Hearing Research 278 (2011) 43e51

Contents lists available at ScienceDirect

Hearing Research
journal homepage: www.elsevier.com/locate/heares

Unbiased counting of neurons in the cochlea of developing gerbils

C.-P. Richter a, b, c, *, G. Kumar a, E. Webster a, S.K. Banas a, D.S. Whitlon a, c, d
a
Department of Otolaryngology, Northwestern University, Feinberg School of Medicine, 303 East Chicago Avenue, Searle 12-561, Chicago, IL 60611-3008, USA
b
Department of Biomedical Engineering, Northwestern University, 2145 Sheridan Road, Tech E310, Evanston, IL 60208, USA
c
The Hugh Knowles Center, Department of Communication Sciences and Disorders, Northwestern University, Evanston, IL 60208, USA
d
Interdepartmental Neurosciences Program, Northwestern University, Chicago, IL 60611, USA

a r t i c l e i n f o a b s t r a c t

Article history: Accurate counting of neurons in the cochlea has a significant impact on the interpretation of research and
Received 22 December 2010 clinically relevant data. However, reports of numbers of neurons in the spiral ganglion are widely
Received in revised form variable across studies, even within the same species. We suggest that the implementation of a more
28 January 2011
standardized, unbiased counting method will improve the consistency and accuracy of neuron counts
Accepted 9 February 2011
Available online 15 February 2011
and will impact scientific interpretations. To test this view, we compared, in different ways, the numbers
of neurons in the spiral ganglia of developing gerbils, previously reported to decrease by 22e27%
between birth and age 7 days. Cochleae from gerbils, aged newborn, 7 days, 20 days, 1.5 years and 2.5
years were embedded in Araldite and serially sectioned at 5 mm. A computer based stereological method
was used to unambiguously count every neuron in serial sections, either throughout the entire cochlea,
or in a 100-mm segment of the cochlea. No significant changes in neuron numbers during cochlear
maturation were found. We demonstrate that in methods using sampling of sections, the identity of the
starting section and the interval between sections impacts the variability of the estimate of neuron
numbers. In addition, we show that packing density differs between the newborn and seven-day old
animals. The data demonstrate that variability in counting methods and the comparison of non-uniform
samples can lead to neuron number estimates that show differences where none exist. We propose that
a standardized counting protocol be implemented across studies and suggest possible approaches to
different types of comparisons between neurons of spiral ganglia from different sources.
Ó 2011 Published by Elsevier B.V.

1. Introduction timesaving. When the samples are non-uniform e for example

Neuron counting in the cochlea is a vital but time-consuming
undertaking that, more often than not, is abbreviated for simplicity
and efficiency. Yet, accurate counting of the total number of
cochlear spiral ganglion cells impacts the interpretation of research
and clinically relevant data. Changes in neuron number are serious
consequences of genetics or cochlear damage, are indicative of
physiologically relevant developmental alterations, and are the
target of interventions designed to maintain, repair or repopulate
cochlear neurons. In the nervous system, counting protocols are
generally abbreviated by estimating overall neuron numbers from
counts of small samples. When the samples are representative for
the entire population, such an approach can be powerful and
* Corresponding author. Northwestern University, Feinberg School of Medicine,
Department of Otolaryngology e Head and Neck Surgery, 303 East Chicago Avenue,
Searle 12-561, Chicago, IL 60611-3008, USA. Tel.: þ1 312 503 1603; fax: þ1 312 503
1616.
E-mail address: cri529@northwestern.edu (C.-P. Richter).
between ages, or between genetic strains, or before and after
cochlear damage, the most accurate method to use is not as
straightforward, and a variety of physical and computational
procedures have been advanced to compensate for under- and
over-estimates of neuron numbers (Coggeshall, 1992; Coggeshall
and Lekan, 1996; Hedreen, 1998; West, 1999; Benes and Lange,
2001; King et al., 2002; West, 2002; Stark et al., 2005; Goto and
Goto, 2006). Cell counting can be “biased” (they contain a statis-
tical sampling or measurement error) or “unbiased” (they have no
systematic error; the number counted represents the true value). In
1996, the Journal of Comparative Neurology published a landmark
statement underscoring the importance of unbiased counting
methods in research (Saper, 1996).
In our perusal of the spiral ganglion literature, we were
surprised not only by the variation in methodology used across
studies, but also by the wide range of reported neuron numbers,
even within one species (see Table 1). We wondered whether
a more standardized, unbiased counting method would improve
the consistency and accuracy e and thus interpretations e of the

0378-5955/$ e see front matter Ó 2011 Published by Elsevier B.V.

doi:10.1016/j.heares.2011.02.003
44 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51

Table 1
The table shows the variability in cell counts for different species obtained by different groups using various techniques for estimating the total cell counts.

