 1981; Fekete et
al., 1982).
The physiology of RFs has been studied extensively over
the last 30 years in a number of mammalian species,
including the cat (Kiang et al., 1965; Liberman, 1978),
chinchilla (Salvi et al., 1982), gerbil (Schmiedt, 1989),
guinea pig (Winter et al., 1990), rabbit (Borg et al., 1988),
and others. In all of these species, there is a close relation-
ship between threshold sensitivity and spontaneous dis-
charge rate (SR): Among the fibers of similar characteristic
frequency (CF), those with the lowest thresholds to sound
Contract grant sponsor: National Institute on Deafness and Other
Communicative Disorders; Contract grant number: RO1 DC00188.
*Correspondence to: M. Charles Liberman, Eaton-Peabody Laboratory,
Massachusetts Eye and Ear Infirmary, 243 Charles Street, Boston, MA
02114. E-mail: mcl@epl.meei.harvard.edu
Received 23 September 1996; Revised 5 December 1996; Accepted 24
December 1996

r 1997 WILEY-LISS, INC.

AUDITORY NERVE LABELING IN GUINEA PIG
are those with the highest SRs, and vice versa. This
relationship has been most extensively documented in the
cat (Liberman, 1978): If the population of cat RFs is
divided into three SR groups (low-SR fibers with SR , 0.5
spikes/second, medium-SR fibers with 0.5 # SR , 17.5
spikes/second, and high-SR fibers with SR $ 17.5 spikes/
second), then thresholds within each SR group are rela-
tively homogenous, and each groups thresholds are signifi-
cantly different from the other two.
Intracellular labeling studies in the cat have shown that
fibers of the three SR groups differ morphologically in a
number of ways with respect to both peripheral and
central projections (Fekete et al., 1982; Liberman, 1982b,
1991, 1993; Liberman and Oliver, 1984). Studies of the
peripheral connections have shown that SR is correlated
with 1) the axonal and dendritic caliber (high-SR fibers are
thickest) and 2) the location of their synapses with IHCs
(high-SR fibers tend to be found on the side closer to the
pillar cells, whereas low-and medium-SR fibers are found
exclusively on the side closer to the modiolus; Liberman,
1982b). This spatial segregation of the different SR groups
is also seen in the osseous spiral lamina and spiral
ganglion, where low- and medium-SR fibers tend to travel
on the scala vestibuli side of the nerve bundle (Kawase and
Liberman, 1992).
The central projections of RFs from the three SR groups
also differ. In cat, the ascending branches of low- and
medium-SR groups are significantly more highly branched
than those of high-SR fibers (Fekete et al., 1982; Liber-
man, 1991, 1993). Much of this profuse branching is
directed toward the small-cell regions, which form a rind
around the magnocellular portions of both the AVCN and
the PVCN (Osen, 1969). Indeed, the small-cell cap receives
RF innervation almost exclusively from low- and me-
dium-SR fibers, with an especially rich projection from
those fibers with the lowest SRs and the highest thresh-
olds (Liberman, 1991). Although high-, medium-, and
low-SR fibers all project to the magnocellular regions of the
VCN, certain cell-types within these regions may receive
input preferentially from some subset of SR groups. Within
the magnocellular portions of the VCN, there may also be a
spatial segregation of projections from the different SR
groups, i.e., within each isofrequency lamina, cells in one
region may receive more inputs from high-SR fibers,
whereas cells in another region of the lamina may be
preferentially innervated by low- and/or medium-SR fibers
(Leake and Snyder, 1989; Liberman, 1991).
The clear-cut morphological distinctions, both peripher-
ally and centrally, among fibers from the three SR groups
in the cat suggest that the SR-based classification of RFs
has a functional significance. The observation that certain
regions and/or cell types within the CN are preferentially
innervated by different SR groups suggests that the groups
may play different roles in the central processing of
auditory information.
Although physiological data on SR and other aspects of
RF response have been analyzed in a number of mamma-
lian species, the kinds of structure-function relationships
revealed by intracellular labeling have been described only
in the cat. The purpose of the present study was to
examine the central and peripheral projections of labeled
RFs of differing SR in a second mammalian species. The
guinea pig was chosen, because it is commonly used in
auditory experimentation. If the same relationships hold
in guinea pig as in the cat, then this strengthens the notion
189
that the cat data represent a generalized mammalian
plan. If that is true, then this raises the likelihood that the
same relationships hold for the human auditory pathways.
MATERIALS AND METHODS
Albino guinea pigs of either sex weighing between 200
and 500 g were used in these experiments. Animals were
anesthetized initially with sodium pentobarbital (15 mg/kg
i.p.) and Innovar-Vet (0.25 ml/kg i.m.). Booster injections
(one-third of the original dose) of each drug were delivered
every 6 and 2 hours, respectively. Animals were tracheosto-
mized, paralyzed with tubocurarine (1.2 mg/kg i.m. every 4
hours), and artificially respired with room air. Rectal
temperature was maintained between 37.5C and 38.5C
by varying the temperature of the ambient air in the
sound-proof chamber. Heart rate was continuously moni-
tored. All experimental methods were approved by the
IACUC of the Massachusetts Eye and Ear Infirmary.
Surgical preparation
After surgical levels of anesthesia were reached, the
animal was prepared for single-fiber recording from the
auditory nerve via a posterior-fossa approach. The skin
and muscle layers over the back of the skull were removed
or reflected to expose the bullas and the occipital bone, and
the cartilaginous ear canal was severed near the skull to
allow sealing of a closed acoustic system near the tympanic
membrane. The posterolateral portion of the bulla was
opened with a scalpel blade to allow placement near the
round window of a fine silver wire, which was used to
record the cochlear compound action potential (CAP). A
circular portion of occipital bone was removed with ron-
geurs, and the dura was opened to allow lateral retraction
of the cerebellum. The cerebellar flocculus was removed by
gentle aspiration, and the auditory nerve bundle was
exposed by placement of cotton balls between the medial
surface of the temporal bone and the lateral surface of the
brainstem.
Sound delivery, acoustic stimuli,
and measurement paradigms
Acoustic stimuli were delivered through a closed, cali-
brated acoustic system (Kiang et al., 1965) consisting of
1-inch Bruel and Kjaer condenser microphone used as a
sound source and a one-quarter-inch Bruel and Kjaer
condenser microphone used to monitor sound pressure
near the eardrum through a probe tube. CAPs were evoked
via 5-msec tone pips (0.5 msec rise-fall with cos2 shaping)
delivered at a rate of 10/second. CAP thresholds were
measured often during the course of the experiment by a
computer-driven algorithm, which tracks the sound pres-
sure level (SPL) required to produce a 10 V CAP as a
function of frequency. The CAP at each frequency/level
combination was measured by amplifying the response
(310,000) from a silver wire near the round window
(referred to the tongue) with a pass band from 300 Hz to
3,000 Hz. At each frequency/level combination, the aver-
age response to 16 tone pips was computed, with micro-
phonic potentials eliminated by reversing the polarity of
the stimulus for half of the presentations. Single-fiber
tuning curves were measured with a computer-driven
algorithm (Liberman, 1978), which tracks an isorate curve
corresponding to a 10 spikes/second increase with respect
to background discharge, by using 50-msec tone bursts (2.5
190
msec rise-fall) delivered at 10/second. SRs were estimated
from 10-second samples.