Spiral ganglion cell counts for different species

Species Method Age Section Separation Count Correction method Reference

thickness (no of sections)
(mm)
Hybrid MF1 Optical 5 days 0.5 1 16,100 (Jensen and Gundersen, (Camarero et al., 2001)
and 129/sv fractionator 1993; Tandrup et al., 1997)
Hybrid MF1 Optical 20 days 0.5 1 16,600 (Jensen and Gundersen, (Camarero et al., 2001)
and 129/sv fractionator 1993; Tandrup et al., 1997)
CBA/J Profile count Adult 4 10 12,577 None (Webster, 1985)
C57BL6/6NNia Profile count 18 months 8 4 3,132 Abercrombie (Park et al., 1990)
mouse (diet)
C57BL6/6NNia Profile count 18 months 8 4 2,399 Abercrombie (Park et al., 1990)
mouse (no diet)
C57/BL6 Profile count Newborn 10 6 8,082 None (Farinas et al., 1994)
WT (C57BL/6) Profile count 0 days 5 6 4,421 Corrected (Liebl et al., 1997)
Hybrid MF1 Optical 5 days 0.5 1 16,100 (Jensen and Gundersen, (Camarero et al., 2001)
and 129/sv fractionator 1993; Tandrup et al., 1997)
Hybrid MF1 Optical 20 days 0.5 1 16,600 (Jensen and Gundersen, (Camarero et al., 2001)
and 129/sv fractionator 1993; Tandrup et al., 1997)
129  C57/BL6 Profile count 8 5 7,800 Abercrombie (Postigo et al., 2002)
background
129  C57 Profiles counts Newborn 8 10 7,148 None (Luikart et al., 2003)
(computer
automated)
BDNFNT3/NT3 mice Profile count 0 days 14 3 6,500 Abercrombie (Agerman et al., 2003)
WT (C57Bl/6) Profile count 0 days 14 3 6,500 Abercrombie (Agerman et al., 2003)
BDNFNT3/NT3 mice Profile count 17 days 14 3 6,100 Abercrombie (Agerman et al., 2003)
WT (C57BL/6) Profile count 17 days 14 3 6,800 Abercrombie (Agerman et al., 2003)
CD-1 Stereological 0 days 5 1 8,240 No correction (Whitlon et al., 2006)
reconstruction
C57BL6/129Svj Stereological 10 wk 5 1 10,402 No correction Richter, (unpublished)
reconstruction

Rat (Sprague-Dawley) Profile count 1e2 months 10 15,800 Konigsmark (Keithley and Feldman, 1979)
Rat (wistar albino) Profile count 6 days 5 10 18,700 Abercrombie (Rueda et al., 2003)
Rat (wistar albino) Profile count 30 days 5 10 18,000 Abercrombie (Rueda et al., 2003)
Rat (wistar albino) Profile count 60 days 5 10 17,404 Abercrombie (Rueda et al., 2003)

Gerbil Profile count 0 days 4 10 22,000 Abercrombie (Echteler and Nofsinger, 2000)
Gerbil Profile count 5 days 4 10 w17,000 Abercrombie (Echteler and Nofsinger, 2000)
Gerbil Profile count 7 days 4 10 15,000 Abercrombie (Echteler and Nofsinger, 2000)
Gerbil Stereological 0 days 5 1 23,573 No correction present study
Gerbil Stereological 7 days 5 1 24,521 No correction present study

Cat Profile count Adult 44,298e57,494 (Howe, 1934)

Cat Profile count Adult 20 39,000 (Schuknecht, 1953) (Schuknecht, 1960)
Cat Profile count P1-8 years 20 10 45,000- (Schuknecht, 1953) (Mair, 1973)
Cat Profile count Adult 20 5 49,175e50,109 No correction (Chen et al., 2010)

Human Profile count 22 months 20 10 33,915 Factor of 0.9 (Hinojosa et al., 1985)
Human Profile count 22e60 years 20 10 33,702 Factor of 0.9 Pollak et al., (1987)
Human Stereological 22e60 years 20 10 33,600e45,700 No correction (Ishiyama et al., 2001)

counts of neuron numbers in the cochlea. Using the guidelines of
Coggeshall and Lekan (1996), which indicate the necessity for cal-
ibrating neuronal counts, we set out to reevaluate neuron counting
in the gerbil cochlea. We took for comparison, the neuron counts in
developing gerbils, which have been reported to decrease by
22e27% between postnatal days 3 and 7 (Echteler and Nofsinger,
2000).
2. Methods
2.1. Animals
Gerbils (Meriones unguiculatus) of either sex aged, newborn,
postnatal day (P) 7, P20, 1.5 and 2.5 years were derived from our
breeding colony at Northwestern University Medical School.
Animal studies were approved by the Animal Care and Use
Committee of Northwestern University.
2.2. Embedding and sectioning the cochleae
Animals were euthanized by a lethal injection of pentobarbital
(>200 mg/kg bodyweight, intraperitoneal) and then decapitated.
The bullae were removed from the skull and were transferred into
Hanks Balanced Salt Solution (HBSS). The cochleae were exposed
and placed into 4% paraformaldehyde in 0.1 M phosphate buffer, pH
7.4 (PB) for 4 h at room temperature (typically at 21  C). Cochleae of
the seven-day old and older animals were then decalcified for one
week by placing them into an ethylene glycol tetraacetic acid
(EGTA, 10%) phosphate buffer (0.1 M), pH 7.4 at 4  C. After fixation
and decalcification, the cochleae were washed in 0.1 M PB three
times for 15 min each, then dehydrated in a graded 6 step acetone
series (25%, 50%, 75%, 90%, 100% and 100%). Cochleae were passed
through acetone solutions of increasing concentrations of Aral-
diteeEpoxy Resin (Acetone:Resin e 7:1, 1:1, 1:7). Cochleae were
washed three times in pure Resin, then embedded in Araldite and
maintained for 12 h at 60  C to cure the resin.
 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51 45

Serial plastic sections were cut along the modiolar plane of the
cochlea at 5 mm on an ultramicrotome (LKB 8800 Ultrotome III).
Number of sections counted in each cochlea: for newborn animals,
230, 192, and 171 sections; for adult animals: 281 and 238 sections.
The sections were placed on glass slides and were stained with 1%
toluidine blue (SigmaeAldrich) and 1% sodium tetraborate solution
(1:20). After staining, several pictures of the spiral ganglia on each
section were captured and stored in a computer.
2.3. Counting neurons e stereological method
Digital imaging and the computer program Adobe Photoshop
were used to unambiguously count neurons by reconstructing
serial sections, as has been previously reported (Whitlon et al.,
2006; Richter et al., 2008). The basic procedure is to locate
profiles of the spiral ganglion cells in section 1, which is called the
lookup section. In section 2 (reference section) all spiral ganglion
cells are counted that are not present in section 1. After counting
the cells in section 2, that section then becomes the lookup section
for section 3, and so on. To do this, every section is photographed.
The first section in which spiral ganglion cells are present is dis-
played in Adobe Photoshop (Fig. 1A). An additional layer to the
image is inserted and a circle of the approximate size of the
neuronal cell nuclei is marked on the layer for each cell counted.
Characteristic landmarks are also added (Fig. 1B). After the cells in
section 1 are counted, the corresponding image of section 2 is
opened. The inserted layer from section 1 is then copied onto
section 2 (Fig. 1C and D). The images are aligned so that the land-
marks superimpose (Fig. 1D). This clearly identifies the cells from
section 1 that were already counted. Another layer is inserted on
the image, and the cells in section 2 that were not counted in
section 1 are then counted and marked on the inserted layer. This
then identifies the already counted cells in section 2 and allows
them to be omitted in the counts of section 3, and so on (Fig. 1D). In
this way, every neuron in the cochlea can be counted without
systematic errors due to over- or under-counting.
To simplify the counting, we did not distinguish between type I
and type II spiral ganglion neurons.
2.4. Measuring of dimensions
The cross-sectional area of Rosenthal’s canal was measured on
digital images using ImageJ (Wayne Rasband, NIH, public domain
software). The program’s scale was calibrated using the image of
a standard slide, and the scale setting was changed from pixels to
micrometers. Area measurements of Rosenthal’s canal were
obtained by tracing the bony opening housing the spiral ganglion
cells. In newborn animals, the bone has not been ossified and the
spiral ganglion cells are tightly packed. In those animals, often
the opening in the cartilage is similar to the shortest line encircling
the spiral ganglion cells, which is Rosenthal’s canal. The total
number of pixels within a circumscribed area was calculated and
converted into square micrometers or square millimeters.
For selected locations along the cochlea, the distance between
the nuclei of neighboring spiral ganglion cells was determined for
newborn and seven-day old animals. For the measurements, the
images were opened in MATLAB and each neural profile in the
cross-section was marked. The distance between a cell and any

Fig. 1. A: Shown is a 5 mm thick section along the mid-modiolar plane of a newborn gerbil cochlea. The section has been stained with Toluidine blue. This image has been displayed
with Adobe Photoshop. An additional, digital layer was inserted. B: Each cell counted is marked on the inserted layer with a black circle approximately the size of the nucleus.
Characteristic landmarks are also added to this layer. After counting the cells in the first (lookup) section, the corresponding image of the next section, which is named the reference
section, is opened. C: The marked up inserted layer of the lookup section is then copied onto the reference section and the sections are aligned so that the landmarks superimpose.
This allows clear identification of the cells from the lookup section that were already counted. D: After inserting a layer on the reference section, those cells that were not counted on
the previous section are counted and marked with a gray circle on the inserted layer. This layer will then serve as a lookup section for the next section.
46 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51

Table 2
Spiral ganglion cell counts for a 100 mm segment along the mid-modiolar section of
the cochlea. Cells were counted on 20 consecutive sections using the stereological
method described in Methods. Cell counts were similar between newborn and adult.

Number of spiral ganglion cells along a 100 mm segment of the spiral ganglion

Age Location 0 days 7 days 20 days 1.5 years 2.5 years

Basal 458  7 330  116 493  89 446  23 369  99
Upper basal 275  9 240  28 339  106 292  34 255  10
Middle 308  9 308  35 283  55 274  59 254  70
Upper middle 532  135 422  99 582  176 646  281 535  9
Apical 734  41 895  211 647  142 685  249 705  90

other cell in the cross-section was calculated and displayed in

a histogram. The average distance provided a measure of the
packing density of the neurons. The larger the average distances
were, the smaller was the packing density.

2.5. Statistics

Averages and standard deviations were calculated for the

neuron counts and the distances between the cells. An analysis of
variance (ANOVA) was performed. If the ANOVA indicated differ-
ences among the means, a posteriori test was used for making
pairwise comparisons among the means. An honestly significant
difference (HSD) test by Tukey was used. The tests are part of
a statistical package provided by IGORÒ (Wavemetrics). Statistical
decisions were made for a probability p ¼ 0.05.

Fig. 2. A: Cross-sectional area measurements of Rosenthal’s canal at different locations

3. Results
Using a sampling approach that differed from those used in
previous studies, every neuron in 20 serial mid-modiolar cochlear
sections (a 100-mm segment along the spiral ganglion) was counted
using the method described. The number of spiral ganglion neurons
in newborn gerbils was compared to the number in 2.5-year old
gerbils. We expected to see a 22e27% decrease in neuron numbers
reflecting that seen previously between newborn and seven-day
old animals (Rueda et al., 1987; Echteler and Nofsinger, 2000).
However, very little difference was found in neuron numbers
between newborn and adult gerbils (Table 2, Fig. 2).
To verify the similarity in neuron numbers that were found
using the stereological method on only 20 consecutive sections,
every neuron in each of three newborn and two P7 gerbil cochleae
was counted unambiguously. The cell counts were 23873  3605
(N ¼ 3) for newborn and 24814  95 (N ¼ 2) for seven-day old
animals (Fig. 3). We further assessed the possibility that different
observers would return different neuron counts by re-counting one
cochlea. The difference in neuron numbers between observers was
1.6%, demonstrating that the counting method is very consistent
and independent of which trained observer did the measurements
(Fig. 3).
To determine whether different sampling techniques could be
responsible for differences between our results (newborn and P7
along the spiral ganglion for different ages of the animals. The cross-sectional area
increases with age. B: spiral ganglion cells were counted for 20, sequential mid-
modiolar sections, each 5 mm thick. The stereotactic method, described in Methods,
was used. The average number of spiral ganglion cells along a 100 mm section of the
spiral ganglion was similar for 0 day, 7 day, 20 day, 1.5-year and 2.5-year old animals.
Differences in spiral ganglion cell counts were not statistically significant.
numbers; or newborn and 2.5-year numbers were essentially the
same) and prior results (neuron numbers decrease by 22e27% or
more by P7), the spiral ganglion neuron counts in each section were
recorded separately (Fig. 4). We then determined what the final
neuron number estimate would have been under conditions where
the following parameters were varied: (a) the starting section; (b)
the length of the sequence; (c) the interval between sampled
sections. Fig. 5 and Table 4 demonstrate that increasing the interval
between sections used to estimate the total number of cells, every
2nd, 3rd, 4th, ., nth section, increased the variability in outcomes.
The packing density was quantified by measuring the area of
Rosenthal’s canal and the distances between the profiles in one
cross-section. The measurements showed that the cross-sectional
area of Rosenthal’s canal is non-uniform across the cochlea as well
as across ages. In newborn animals, the area, expressed in mm2 is
0.01  0.004 at the basal turn, 0.005  0.001 at the upper basal turn,
0.005  0.001 at the middle turn, 0.01  0.007 at the upper middle
turn, and 0.019  0.008 at the apical turn (Table 3, Fig. 2A). The