Intracellular recording and labeling
Single-fiber activity was measured with glass micropi-
pettes beveled to final resistances of 4060 megohms. Two
different neuronal tracers were used: horseradish peroxi-
dase (HRP: Sigma type VI) or biocytin (Sigma). HRP was
used at a concentration of 48% in 0.05 M Tris buffer and
0.15 M KCl; biocytin was used at a concentration of 2% in
0.05 M Tris buffer and 0.45 M KCl. Following measure-
ment of tuning curve and spontaneous activity, selected
fibers were labeled iontophoretically with 50-msec current
pulses (electrode positive) delivered 10/second. Iontophore-
sis was not attempted unless the membrane potential was
more negative than 20 mV. Labeling was usually success-
ful if 4 nA of current were passed for at least 23 minutes
(Liberman and Oliver, 1984); however, HRP labeling was
most successful for peripheral projections, whereas biocy-
tin was preferred for the central projections.
Fixation and sectioning
The anesthetized and curarized animal was maintained
on the respirator for 812 hours after the last injection to
allow for neuronal transport of tracer. After this survival
time, the cochlea and brainstem were fixed by intravascu-
lar perfusion of physiological saline followed by a solution
of glutaraldehyde and paraformaldehyde in 0.065 M phos-
phate buffer, pH. 7.3. For tissue in which HRP was the
tracer, 1.5% glutaraldehyde and 1% paraformaldehyde
were used. For tissue with biocytin injections, 0.5% glutar-
aldehyde and 4% paraformaldehyde was used. After fixa-
tion, the cochlea was removed, and cold fixative was
flushed through the oval and round windows. Following 12
hours of postfixation in the same solution at 4C, the
cochlear bone was thinned with dental drills, and the
remaining bone was decalcified by immersion in EDTA for
3 days at 4C. After complete decalcification, the cochlea
was infused with a gelatin-albumin mixture, which was
hardened with glutaraldehyde, and then sectioned on a
Vibratome at 80 m in the horizontal plane. The brainstem
was also postfixed for 12 hours in the same fixative at 4C,
partially cryoprotected by immersion in 20% sucrose (phos-
phate buffered) for 12 hours, then frozen with liquid nitro-
gen, and thawed prior to embedding in gelatin-albumin
and cutting at 80 m in a transverse plane. This brief
freezing step improved the darkness of the ultimate reac-
tion product, presumably by disrupting cell membranes to
improve penetration of reagents.
Histochemical reactions and counterstains
Tissue in which HRP was the neuronal tracer was
processed as described previously (Liberman and Oliver,
1984). In brief, the sections were reacted before they were
mounted on slides by 3060 minute incubation in a
solution of cobalt chloride followed by 6090 minute incu-
bation in a solution containing diaminobenzidine (Adams,
1977). Tissue in which biocytin was the neuronal tracer
was processed by using an ABC kit (Vector Laboratories).
First, endogenous peroxidases were inactivated by a 1-hour
soak in hydrogen peroxide and methanol. Then, the tissue
was incubated overnight in the avidin-biotin-HRP complex
from the ABC kit. Finally, the sections were reacted by
soaking in a phosphate-buffered solution of diaminobenzi-
dine, cobalt chloride, nickel-ammonium sulfate, and di-
J. TSUJI AND M.C. LIBERMAN
methyl sulfoxide. After mounting on slides, CN tissue was
counterstained with thionin and azure. Cochlear tissue
was not counterstained.
Cochlear reconstructions
Cochleas were graphically reconstructed (Schuknecht,
1953). The peripheral projections, in general, were more
lightly labeled than the central projections. Peripherally,
some fibers could not be followed past the spiral ganglion.
Most fibers could be followed to the habenula perforata,
just beneath the organ of Corti; however, relatively few
fibers could be followed to their terminal swellings on
IHCs. Only those fibers that could be followed at least to
the habenula were used in the construction of a cochlear
frequency map. Measurements of fiber diameter and cell
body morphology were done with a drawing tube at a
magnification of 32,150. The drawings were digitized via
computerized planimetry.
CN reconstructions
Labeled fibers were reconstructed initially with a draw-
ing tube at a magnification of 385. The low-power draw-
ings were superimposed (four sections to the page), and
CN subdivisions were determined for the second section of
each four-section series. CN subdivisions were based on
descriptions in the literature of the guinea pig CN (Hack-
ney and Pick, 1986; Hackney et al., 1990). After the
low-power reconstructions were complete, each fiber was
reanalyzed with high-power Nomarski optics: The num-
bers of terminals were counted, both terminal and en
passant. Those terminals clearly contacting the somata of
CN cells were identified. Each of these somatic terminals
was traced with a drawing tube (at 32,150), and the
tracings were digitized by computerized planimetry. So-
matic endings tended to be larger and more complex than
nonsomatic endings. For the terminal counts, a given
somatic contact was counted only once, regardless of how
many small swellings contributed to that particular con-
tact on that one cell. For nonsomatic endings, each en
passant and terminal swelling was counted individually.
RESULTS
Relation between CF and cochlear location
In the guinea pig, there are many measures of the
relation between cochlear location and CF for the lower
basal turn (CFs . 10 kHz) and a few frequency-location
values for apical turns (Fig. 1B); however, no explicit
cochlear frequency map has yet been published. Figure 1A
shows that the present study provides data that can be
used to construct a complete cochlear map for the guinea
pig. Although we had difficulty in preserving the unmyelin-
ated terminals of the labeled RFs within the organ of Corti,
a total of 22 fibers could be traced unambiguously to the
habenula perforata, allowing an accurate determination of
the cochlear locus of the peripheral terminal to within a
few tens of microns. The 22 map points in Figure 1A are
well fit by a linear-distance to log-frequency relationship:
the best-fit straight line on this semilog plot has a correla-
tion coefficient of 0.995. The correlation is lower (R 5
0.986) when the data are plotted vs. absolute cochlear
location (see Fig. 1B). Careful inspection of the distribu-
tion of data points around the line in Figure 1A hint of a
slight decrease in slope of the map function as CF de-
AUDITORY NERVE LABELING IN GUINEA PIG
Fig. 1. Cochlear frequency maps for the guinea pig. A: Cochlear
frequency map derived from intracellular labeling in the present study
based on 22 labeled fibers. The best-fit straight line to the data is also
shown along with the equation representing that logarithmic relation-
ship between percent distance (%d) from base and characteristic
frequency (CF; in kHz). B: Comparison of the cochlear frequency map
from the present study (solid circles) with data derived from a number
of other studies in the literature, including measurements of basilar
creases, as is clearly seen in HRP-based cochlear maps in
other species (see, e.g., Liberman, 1982a); however, the
high correlation coefficient for the straight-line fit to these
data argue against the search for a better fit with a more
complex function of the type described by Greenwood
(1996).