Table 3
Averages and standard deviation for the cross-sectional area measurements of Rosenthal’s canal are provided. The cross-sectional area increases with age, in particular
between newborn and P7.

Cross-sectional area in square millimeter

Age Location 0 days 7 days 20 days 1.5 years 2.5 years

Basal 0.0102  0.0043 0.0107  0.0027 0.0243  0.0096 0.0233  0.0095 0.0379  0.0014
Upper basal 0.0049  0.0007 0.0067  0.0028 0.0172  0.0062 0.0145  0.0002 0.0203  0.0039
Middle 0.0046  0.0012 0.0092  0.0024 0.0156  0.0002 0.0155  0.0043 0.0243  0.0041
Upper middle 0.0108  0.0067 0.0202  0.0180 0.0251  0.0032 0.0315  0.0144 0.0520  0.0019
Apical 0.0188  0.0077 0.0303  0.0130 0.0282  0.0060 0.0301  0.0127 0.0569  0.0116
 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51
Fig. 3. The total count of spiral ganglion cells is graphed as a category plot. The
numbers were obtained from three newborn animals and from two seven-day old
animals. The stereological method described in Methods was used to count the cells.
The numbers in the bars are the results for each of the animals. The numbers in
parentheses are the recounts from a second independent observer. The numbers
indicate actual inter-animal variability.
cross-sectional areas were clearly larger in the 2.5-year old animals,
although the numbers of neurons counted were similar in newborn
adult gerbils e hence a difference in packing density. Correspond-
ing area measures in 2.5-year old animals in mm2 were
0.04  0.001 at the basal turn, 0.02  0.004 at the upper basal turn,
0.02  0.004 at the middle turn, 0.05  0.002 at the upper middle
turn and 0.06  0.012 at the apical turn. Since the number of
neurons changed little, we were interested in whether the packing
density changed with the increase of Rosenthal’s canal. The packing
density of the neurons was estimated by the distance between
selected neurons and their neighbors. The mode for the distances
between neurons increased for the first turn locations between
newborn and seven-day old animals from about 30 mm to about
50 mm. For the upper turns the mode of the distances between the
cells changed little between the newborn and the seven-day old
animals and was about 40 mm (Fig. 6). The decrease in packing
density caused the locations of the neurons in the P7 animals to be
less uniform than in the newborn animals. The changes indicate
that calibration (correction of profile counts to actual counts) is
necessary for each age group and each location along the cochlea.
The estimates of neuron number should only be compared directly
if similar conditions for the counting exist.
Having reconstructed and counted every slice in the gerbil
cochleae, allows us to determine the effect of selecting the starting
point, increasing the thickness of the slice, and the effect of the
length of the sequence. Fig. 4 shows the cell counts for each section
Fig. 4. Traces show the number of neurons counted on each section. Thin lines are
obtained from the three newborn animals and the thick lines show the counts of the
seven-day old animals. The traces show that the number of neurons counted on each
section varies along the cochlea. Consequently, the selection of the interval between
sections used to estimate the total number of neurons will affect the estimated cell
counts. The variability of the traces predicts that the total number of cells estimated
from samples will also depend on the starting point of the sequence (see Fig. 5).
47
Fig. 5. The x-axis in booth panels shows the number of sections, which were skipped.
For the case that five sections are skipped, spiral ganglion cells were counted on the
0th, 5th, 10th, ., nth section and provided one estimate for the total spiral ganglion
cell count. The counting was repeated by using the 1st, 6th, 11th, ., (n þ 1)th section
and a second estimate for the total number was obtained. The procedure was repeated
until counting started at the 5th section. From the five estimated total counts of the
spiral ganglion cells, the minimum and the maximum were selected and are shown in
panel A. Panel B shows the same values obtained in A; the changes are normalized to
the estimate when all the neurons were counted. The figure shows that the estimation
of the spiral ganglion cells depends on the length of the sequence and the selected
starting point.
along the cochlea. Thin lines represent the newborn animals; the
P7 animals are shown by the thick lines. The counts on each section
differ largely indicating that the distribution of the cells is not
uniform in Rosenthal’s canal. It is also apparent that the starting
point for a series of sections to estimate the total cell count as well
as the number of sections between the sections selected can alter
the outcome of the estimates. In the following we have varied the
number of sections skipped and the starting point of the sequence.
For each of the variations the total estimate was determined. For all
the results obtained from such variations, the maximum and
minimum estimated cell count is shown in Fig. 5. The x-axis in Fig. 5
shows the number of sections, which were skipped. The following
example will demonstrate the procedure that was used to arrive at
the corresponding y-values. For the case that 5 sections are skipped
(x-axis-value ¼ 5), spiral ganglion cells were counted on the 0th,
5th, 10th, ., nth section and provided one estimate for the total
count of neurons. Counting was repeated by using the 1st, 6th, 11th,
., (n þ 1)th section and a second estimate for the total number of
neurons was obtained. The procedure was repeated until counting
started at the 5th section. From the five estimated total counts of
the spiral ganglion cells, the minimum and the maximum value is
shown in Fig. 5. Fig. 5A provides the total numbers. Solid lines show
the maximum counts and broken lines the minimum count. The
results show that it is possible to err by 50% if every tenth section is
counted. The difference between the two values provides insight in
the susceptibility to error of skipping sections in estimating the
total counts of spiral ganglion cells.
If the same starting section is used but the thickness of the
section is increased and the number of skipped sections is varied,
 48
Table 4
The table demonstrates the effect of changing the thickness of the section (increasing thickness along the x-axis) and the number of sections skipped for estimating the total number of neurons (y-axis). The numbers in the table
provide the deviation in % from the total counts obtained by counting each section for the entire cochlea. Each table represents a different animal. Results from two newborn and two P7 animals are shown. The deviation increases
for smaller sections and for longer intervals between sections counted.

Thickness of the section Thickness of the section

(mm) e newborn (mm) e newborn

C.-P. Richter et al. / Hearing Research 278 (2011) 43e51

Distance to start of next 5 10 15 20 25 30 35 40 45 50 Distance to start of next 5 10 15 20 25 30 35 40 45 50
section counted (mm) 5 0 section counted (mm) 5 0
10 1.64 0 10 1.81 0
15 3.94 0.50 0 15 1.27 1.62 0
20 2.63 1.22 1.23 0 20 1.35 0.98 1.41 0
25 6.11 0.62 2.31 0.30 0 25 0.18 1.65 0.48 0.69 0
30 0.69 0.66 1.08 0.77 0.30 0 30 1.81 0.59 1.49 0.94 0.64 0
35 3.24 0.99 1.06 2.02 0.40 0.71 0 35 0.81 0.65 1.37 0.10 1.24 0.59 0
40 4.55 2.30 1.71 1.25 1.14 0.88 1.03 0 40 1.62 3.99 2.35 1.19 1.17 0.11 0.88 0
45 5.87 2.91 2.04 1.69 1.48 1.13 1.73 0 45 6.32 2.66 4.53 1.29 0.53 0.56 0.23 0.10 0
50 2.79 2.26 0.41 2.13 0.93 0.79 0.33 1.00 0.70 0 50 3.17 3.06 1.02 0.21 1.62 0.88 1.07 1.04 0.19 0
Thickness of the section Thickness of the section
(mm) e 7 days (mm) e 7 days
Distance to start of next 5 10 15 20 25 30 35 40 45 50 Distance to start of next 5 10 15 20 25 30 35 40 45 50
section counted (mm) 5 0 section counted (mm) 5 0
10 3.56 0 10 2.05 0
15 0.05 1.14 0 15 1.72 0.23 0
20 4.60 0.25 0.67 0 20 1.83 0.96 0.11 0
25 3.65 1.99 1.86 2.04 0 25 0.77 1.28 1.13 0.02 0
30 5.08 2.50 2.08 0.26 1.09 0 30 4.92 4.38 2.13 1.32 0.55 0
35 7.22 1.45 3.98 2.73 3.02 1.46 0 35 5.05 2.98 2.04 0.44 0.51 0.60 0
40 1.68 5.04 3.43 3.39 0.76 1.09 0.18 0 40 3.24 3.38 0.56 2.74 2.11 1.34 1.23 0
45 7.02 4.72 2.15 0.73 0.39 0.28 0.55 0.87 0 45 8.59 2.48 3.06 1.94 0.60 0.71 0.32 1.26 0
50 16.37 3.02 3.17 4.71 2.69 0.73 1.68 1.26 0.53 0 50 4.22 0.06 0.89 1.81 1.54 2.24 0.35 0.22 0.03 0
 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51
Fig. 6. The distance between each of the profiles (nucleoli of the neurons), which can
be seen in a cross-section at the first turn, the second turn, and the apex of the gerbil
cochlea were determined. Newborn animals and gerbils at postnatal day seven were
compared. The distance of the profiles increased with increasing age in the first and
the second turn. Little changes can be observed for the apical turn. In the basal and the
middle turn the average distance between the profiles increased from 30 mm to 50 mm.
The increase in distance indicates that the packing density decreases.
deviations from the expected count are also evident (Table 4). A
good example is presented with the lower left cochlea in Table 4:
For 5 mm thick sections (x-axis, first column) and every 10th section
considered (y-axis, tenth row), the estimated number of neuron is
16.4% lower than the actual number. As can be seen from the table,
the estimation of the total number of cells can be improved if the
sections are selected thicker (or moving along the x-axis to the
right), or by increasing the number of sections which are consid-
ered (or moving upwards along the y-axis). The four examples
shown in Table 4 show that increasing the thickness of the sections
to 10e15 mm and decreasing the number of sections skipped to ten
or less (<50 mm) would allow a practical balance of accuracy and
efficiency.
49
4. Discussion
By counting, unambiguously, every neuron in the gerbil spiral
ganglion, we have observed that total neuron counts do not change
significantly between birth and P7. This observation differs from
those of Rueda et al. (1987) for rat and Echteler and Nofsinger
(2000) for gerbil, who calculated a decrease in neuron number
between P3 and P7. We have shown with an in depth analysis of
neuron numbers in each section throughout the cochlea, that the
causes of this discrepancy must lie with the method by which
neuronal population was estimated.
In the nervous system, total counts of neurons are usually
acquired by estimating from counts of small samples. Logically, use
of such an approach requires that the samples be representative for
the entire population. Methods to count cells include (1) profile
counts, (2) reconstruction of serial sections (as in the present
work); (3) the combination of (1) and (2), and (4) other stereo-
logical methods (Coggeshall and Lekan, 1996). Common pitfalls of
any given counting method are no less likely to occur in the cochlea
than in any other region of the nervous system. If these factors are
not identified and controlled for, they can result in counts that
generate differences that do not actually exist between samples.
4.1. Counting
Counting the profiles of entire cells, cell nuclei, or nucleoli in
sections taken at constant intervals, is the most common approach
to estimating cell numbers. The number of sections that are eval-
uated can vary, but typically, every 10th section is counted and the
resulting counts are then multiplied by 10. Estimating cell numbers
with this method is based on the assumption that each profile
counted represents one cell. However, one profile can be repre-
sented on more than one section. To account for the over-repre-
sentation of profiles during counting, corrections of the raw counts
that take into account the thickness of the sections and the size of
the profile have been established (e.g. Abercrombie, 1946;
Konigsmark et al., 1969) and used in some of the prior studies of
spiral ganglion neurons. Abercrombie’s equation, which is a modi-
fied correction from Linderstrøm-Lang et al. (1935), converts the
number of counted profiles, to an estimate of the true number of
objects. The following equation is commonly used to correct the
counts: Countcells ¼ Countprofiles  x/(x þ d), where x is the thick-
ness of the section, and d is the diameter of the profile. The