Relation between SR and threshold
Auditory-nerve fibers (ANFs) in the guinea pig (Winter
et al., 1990; Yates, 1991) show the same general relation-
ship between SR and threshold reported in the cat (Liber-
man, 1978), although, in the guinea pig, the evidence for
division into three SR groups is not as compelling. This
discrepancy may arise from differences in data acquisition
and analysis techniques; thus, in each of two guinea pigs in
the present study, an effort was made to record from large
enough samples of nerve fibers (83 in one animal and 151
in the other) that the relation between threshold and SR
could be evaluated in a manner similar to that used
previously for the cat. The results of the analysis from
these two ears are shown in Figures 2 and 3.
The distribution of SRs seen in the guinea pig appears to
be fundamentally bimodal (Fig. 2B), as is the case for the
cat, although the separation between the low-rate and
high-rate peaks is not quite so clear cut. As is the case for
the cat, there is no obvious difference in the SR distribu-
tions in different CF regions (Fig. 2A).
The relationship between SR and relative threshold
extracted from these two animals is shown in Figure
3AD. In Figure 3A,B, the relationship is shown with a
linear SR axis, as originally published for the cat; in Figure
3C, a logarithmic SR axis is used. Although the relation-
ships shown in Figure 3AC does not provide compelling
evidence for an SR-based division of RFs into three dis-
191
membrane (BM) motion, intracellular recordings from inner hair cells
(IHCs; Russell and Sellick, 1978), differential recordings of cochlear
microphonic (CM; Dallos, 1973; Schmiedt and Zwislocki, 1977), single
cell recordings from the spiral ganglion cells (SGC; Robertson et al.,
1980), and a study using horseradish peroxidase (HRP) labeling of
spiral ganglion cells (Robertson, 1984). Note that the vertical scale in
B is the absolute distance from the base, whereas that in A is
normalized to the total length.
crete classes, it is nonetheless clear that dividing the SR
continuum into three groups has practical value. If the
groups are defined by boundaries at 1 and 20 spikes per
second, as shown in the figures, then: 1) each of the
resultant SR groups has relatively homogeneous relative
thresholds (65 dB range for high- and medium-SR groups
and 610 dB for low-SR fibers); 2) within each SR group,
there is little relation between SR and threshold; and 3)
there is a highly significant difference in the mean relative
threshold between any two groups (Fig. 3D). Based on this
analysis, high-, medium-, and low-SR groups constitute
73%, 15%, and 12%, respectively, of the fibers in the guinea
pig auditory nerve.
Relation between SR and peripheral
morphology
Neuronal caliber. Measurements of axonal caliber
were made at a number of locations on each darkly labeled
fiber. Figure 4 shows that in none of the regions analyzed
was there a clear relation between CF and fiber caliber.
However, low- and medium-SR fibers were clearly smaller
in diameter than high-SR fibers for both peripheral (Fig.
4A) and central axons (Fig. 4B), as measured in the
immediate vicinity of the cell body within the spiral
ganglion. Furthermore, the peripheral axons of fibers from
all SR groups were smaller in caliber than the central
axons, as has also been reported for the cat (Liberman and
Oliver, 1984).
On the other hand, there do not appear to be any
SR-based differences in the size or shape of the cell body in
the spiral ganglion. Figure 5 shows that the silhouette
areas of high-, medium-, and low-SR somata appear to
fully overlap (Fig. 5A), as do their shapes (Fig. 5B). The
shape measure used in the present study was a coefficient
192
Fig. 2. A,B: Distributions of spontaneous discharge rates seen in 234 units recorded from two guinea
pigs. Spontaneous discharge rates are based on a 10-second sample. sp/sec, spikes/second.
of circularity, for which a value of 1.0 corresponds to a
perfect circle, whereas lower values are increasingly fusi-
form.
In the present material, we were unable to routinely
preserve the unmyelinated terminals within the organ of
Corti. Similar technical problems were encountered when
material from the cat was processed as serial sections:
Previous measurements of terminal caliber were made on
cochlear material processed as whole-mounts and viewed
as plastic-embedded surface preparation. Thus, no mea-
surements of peripheral terminal caliber are available in
this study.
Spatial location. In cat, fibers from different SR groups
are spatially segregated around the IHCs, and this segre-
gation is reflected in an SR-based spatial organization in
the osseous spiral lamina and the spiral ganglion (Kawase
and Liberman, 1992). The limited recovery of labeled
peripheral terminals in the present study precluded an
analysis of the spatial organization of the three SR groups
around the IHCs; however, an analysis of the organization
within the osseous spiral lamina and spiral ganglion was
carried out. Figure 6A shows that there is no sign of spatial
organization of the three SR groups within the spiral
ganglion. However, within the osseous spiral lamina (Fig.
6B), in a region near the habenula perforata, all four
low-SR fibers were found in the upper half of the lamina
(i.e., near the scala vestibuli side), and all of the fibers in
the lowest 30% of the lamina were from high-SR fibers.
Relation between SR and central projections
Terminal counts. In the cat CN, the ascending
branches of low- and medium-SR fibers are more highly
branched than high-SR fibers (Ryugo and Fekete, 1982),
and the difference is due to a profuse innervation of the
small-cell cap by low- and medium-SR fibers (Liberman,
1991). In the present study, only the ascending branches of
each labeled fiber were evaluated. For each fiber, all of the
terminal or en passant swellings were evaluated with
high-power Nomarski optics, and those terminals clearly
contacting a CN cell were identified as somatic. We
J. TSUJI AND M.C. LIBERMAN
presume that the nonsomatic terminals are contacting
the dendrites of CN cells or other neural structures that
are not visible in a light-microscopic examination of this
Nissl-stained material.
Figure 7A,B shows that there are many more nonso-
matic than somatic terminals for each fiber. These data
also show that the ascending branches of low- and me-
dium-SR fibers in guinea pig are significantly more highly
branched than those of high-SR fibers; however, the differ-
ences arise only in the numbers of nonsomatic terminals
(Fig. 7B). With respect to these nonsomatic terminals, the
counts for low- and medium-SR fibers are roughly three
times greater than those for high-SR fibers, and there is
almost no overlap. There also appears to be a trend toward
more terminals as CF increases.
The regions of the CN innervated by the ascending
branches consist of a large central core of large cells and a
thin rind or cap of small cells (Osen, 1969; Hackney et
al., 1990). This rind of small cells has also been termed the
juxtagranular region (Liberman, 1991), because it lies
just beneath the layer of granule cells covering much of the
VCN. One example of an ANF projection to the juxtagranu-
lar small-cell region is shown in the photomicrograph in
Figure 8. The labeled terminals of this medium-SR fiber
end in the small-cell cap, within a few microns of the
granule cell area. In the present study, there was a profuse
and selective innervation of the small-cell regions by low-
and medium-SR fibers. In Figure 7C,D, all of the terminals
of each ANF (both somatic and nonsomatic) are divided
into those innervating the magnocellular core regions (Fig.