correction indicates that the thickness of the histological section
should be significantly larger than the size of the structure of
interest, so that x/(x þ d) approximates 1 and Countcells comes as
close as possible to Countprofiles. For example, the nucleus of
a neuron is typically about 7 mm in diameter. For 4 mm thick
sections this structure can be counted, in theory, on three sections
giving an over-estimate of cell number. To reduce the error of the
total number of neurons, the structure to be counted should be
small relative to the thickness of the section. Based on our results,
a minimum ratio of 1:10 should be acceptable. Due to the their
small size (1e2 mm), nucleoli are preferred to estimate the number
of neurons. However, although most nuclei have one nucleolus,
none, one, or even two and three nucleoli can be found in a nucleus.
The number of nucleoli is not consistent for different conditions,
e.g. during the development of the cochlea. Nonetheless, despite
those limitations, profile counts can be fairly reliable if a calibration
is established (see below).
4.2. Stereological methods
In 1984, Sterio introduced the dissector principle to allow for
unbiased estimates of cell number that are independent of the size
50 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51
and shape of the particles to be estimated. The physical dissector
method uses sequences of sections and counts only the last profile of
the object in each section. Instead of using series of sections for
which an object is tracked, the optical dissector uses two or more
different optical planes in a thick section. (Sterio, 1984; Gundersen,
1988; Gundersen et al., 1988b; Gundersen et al., 1988a; Coggeshall
and Lekan, 1996). With the development of novel imaging tech-
niques (Lareida et al., 2009; Richter et al., 2009; Santi et al., 2010)
spiral ganglion cells can be visualized with tomographic approaches
and computer based stereological counting techniques will become
available. With those techniques it will be feasible to count the entire
cell population within a reasonable time. For the spiral ganglion local
variations of cell density and size of the spiral ganglion can be
considered (Johnson et al., this issue).
4.3. Calibration
For calibration, the actual counts of well identifiable and recon-
structable targets, in this case neurons, are compared with the raw
counts of the profiles (e.g. Coggeshall, 1992). To adjust the
raw counts to the actual counts, the ratio of the actual count to the
raw count is multiplied by the raw profile counts. The calibration has
to be repeated if changes in target size, or packing density occur.
Particularly prone to error are the cases where there are non-
uniformly distributed cell populations or if only local changes occur.
For elliptical objects, the orientation of the major axis is another
confounding factor. Calibration can account for such variability.
4.4. Packing density
When neurons are densely packed, they may develop a regular
distribution pattern (similar to that in a crystal) in which common
distances exist between the neurons. In this case, the sampling
frequency, which is the number of sections skipped for estimating
the total count, can make a difference in the accuracy of the count.
For example, if the most common distance (mode) between neurons
is 30 mm, estimates of the total number of neurons based on
a sampling interval of 30 mm will differ from those obtained with
a sampling interval of 50 mm. We have determined the most
common distance of the nucleoli in several sections and at several
locations along the cochlea and for different ages. The distances are
not equally distributed; the mode of the distribution changes with
age and location (Fig. 6). With decreasing packing density, the most
common distance between neurons increases. This observation
holds for profiles as well as the neurons themselves. For the present
assessment, the distances were measured between each profile
(nucleolus) and all its possible neighbors in Rosenthal’s canal shown
in one section. The distribution of the distances provides a single
mode. The distances increased from about 30 mm in newborn to
about 50 mm in P7 animals. Changes were observed in the first and
second turns of the cochlea but not in the apical segment. In other
words, changes in packing density of the neurons were observed and
were not uniform along the cochlea. Estimating total counts of spiral
ganglion neurons would therefore require a calibration for the two
age groups and for the different locations along the cochlea.
4.5. Variability of total spiral ganglion cell numbers
As can be seen in Table 1, most estimates of the spiral ganglion
neuron number have been acquired with profile counts. In many
cases assumption based corrections or the Method by Abercrombie
has been applied to correct the raw counts of the profiles, but no
calibrations have been used. As expected, the numbers of spiral
ganglion neurons, even for one species, vary widely among groups.
For example, as mentioned above, the number of spiral ganglion
cells has been estimated for the developing gerbil and the rat. Both
studies used profile counts to estimate the total number of neurons
and reported that between newborn and postnatal day seven
approximately 22e27% of the neurons disappear. In the present
study, the number of neurons for 100 mm segments along the
cochlea did not change significantly between newborn and 2.5-
year old animals.
Previous studies did not discuss the possible implications of
morphological changes that occur at the same time and might