7C) vs. the small-cell cap regions (Fig. 7D). Only one of the
high-SR fibers in the present sample sent any branches
into the small-cell cap region, and only one of the low- or
medium-SR fibers did not. There is evidence that the most
profuse projections to the magnocellular core regions from
single fibers also arise from low- and medium-SR fibers
(Fig. 7C); however, not all low- and medium-SR fibers
branch profusely in these regions.
Terminal morphology. The central terminals of ANFs
show a wide range of morphologies, from small, simple,
AUDITORY NERVE LABELING IN GUINEA PIG
Fig. 3. AD: The relationship between spontaneous rate (SR) and
relative threshold in 234 units from two guinea pigs. These two
animals were selected because 1) a large number of fibers was sampled
in each (82 in one and 151 in the other), and 2) both were particularly
stable, showing cochlear compound action potential (CAP) thresholds
within 5 dB at all frequencies from the first measure to the last
measure. The same data are plotted in three different ways in AC: in
each, relative threshold is defined as the unit threshold minus the
average thresholds for high-SR fibers (SR . 20 spikes/second) at the
same CF region in the same ear. In A and B, the SR is plotted on a
linear scale; in C, a logarithmic scale is used. In AC, the SR values of
spherical swellings to the large and elaborate, clawlike
endbulbs of Held on spherical cells (Ryugo and Fekete,
1982; Sento and Ryugo, 1989). In the present study, we
traced and measured the silhouette areas of all somatic
terminals and the CN cells they contacted. No systematic
SR-based differences were found in the sizes of terminals
or in the sizes of the target CN cells (not illustrated).
In previous studies (Rouiller et al., 1986; Liberman,
1991), it has been shown that the terminals on spherical
193
1 spike/second and 20 spikes/second are indicated by vertical dotted
lines. In D, the means and standard errors for the relative thresholds
in each of the three SR groups (defined in C) are shown. The relative
threshold is computed with respect to a normalization curve for each
animal derived by 1) averaging the thresholds for all high-SR units in
each of 12 logarithmically spaced CF bins, 2) setting the frequency of
each bin to the geometric mean of its boundaries, and 3) interpolating
normalization values for all frequencies between these 12 mean
values. The interpolation process introduces random variations, such
that the ensemble average for high-SR fibers shown in D is nonzero.
cells tended to be the largest somatic terminals in the CN.
A number of these large ANF terminals are illustrated in
the photomicrographs in Figure 9: The CN cell contacted
by the terminal is clearest in Figure 9A.1 Figure 10A shows
1The CN cell body is best seen in Nomarski optics. However, these
micrographs were obtained with brightfield optics to maximize the depth of
field, thereby capturing more of the terminal in the image.
194 J. TSUJI AND M.C. LIBERMAN

Fig. 4. Average diameter of auditory nerve fibers of the three SR
groups displayed as function of their CFs. The symbol key in A applies
to both A and B. The cochlear peripheral (A) and central axons (B)
were measured near the cell body. For each measurement, a 100-m
length of labeled fiber was traced, and the total area was divided by
the path length. Nodal regions, where diameter is locally constricted
(Liberman and Oliver, 1984), were not included. The data base for
central axons was larger than that for peripheral axons, because
several fibers could not be traced peripheral to the cell body.

Fig. 5. Size (A) and shape (B) of the spiral ganglion cell somata of labeled fibers from the three SR
groups displayed as function of their CF. The soma roundness is a coefficient of circularity, i.e., the ratio
between the measured area and the area of perfect circle with the same perimeter. The symbol
conventions are the same as shown in Figure 4A.

that spherical cell contacts were always at least 20 m2 in
area. Although most somatic contacts on nonspherical cells
were significantly smaller, a few were as large as moderate-
sized spherical cell terminals. If a class of large endbulbs
is arbitrarily defined as any terminal greater in silhouette
area than 20 m2, as was done for Figure 10B, then each
fiber has at least one endbulb; however, some have as
many as five. There do not appear to be any strong
SR-based differences in the numbers of endbulbs per fiber.
However, the probability of one fiber giving rise to multiple
large endbulbs appears to be greater for high-CF fibers
than for low-CF fibers.
Previous study of the CN terminals of cat ANFs has
suggested that high-SR endbulbs tend to be less complex
in shape than those from low- and medium-SR fibers
(Sento and Ryugo, 1989). Such SR-based differences in
AUDITORY NERVE LABELING IN GUINEA PIG
Fig. 6. The relative positions of fibers from the three SR groups
within the spiral ganglion (panel A) and in the osseous spiral lamina
near the habenula perforata (panel B). The schematic cross section
through the cochlear duct in (panel C) illustrates how relative position
is defined. For ganglion position (A), the height (H) of the ganglion is
measured along an axis perpendicular to the osseous spiral lamina,
and the relative position of the labeled cell (X) is measured along the
same axis. Normalized position is simply X/H. For the osseous spiral
complexity can also be seen in guinea pig endbulbs in the
photomicrographs in Figure 9: The endings of low- and
medium-SR fibers shown in Figure 9A,C are more fenes-
trated and have a more lacy appearance compared with
those of the high-SR fibers, as shown in Figure 9B,D. In an
attempt to quantify this morphological difference, a form
factor was defined (Sento and Ryugo, 1989) as the ratio of
the area to the perimeter. Endbulbs that are highly
complex have a larger perimeter and, thus, a lower form
factor. The camera lucida tracings and the scatterplot of
form factors in Figure 11 indicate that endbulbs from
guinea pig ANFs show the same relationship as those in
the cat: Low- and medium-SR fibers tend to produce
195
lamina (panel B), the height of the lamina (h) and the position of the
fiber (x) are measured along the same axis, and normalized position is
defined as x/h. For the data in panel B, the measurements were always
made within 100 m of the habenula. There are fewer points in panel
B, because not all of the labeled fibers could be traced with confidence
peripheral to the ganglion. For SR symbol conventions, see Figure 4A.
SV, scala vestibuli; ST, scala tympani.
endbulbs with lower form factors than those from high-SR
fibers.
DISCUSSION
Cochlear frequency maps
Cochlear frequency maps are now available for a num-
ber of mammalian species, including several of those most
commonly used for auditory research, i.e., cat, gerbil,
mustache and horseshoe bats, and chinchilla, as well as for
a number of other species, including rat, mole rat, and
opossum (for recent reviews, see Greenwood, 1996; Muller,
196
Fig. 7. AD: Numbers of cochlear nucleus (CN) terminals per fiber
for labeled fibers from the three SR groups. Only data from completely
reconstructed fibers are shown, and only terminals from the ascending
branch of each fiber were counted. For each fiber shown, the total
number of terminals is the sum of the values shown in A and B:
Terminals are divided according to whether a clear contact with a CN
1996). A number of different techniques have been used to
construct such maps, including single-fiber labeling (Liber-
man, 1982a), multifiber labeling (Muller, 1991), and the
correlation of structural and functional change in pathologi-
cal ears (Eldredge et al., 1981).