affect the estimates of the total number of spiral ganglion cells.
During development, thyroid hormone stimulates the opening of
Rosenthal’s canal. In the gerbil cochlea the cross-sectional area
changes approximately by a factor of two between newborn and P7.
This is similar to what has been reported by Echteler and Nofsinger
(2000). Differences between the latter and our study exist for the
absolute values and also for developmental changes in the most
basal portion of the cochlea (12e15% from the basal and of the
cochlea). While we did not find changes at the very basal end of the
cochlea, Echteler and Nofsinger (2000) reported an increase of
Rosenthal’s canal by a factor of w2. Discrepancies in values and in
developmental changes observed for the cross-sectional area of
Rosenthal’s canal may originate in the method of measuring the
area. While Echteler and Nofsinger (2000) traced the smallest
perimeter of modiolar tissue that surrounded all ganglion cells
within each cochlear turn, we traced the opening in the bone/
cartilage. In the study by Echteler and Nofsinger (2000), the values
for the cross-sectional area of Rosenthal’s canal are determined by
the packing density of the spiral ganglion cells and thus the changes
in cross-sectional area can be caused by the change in packing
density of the spiral ganglion cells.
Coinciding with this developmental change, it has been repor-
ted that the total number of spiral ganglion cells decreases in the rat
(Rueda et al., 1987) and the gerbil (Echteler and Nofsinger, 2000)
but not in the mouse (Camarero et al., 2001; Agerman et al., 2003).
In one study, neonatal hypothyroidism was introduced during the
critical developmental phase and the spiral ganglion neuron counts
were reported to remain constant from P0 to P60, leading the
authors to propose that the developmental decrease in spiral
ganglion neurons is due to thyroid hormone (Rueda et al., 2003). An
alternative interpretation now presents itself. In contrast to the
structural changes occurring during normal development, hypo-
thyroid rats do not exhibit large increases in the size of Rosenthal’s
(Uziel et al., 1983). This would make the P7 cochleae more struc-
turally similar to those at birth. This would allow the direct
comparison between newborn and P7 hypothyroid rats that, due to
morphologic differences, was not accurate in the normal rats.
4.6. Recommendations
For counting we suggest the follow approach:
 Select the section thickness about 10 times the thickness of the
profile to be counted. For counting nucleoli in the cochlea the
sections should be at least 10 mm thick.
 For each condition and location along the cochlea, calibrate the
counts. This is done by using no less than 10 serial sections to
compare the actual number of neurons in each section with the
profile counts. Calculate a correction factor for the profile
counts, i.e. number of actual neurons/number of profiles.
 Use profile counts for each condition and location along the
cochlea multiplying the resulting number with the correction
factor obtained from the calibration.
 To estimate the total number of neurons in a cochlea, use 10 mm
sections sampled at intervals of 50 mm. Use the correction
factor (s) then multiply by 5.
 C.-P. Richter et al. / Hearing Research 278 (2011) 43e51
Total numbers versus local changes:
 For a few questions, including developmental changes or
changes caused by genetic modifications the total number of
the spiral ganglion cells is required.
 Frequently local changes of spiral ganglion cell numbers in the
cochlea are of interest. To determine local changes, the neurons
are counted at selected and corresponding locations. For each
condition and each location a calibration of counts is per-
formed and the numbers can be compared. At least 10 sections
should be included in the analysis at a location.
Acknowledgements
This project has been supported by the American Hearing
Research Foundation, the Hugh Knowles Center, Northwestern
University, and the Department of Otolaryngology, Northwestern
University.
References
Abercrombie, M., 1946. Estimation of nuclear population from microtome sections.
Anat. Rec. 94, 239e247.
Agerman, K., Hjerling-Leffler, J., Blanchard, M.P., Scarfone, E., Canlon, B., Nosrat, C.,
Ernfors, P., 2003. BDNF gene replacement reveals multiple mechanisms for
establishing neurotrophin specificity during sensory nervous system develop-
ment. Development 130, 1479e1491.
Benes, F.M., Lange, N., 2001. Two-dimensional versus three-dimensional cell
counting: a practical perspective. Trends Neurosci. 24, 11e17.
Camarero, G., Avendano, C., Fernandez-Moreno, C., Villar, A., Contreras, J., de
Pablo, F., Pichel, J.G., Varela-Nieto, I., 2001. Delayed inner ear maturation and
neuronal loss in postnatal Igf-1-deficient mice. J. Neurosci. 21, 7630e7641.
Chen, I., Limb, C.J., Ryugo, D.K., 2010. The effect of cochelar-implant-mediated
electrical stimualtion on sopiral ganglion cells in congenitally deaf white cats.
J. Assoc. Res. Otolaryngol.. doi:10.1007/s10162-010-0234-3.
Coggeshall, R.E., 1992. A consideration of neural counting methods. Trends Neu-
rosci. 15, 9e13.
Coggeshall, R.E., Lekan, H.A., 1996. Methods for determining numbers of cells and
synapses: a case for more uniform standards of review. J. Comput. Neurol. 364, 6e15.
Echteler, S.M., Nofsinger, Y.C., 2000. Development of ganglion cell topography in the
postnatal cochlea. J. Comput. Neurol. 425, 436e446.
Farinas, I., Jones, K.R., Backus, C., Wang, X.Y., Reichardt, L.F., 1994. Severe sensory and