A wealth of data exists on the relation between cochlear
location and best frequency for the basalmost one- tent with all available data on the cochlear frequency map
quarter of the guinea pig cochlea, and there is very good
agreement across studies, as summarized in Figure 1B.
These data derive from studies of 1) basilar membrane
motion (Kohloffel, 1972; Wilson and Johnstone, 1975;
Eldredge et al., 1981), 2) intracellular responses of IHCs
(Russell and Sellick, 1978), and 3) recordings from spiral
ganglion cells (Robertson et al., 1980; Robertson, 1984).
J. TSUJI AND M.C. LIBERMAN
cell was visible (somatic terminal) or not (nonsomatic terminal). For C
and D, both somatic and nonsomatic terminals were counted; thus, for
each fiber, the sum of the values plotted in C and D is the same as the
sum of the values in A and B. For SR symbol conventions, see Figure
4A. For C and D, magnocellular areas include all regions of the CN not
classified as small-cell cap.
Few of these studies provide total cochlear lengths; thus,
cochlear location is expressed as absolute distance from
the basalmost end. In the apical three-quarters of the
cochlea, the only map points available outside the pres-
ent study are from differential recordings of cochlear
microphonics. The data from the present study are consis-
for the guinea pig.
Previous single-fiber labeling studies in the cat have
shown that the cross-animal scatter in the map is reduced
if the cochlear length is normalized. The HRP map data in
Figure 1A are remarkably well fit by a simple logarithmic
relation between normalized position and CF. In this
respect, it differs from the maps in many other species,
AUDITORY NERVE LABELING IN GUINEA PIG
Fig. 8. Photomicrograph showing a cluster of endings (within
circle) from a biocytin-labeled, medium-SR fiber (CF 5 15.6, SR 5
10.1) terminating in the small-cell cap, very close to the granule cell
area. The approximate border of the granule cell area runs along the
dotted line. This image (and the images in Fig. 9) was obtained
digitally via Image 1 software (Universal Imaging) and was adjusted
after acquisition to maximize contrast and resolution.
which show a prominent curvature on a semilog plot: For
cochlear locations approaching the apex, the curve slope
approaches 0 as each millimeter of organ of Corti serves a
larger and larger fraction of an octave. The mathematical
description of such map functions, which were first sug-
gested by Greenwood (1961), is
f 5 A * (10ax 2 k)
[f in Hz; x in normalized distance from apex (0 to1)]
For the present guinea pig data, the best-fit straight line
to the points in Figure 1A gives A 5 132, a 5 2.63, and k 5
0. The suggestion that the constant k is 0 in the guinea pig
reflects the fact that a simple logarithmic relation between
cochlear location and CF produced such a good fit (r 5
.997). Thus, although there is a slight hint in Figure 1A
that the map slope at low frequencies becomes less nega-
tive, the search for low k value that might produce a
better fit was unjustifiable given the paucity of apical
data points. The suggestion that no apical curvature exists
in the guinea pig cochlea map is interesting in light of the
idea that the decrease in minimum cochlear best frequency
associated with this map feature serves to decrease travel-
ing wave reflections in the apex that may occur for
frequencies below the lowest mechanical best frequency
(Greenwood, 1996). However, it must be noted that the
repositioning of the single most apical point in Figure 1A
could change the k value significantly. Given the fact that
the measured CF of that apicalmost fiber was near the
lower frequency limit of our tuning-curve algorithm, the
possibility exists that the true CF was lower than the
measured CF, which would alter the map in the direction
of suggesting a nonzero value of k.
The best-fit line to the cochlear map values suggests that
the apical and basal extremes are tuned to 132 Hz and
197
54,720 Hz, respectively, for a total range of 8.7 octaves. The
a coefficient of 2.6, another measure of the frequency span
of the cochlea, is slightly higher than the values reported
for cat, human, chinchilla, and gerbil, which range from
2.1 to 2.2 (Greenwood, 1996). The a coefficients in all of the
above-mentioned species are significantly higher than the
coefficients for the opossum and rat (,1), where many
fewer octaves are spanned by the cochlea (Greenwood,
1996).
SR groups and the relation between
SR and threshold
Early recordings from the cat auditory nerve showed
that 1) the distribution of SRs is fundamentally bimodal,
with two peaks separated by a valley at rates of roughly
17.5 spikes/second; and 2) fibers in the high-rate peak tend
to have lower thresholds than those in the low-rate peak
(Kiang et al., 1965). The relation between SR and relative
threshold was revisited after computerization of data-
gathering techniques made it feasible to obtain data from
hundreds of fibers in a single animal (Liberman, 1978).
Analysis of data from individual cats showed that, if
thresholds were normalized to the average threshold for
fibers in the high-rate peak (high-SR fibers, SR . 17.5
spikes/second), then the high-SR group was homogenous
in threshold, with high-SR fibers deviating by no more
than 5 dB in relative threshold, whereas the low-rate peak
(SR , 17.5 spikes/second) contained fibers with relative
thresholds varying over a 60 dB range. Splitting the
low-rate peak into two groups produced a medium-SR
group (0.5 , SR , 17.5 spikes/second), which was also
homogeneous in threshold (intragroup variation of less
than 5 dB), with relative thresholds roughly 10 dB higher
than high-SR fibers. The low-SR group so defined con-
tained exclusively fibers with relative thresholds 10 dB
higher than the high-SR fibers; however, this group re-
mained heterogeneous, in that relative thresholds of low-SR
fibers ranged from 10 dB to 60 dB. Nevertheless, dividing
the SR dimension into these three groups has practical
value, in that it increases the accuracy with which nor-
mal auditory nerve thresholds can be defined (Liberman
and Kiang, 1978).
Although the fundamental correlation between SR and
threshold sensitivity has been replicated in the auditory
nerves of a number of other mammalian species, the
clarity with which two, three, or more SR groups can be
defined has varied. For example, Winter et al. (1990)
published data from the guinea pig showing increasing
threshold as SR declines, with low-SR fibers showing
significantly higher thresholds than high-SR fibers. How-
ever, in contrast to the results of the present study (Fig.
3D), the thresholds of low-SR and medium-SR fibers in the
data of Winter et al. were not significantly different from
each other, nor were the medium- and high-SR groups. In
the study of Winter et al., data were pooled across ten
animals, whereas the present study pooled across only
two. More importantly, in the Winter et al. study, the
relative threshold was computed from the CAP threshold
functions measured from the round window, whereas, in
the present study, relative threshold was computed from
the average threshold of high-SR fibers (only feasible with
198
Fig. 9. Photomicrographs of labeled endbulbs showing the SR-
based differences in ending complexity. The form factors for the
endings shown in AD were 0.74, 1.00, 0.55, and 1.01, respectively.
The scale bar in D also applies to AC. A: An HRP-filled endbulb from a
medium-SR fiber (CF 5 19.6 kHz, SR 5 2.5 spikes/second). B: A
large numbers of fibers sampled in each animal). Normal-
izing with respect to the CAP probably introduces addi-
tional scatter in the data, because the relation between
CAP threshold and ANF threshold probably varies with ings than routine animals raised in less controlled acous-
CF (Liberman, unpublished). One would expect that the
CAP thresholds might increasingly overestimate single
fiber threshold as CF decreases, because the density of
ANFs per octave of CF decreases steadily for CFs below
about 10 kHz.
Data from the present study suggest that, if the same
techniques are used to obtain and normalize the data, then
the relation between SR and threshold is very similar in
the cat and the guinea pig. In cat data, the clarity with
which three SR groups are separable in plots of relative
J. TSUJI AND M.C. LIBERMAN
biocytin-filled endbulb from a high-SR fiber (CF 5 0.4 kHz, SR 5 75
spikes/second). C: A biocytin-filled endbulb from a medium-SR fiber
(CF 5 15.6 kHz, SR 5 10.1 spikes/second). D: A biocytin-filled endbulb
from a high-SR fiber (CF 5 4.6 kHz, SR 5 34.4 spikes/second).
threshold vs. SR (e.g., Fig. 3A) depends on acoustic history:
Animals born and raised in a controlled low-noise environ-
ment show more clear-cut evidence of discrete SR group-
tic environments (Liberman, 1978). Thus, had the acoustic
history of the present animals been more carefully con-
trolled (they were born in a commercial facility and
shipped to our animal care facility, where they were
housed for variable times), the SR-based threshold differ-
ences might have been significantly cleaner.
Nevertheless, the question remains unresolved whether
the three SR groups in either species represent qualita-
tively different fiber populations or simply a useful way to
subdivide a continuum of response properties. The answer
AUDITORY NERVE LABELING IN GUINEA PIG
Fig. 10. The sizes of somatic terminals in the CN from all labeled
fibers in the present study (A) and the number of large somatic
terminals per fiber for fibers from the three SR groups (B). A: The
silhouette areas of all somatic terminals from labeled fibers are plotted
vs. the CF of the parent fiber. They are divided into two groups: those
to this question must await a complete description of the
mechanisms underlying the differences in SR. A thorough
discussion of the existing evidence on this subject can be
found elsewhere (Merchan-Perez and Liberman, 1996). In
brief, fibers from all three SR groups can innervate the
same IHC, and there is evidence for SR-based differences
in the ultrastructure of the IHC/RF synapse as well as
SR-based differences in the caliber of the peripheral
terminals of the RF, either of which could contribute to the
sensitivity differences of the different SR groups. Ulti-
mately, the functionally important differences may prove
to be difference in the numbers or types of ion channels
(e.g., voltage-gated calcium channels or neurotransmitter
receptors) in the IHC or RF synaptic membrane. Clearly,
differences in channel type would be a result consistent
with qualitative intergroup differences.
SR and peripheral fiber morphology
The existing intracellular labeling studies of ANFs in
the cat have shown a number of correlations between SR
and the light microscopic morphology. Within the organ of
Corti, it has been shown that the unmyelinated peripheral
terminals of low- and medium-SR fibers are smaller in
caliber than those of high-SR fibers and are spatially
segregated around the hair cell circumference: The pillar
side of the IHC (closer to the outer hair cells) is contacted
exclusively by high-SR fibers, whereas the modiolar side is
the site of all low- and medium-SR contacts as well as some
high-SR contacts (Liberman, 1982b; Liberman and Oliver,
1984; Merchan-Perez and Liberman, 1996).
It has been suggested that the small caliber of the low-
and medium-SR terminals could contribute to their de-
creased sensitivity and level of SR (Geisler et al., 1985): If
the spike generator sits at the first node of Ranvier
(central to the habenula), then the synaptic potentials
(EPSPs) generated at the RF/IHC synapse must be con-
199
on spherical cells (circles) and those on all other CN cell types
(squares). The division between large and small terminals is arbitrary.
B: The number of large terminals per fiber is plotted vs. fiber CF. Large
is defined as shown in A. For SR symbol conventions in B, see Fig-
ure 4A.
ducted passively along the unmyelinated fiber, and de-
creases in fiber diameter would increase the internal
resistance and decrease the size of the potential at the
node. Thus, a summation of more EPSPs would be re-
quired to generate action potentials in thinner fibers.
In data from the cat, the SR-based caliber differences are
reflected in the myelinated regions of the fibers as well
(Liberman and Oliver, 1984): Both peripheral and central
axons of low- and medium-SR fibers are clearly thinner (by
roughly 20%) than those of high-SR fibers. The guinea pig
data in Figure 4 indicate the same SR-based differences:
For both peripheral and central axons, the low- and
medium-SR fibers were roughly 2040% thinner than
those for high-SR fibers. Caliber data for the unmyelinated
portions were not available in the present study.
The functional significance of the SR-based spatial
segregation of RF terminals around the IHC is not yet
clear, although it has been noted that the inner spiral
bundles, which contain spiraling efferent fibers from the
lateral olivocochlear (LOC) system, run closer to the
modiolar than the pillar side of the IHC, and, correspond-
ingly, modiolar-side afferent fibers (predominately those
with medium- and low-SRs) are much more heavily inner-
vated by LOC synapses than those afferents from the pillar
side (Liberman et al., 1990). The asymmetry of this
efferent innervation could be involved in the development
of the spatial segregation of RF response types.
In data from the cat, the spatial segregation at the level
of the IHC is reflected in a degree of spatial segregation
within the osseous spiral lamina and a weak, but statisti-
cally significant segregation within the ganglion (Kawase
and Liberman, 1992). In the osseous spiral lamina, espe-
cially close to the organ of Corti, RFs in cat show the
organization expected if there is minimal crossing of fiber
trajectories as they travel from the IHC: The region of the
lamina closest to scala tympani contains fibers originating
200
Fig. 11. Camera lucida tracings of selected endbulbs on spherical
cells (top) and a scatterplot (bottom) of the form factors of all highest thresholds (Liberman, 1991). Physiological studies
spherical cell endbulbs from the present study. The form factor is a
measure devised by Sento and Ryugo (1989) to describe the extent to
which the ending is fenestrated: It is defined as the ratio of the
silhouette area to silhouette perimeter. For SR symbol conventions,
see Figure 4A.
on the pillar side of the IHC (high-SR only), whereas the
scala vestibuli side contains most of the low- and me-
dium-SR fibers as well as some high-SR fibers. The guinea
pig data from the present study (Fig. 6) did not allow
analysis of the unmyelinated terminals within the organ of
J. TSUJI AND M.C. LIBERMAN
Corti. However, the peripheral axons show signs of a
similar segregation within the osseous spiral lamina: All of
the low-SR fibers were seen in the scala vestibuli half of
the lamina, and the portion of the lamina closest to the
scala tympani contained exclusively high-SR fibers. The
data from the ganglion in the guinea pig (Fig. 6A) do not
show any SR-based organization; however, micropipette
recordings from the ganglion in the basal turn of the
guinea pig do show a statistically significant tendency for
the low- and medium-SR fibers to be found at the scala
vestibuli side of the ganglion (Brown, 1994). Given the
difficulties inherent in attempting to define a common,
one-dimensional spatial metric for a ganglion that changes
in its shape and orientation from base to apex and that is
cut in many different planes in our serially sectioned
material, and given the smaller database in the present
study than in the previous cat material, it is perhaps not
surprising that the present findings failed to yield convinc-
ing data on the issue of spatial organization.
In summary, the data from the present study clearly
show the same SR-based caliber differences reported in the
cat. With respect to the SR-based spatial organization, the
guinea pig data are less compelling, but they are not
inconsistent with the pattern that has been clearly demon-
strated in the cat.
SR and central fiber morphology
Existing intracellular labeling studies in the cat have
shown that fibers from all three SR groups 1) project to the
CN and 2) bifurcate into an ascending branch that inner-
vates the AVCN and a descending branch that innervates
the PVCN and the DCN (Fekete et al., 1982). Fibers of
different SR differ systematically in their degree of branch-
ing: low- and medium-SR fibers give off many more
swellings (Liberman, 1991; Fekete et al., 1982), both
terminal and en passant, which are presumed sites of
synaptic interactions with other neural elements within
the CN. These SR-based differences in degree of branching
are seen only for the ascending branches; thus, the present
study concentrated on the ascending branch.
Much of the increased branching and associated termi-
nal swellings given off by the ascending branches of low-
and medium-SR fibers is directed toward a particular
subdivision of the CN called the small-cell cap (Liberman,
1991). In cat, the small-cell cap constitutes a thin rind of
small multipolar cells capping the core regions of the
AVCN that contain the relatively large projection neurons,
such as the bushy cells and the large multipolar cells
(Osen, 1969; Hackney et al., 1990). In cat, virtually all
auditory nerve projections to the small-cell cap region
were from low- and medium-SR fibers, and especially large
projections were seen from those low-SR fibers with the
of the small-cell cap in cat have recently shown that there
are large numbers of cells with especially wide dynamic
range and low SRs (Ghoshal and Kim, in press), consistent
with a selective innervation by low- and medium-SR fibers.
In the magnocellular core regions of the AVCN in the cat,
the projections to spherical cells in anterior AVCN and
multipolar cells in posterior AVCN were similar in all
three SR groups, whereas projections to multipolar cells in
anterior AVCN were largely from low- and medium-SR
fibers, and projections to globular cells were largely from
high-SR fibers (Liberman, 1991, 1993). Labeling studies in
cat have also shown that the large endbulb terminals
AUDITORY NERVE LABELING IN GUINEA PIG
made by ANFs on spherical cells are morphologically
different according to the SR of the primary fibers (Sento
and Ryugo, 1989).
Although all aspects of CN physiology and anatomy are
better understood in the cat than in the guinea pig,
existing anatomical information from the guinea pig sug-
gests that the fundamental organization of the nucleus is
similar in the two species, including the existence of a rind
of small cells capping the magnocellular portions of the
AVCN (Hackney and Pick, 1986; Hackney et al., 1990).
Existing physiological studies suggest that the same funda-
mental response types are represented in the guinea pig in
the analogous areas of the nucleus (Winter and Palmer,
1990); the small-cell cap region has not been studied
physiologically in the guinea pig.
Data from the present study show that the basic SR-
based differences in central projection seen in the cat are
also present in the guinea pig. Specifically, there is a clear
difference in the degree of branching, with low-and me-
dium-SR fibers showing significantly more terminals (espe-
cially nonsomatic terminals) than high-SR fibers (Fig. 7B).
Furthermore, as reported for the cat, most of that striking
difference in branching complexity is accounted for by the
projections of low- and medium-SR fibers to the small-cell
regions (Fig. 7C,D). As reported for the cat, there were no
clear SR-based differences in the numbers of large termi-
nals, or endbulbs, per fiber (Fig. 10B), most of which are on
spherical cells. Of course, because there are fewer low- and
medium-SR fibers than high-SR fibers, most of the end-
bulbs in the CN will arise from high-SR fibers. The
SR-based differences in endbulb morphology reported in
the cat were also clearly demonstrable in the guinea pig
data. Sento and Ryugo (1989) showed that the endbulbs of
low- and medium-SR fibers are more complex and more
highly fenestrated than those of medium-SR fibers: The
same relationship held for the guinea pig, as shown in
Figures 9 and 11. The functional significance of this
morphological heterogeneity is not clear: Sento and Ryugo
suggest that high-SR terminals may contain more or
larger mitochondria (due to the increased levels of activ-
ity), and the distribution of this intracellular machinery
may dictate the coarser shape of the high-SR endbulb.
We did not attempt to separate multipolar cells into
those from the AVCN vs. the PVCN, and most globular
cells in the guinea pig are innervated by the descending
branches (which were not analyzed): Thus, the other
SR-based trends reported for the cat VCN were not tested
by the present study.
In summary, several of the important SR-based differ-
ences in fiber morphology, branching complexity, and
innervation of the small-cell cap have now been demon-
strated in both guinea pig and cat, two mammalian species
that are not particularly close phylogenetically. This dem-
onstration in two mammalian species increases the likeli-
hood that these structure-function relations constitute a
general mammalian plan and, thus, apply to the human
case as well.
LITERATURE CITED
Adams, J.C. (1977) Technical considerations on the use of horseradish
peroxidase as a neuronal marker. Neuroscience 2:141145.
Borg, E., B. Engstrom, G. Linde, and K. Marklund (1988) Eighth nerve fiber
firing features in normal-hearing rabbits. Hearing Res. 36:191202.
Brown, M.C. (1994) Antidromic responses of single units from the spiral
ganglion. J. Neurophysiol. 71:18351847.
201
Dallos, P. (1973) Cochlear potentials and cochlear mechanics. In A.R. Moller
(ed): Basic Mechanisms in Hearing. New York: Academic Press, pp.
335372.
Eldredge, D.H., J.D. Miller, and B.A. Bohne (1981) A frequency-position
map for the chinchilla cochlea. J. Acoust. Soc. Am. 69:10911095.
Fekete, D.M., E.M. Rouiller, M.C. Liberman, and D.K. Ryugo (1982) The
central projections of intracellularly labeled auditory nerve fibers in
cats. J. Comp. Neurol. 229:432450.
Geisler, C.D., L. Deng, and S.R. Greenberg (1985) Thresholds for primary
auditory nerve fibers using statistically defined criteria. J. Acoust. Soc.
Am. 77:11021109.
Ghoshal, S., and D.O. Kim (1996) Marginal shell of the anteroventral
cochlear nucleus: Acoustically weakly driven and not-driven units in
the unanesthetized decevesrate cat. Acta Otolaryngol. (Stockh.) 116:280
283.
Greenwood, D.D. (1961) Critical bandwidths and the frequency coordinates
of the basilar membrane. J. Acoust. Soc. Am. 33:13441356.
Greenwood, D.D. (1996) Comparing octaves, frequency ranges, and cochlear-
map curvature across species. Hearing Res. 94:157162.
Hackney, C.M., and G.F. Pick (1986) The distribution of spherical cells in
the anteroventral cochlear nucleus of the guinea pig. Br. J. Audiol.
20:215220.
Hackney, C.M., K.K. Osen, and J. Kolston (1990) Anatomy of the cochlear
nuclear complex of guinea pig. Anat. Embryol. 182:123149.
Kawase, T., and M.C. Liberman (1992) Spatial organization of the auditory
nerve according to spontaneous discharge rate. J. Comp. Neurol.
319:312318.
Kiang, N.Y.S., T. Watanabe, E.C. Thomas, and L.F. Clarke (1965) Discharge
Patterns of Single Fibers in the Cats Auditory Nerve. Cambridge:
M.I.T. Press.
Kiang, N.Y.S., M.C. Liberman, J.S. Gage, C.C. Northrop, L.W. Dodds, and
M.E. Oliver (1984) Afferent innervation of the mammalian cochlea. In
L. Bolis, R.D. Keynes, and S.H.P. Maddrell (eds): Comparative Physiol-
ogy of Sensory Systems. Cambridge University Press, pp. 143161.
Kohloffel, L.U.E. (1972) A study of basilar membrane vibrations. Acoustica
27:4958.
Leake, P.A., and R.L. Snyder (1989) Topographic organization of the central
projections of the spiral ganglion in cats. J. Comp. Neurol. 281:612629.
Liberman, M.C. (1978) Auditory-nerve response from cats raised in a
low-noise chamber. J. Acoust. Soc. Am. 63:442455.
Liberman, M.C. (1980) Morphological differences among radial afferent
fibers in the cat cochlea: An electron-microscopic study of serial
sections. Hearing Res. 3:4563.
Liberman, M.C. (1982a) The cochlear frequency map for the cat: Labeling
auditory-nerve fibers of known characteristic frequency. J. Acoust. Soc.
Am. 72(5):14411449.
Liberman, M.C. (1982b) Single-neuron labeling in the cat auditory nerve.
Science 216:12391241.
Liberman, M.C. (1991) Central projections of auditory nerve fibers of
differing spontaneous rate, I: Anteroventral cochlear nucleus. J. Comp.
Neurol. 313:240258.
Liberman, M.C. (1993) Central projections of auditory nerve fibers of
differing spontaneous rate. II: Posteroventral and Dorsal cochlear
nuclei. J. Comp. Neurol. 327:1736.
Liberman, M.C., and N.Y.S. Kiang (1978) Acoustic trauma in cats, cochlear
pathology and auditory-nerve activity. Acta Otolaryngol. 358:563.
Liberman, M.C., and M.E. Oliver (1984) Morphometry of intracellularly
labeled neurons of the auditory nerve: Correlations with functional
properties. J. Comp. Neurol. 223:163176.
Liberman, M.C., L.W. Dodds, and S. Pierce (1990) Afferent and efferent
innervation of the cat cochlea: Quantitative analysis with light and
electron microscopy. J. Comp. Neurol. 301:443460.
Lorente de No, R. (1981) The Primary Acoustic Nuclei. New York: Raven
Press.
Merchan-Perez, A., and M.C. Liberman (1996) Ultrastructural differences
among afferent synapses on cochlear hair cells: Correlations with
spontaneous discharge rate. J. Comp. Neurol. 371:208221.
Muller, M. (1991) Frequency representation in the rat cochlea. Hearing
Res. 51:247254.
Muller, M. (1996) The cochlear place-frequency map of the adult and
developing mongolian gerbil. Hearing Res. 94:148156.
Osen, K.K. (1969) Cytoarchitecture of the cochlear nuclei in the cat. J.
Comp. Neurol. 136:453484.
202
Osen, K.K. (1970) Course and termination of the primary afferents in the
cochlear nuclei of the cat: An experimental anatomical study. Arch. Ital.
Biol. 108:2151.
Robertson, D. (1984) Horseradish peroxidase injection of physiologically
characterized afferent and efferent neurones in the guinea pig spiral
ganglion. Hearing Res. 15:113121.
Robertson, D., A.R. Cody, G. Bredberg, and B.M. Johnstone (1980) Response
properties of spiral ganglion neurons in cochleas damaged by direct
mechanical trauma. J. Acoust. Soc. Am. 67:12951303.
Rouiller, E.M., R. Cronin-Schreiber, D.M. Fekete, and D.K. Ryugo (1986)
The central projections of intracellularly labeled auditory nerve fibers
in cats: An analysis of terminal morphology. J. Comp. Neurol. 249:261
278.
Russell, I.J., and P.M. Sellick (1978) Intracellular studies of hair cells in the
mammalian cochlea. J. Physiol (London) 284:261290.
Ryugo, D.K., and D.M. Fekete (1982) Morphology of primary axosomatic
endings in the anteroventral cochlear nucleus of the cat: A study of the
endbulbs of Held. J. Comp. Neurol. 210:239257.
Salvi, R., J. Perry, R.P. Hamernik, and D. Henderson (1982) Relationships
between cochlear pathologies and auditory nerve and behavioral re-
sponses following acoustic trauma. In R.P. Hamernik, D. Henderson,
and R. Salvi (eds): New Perspectives on Noise-Induced Hearing. New
York: Raven Press, pp. 165188.
Schmiedt, R.A. (1989) Spontaneous rates, thresholds and tuning of auditory-
J. TSUJI AND M.C. LIBERMAN
nerve fibers in the gerbil: Comparisons to cat data. Hearing Res.
42:2336.
Schmiedt, R.A., and J.J. Zwislocki (1977) Comparison of sound-transmis-
sion and cochlear-microphonic characteristics in Mongolian gerbil and
guinea pig. J. Acoust. Soc. Am. 61:133149.
Schuknecht, H.F. (1953) Techniques for study of cochlear function and
pathology in experimental animals. Acta Otolaryngol. (Stockholm)
58:377391.
Sento, S., and D.K. Ryugo (1989) Endbulbs of Held and spherical bushy
cells in cats: Morphological correlates with physiological properties. J.
Comp. Neurol. 280:553562.
Spoendlin, H.H. (1969) Innervation of the organ of Corti of the cat. Acta
Otolaryngol. (Stockholm) 67:239254.
Wilson, J.P., and J.R. Johnstone (1975) Basilar membrane and middle ear
vibration in the guinea pig measured by capacitive probe. J. Acoust. Soc.
Am. 57:705723.
Winter, I.M., and A.R. Palmer (1990) Responses of single units in the
anteroventral cochlear nucleus of the guinea pig. Hearing Res. 44:161
178.
Winter, I.M., D. Robertson, and G.K. Yates (1990) Diversity of characteristic
frequency rate-intensity functions in guinea pig auditory nerve fibres.
Hearing Res. 45:191202.
Yates, G.K. (1991) Auditory-nerve spontaneous rates vary predictably with
threshold. Hearing Res. 57:5762.