sympathetic deficits in mice lacking neurotrophin-3. Nature 369, 658e661.
Goto, N., Goto, J., 2006. Morphometric evaluations of the human nervous system.
Hum. Cell. 19, 49e64.
Gundersen, H.J., 1988. The nucleator. J. Microsc. 151, 3e21.
Gundersen, H.J., Bendtsen, T.F., Korbo, L., Marcussen, N., Moller, A., Nielsen, K.,
Nyengaard, J.R., Pakkenberg, B., Sorensen, F.B., Vesterby, A., West, M.J., 1988a.
Some new, simple and efficient stereological methods and their use in patho-
logical research and diagnosis. APMIS 96, 379e394.
Gundersen, H.J., Bagger, P., Bendtsen, T.F., Evans, S.M., Korbo, L., Marcussen, N.,
Moller, A., Nielsen, K., Nyengaard, J.R., Pakkenberg, B., Sørensen, F.B.,
Vesterby, A., West, M.J., 1988b. The new stereological tools: disector, fraction-
ator, nucleator and point sampled intercepts and their use in pathological
research and diagnosis. APMIS 96, 857e881.
Hedreen, J.C., 1998. What was wrong with the Abercrombie and empirical cell
counting methods? A review. Anat. Rec. 250, 373e380.
Hinojosa, R., Seligsohn, R., Lerner, S.A., 1985. Ganglion cell counts in the cochleae of
patients with normal audiograms. Acta Otolaryngol. 99, 8e13.
Howe, H.A., 1934. The reaction of the cochlear nerve to distrution of its endorgan:
a study on deaf albino cats. J. Comput. Neurol. 62, 73e79.
Ishiyama, A., Agena, J., Lopez, I., Tang, Y., 2001. Unbiased stereological quantification
of neurons in the human spiral ganglion. Neurosci. Lett. 304, 93e96.
51
Jensen, E.V.B., Gundersen, H.J.G., 1993. The rotator. J. Microsc. 170, 35e44.
Keithley, E.M., Feldman, M.L., 1979. Spiral ganglion cell counts in an age-graded
series of rat cochleas. J. Comput. Neurol. 188, 429e442.
King, M.A., Scotty, N., Klein, R.L., Meyer, E.M., 2002. Particle detection, number
estimation, and feature measurement in gene transfer studies: optical frac-
tionator stereology integrated with digital image processing and analysis.
Methods 28, 293e299.
Konigsmark, B.W., Kalyanaraman, U.P., Corey, P., Murphy, E.A., 1969. An evaluation of
techniques in neuronal population estimates: the sixth nerve nucleus. Johns
Hopkins Med. J. 125, 146e158.
Lareida, A., Beckmann, F., Schrott-Fischer, A., Glueckert, R., Freysinger, W., Muller, B.,
2009. High-resolution X-ray tomography of the human inner ear: synchrotron
radiation-based study of nerve fibre bundles, membranes and ganglion cells.
J. Microsc. 234, 95e102.
Liebl, D.J., Tessarollo, L., Palko, M.E., Parada, L.F., 1997. Absence of sensory neurons
before target innervation in brain-derived neurotrophic factor-, neurotrophin
3-, and TrkC-deficient embryonic mice. J. Neurosci. 17, 9113e9121.
Linderstrøm-Lang, K., Holter, H., Ohlsen, A.S., 1935. Studies on enzymatic histo-
chemistry: XIII. The distribution of enzymes in the stomach of pigs as a function
of its histological structure. Comptes-rendus du Lab. Carlsberg 20, 66e125.
Luikart, B.W., Nef, S., Shipman, T., Parada, L.F., 2003. In vivo role of truncated trkb
receptors during sensory ganglion neurogenesis. Neuroscience 117, 847e858.
Mair, I.W., 1973. Hereditary deafness in the white cat. Acta Otolaryngol. Suppl. 314,
1048.
Park, C.J., Cook, K.C., Verde, E.A., 1990. Dietary restriction slows the abnormally
rapid loss of spiral gagnlion neurons in C57/BL6 mice. Hear. Res. 48, 275e280.
Postigo, A., Calella, A.M., Fritzsch, B., Knipper, M., Katz, D., Eilers, A., Schimmang, T.,
Lewin, G.R., Klein, R., Minichiello, L., 2002. Distinct requirements for TrkB and
TrkC signaling in target innervation by sensory neurons. Genes Dev. 16, 633e645.
Richter, C.P., Shintani-Smith, S., Fishman, A., David, C., Robinson, I., Rau, C., 2009.
Imaging of cochlear tissue with a grating interferometer and hard X-rays.
Microsc. Res. Tech. 72, 902e907.
Richter, C.P., Bayon, R., Izzo, A.D., Otting, M., Suh, E., Goyal, S., Hotaling, J.,
Walsh Jr., J.T., 2008. Optical stimulation of auditory neurons: effects of acute and
chronic deafening. Hear. Res. 242, 42e51.
Rueda, J., de la Sen, C., Juiz, J.M., Merchan, J.A., 1987. Neuronal loss in the spiral
ganglion of young rats. Acta Otolaryngol. 104, 417e421.
Rueda, J., Prieto, J.J., Cantos, R., Sala, M.L., Merchan, J.A., 2003. Hypothyroidism
prevents developmental neuronal loss during auditory organ development.
Neurosci. Res. 45, 401e408.
Santi, P., Johnson, S., Schmitz, H., Ramakrishnan, K., 2010. 3D anatomical atlas of the
cochlea of the CBA/J mouse. Abstr. Assoc. Res. Otolaryngol. 33, 101.
Saper, C.B., 1996. Any way you cut it: a new journal policy for the use of unbiased
counting methods. J. Commun. Network 364, 5.
Schuknecht, H.F., 1953. Techniques for study of cochlear function and pathology in
experimental animals; development of the anatomical frequency scale for the
cat. AMA Arch. Otolaryngol. 58, 377e397.
Schuknecht, H.F., 1960. Neuroanatomical correlates of auditory sensitivity and pitch
discrimination in the cat. In: Rasmussen, G.L., Windle, W.F. (Eds.), Neural Mech-
anisms of the Auditory and the Vestibular Systems. Thomas, Springfield, IL, p. 76.
Stark, A.K., Pelvig, D.P., Jorgensen, A.M., Andersen, B.B., Pakkenberg, B., 2005.
Measuring morphological and cellular changes in Alzheimer’s dementia:
a review emphasizing stereology. Curr. Alzheimer Res. 2, 449e481.
Sterio, D.C., 1984. The unbiased estimation of number and sizes of arbitrary particles
using the disector. J. Microsc. 134, 127e136.
Tandrup, T., Gundersen, H.J., Jensen, E.B., 1997. The optical rotator. J. Microsc. 186,
108e120.
Uziel, A., Legrand, C., Ohresser, M., Marot, M., 1983. Maturational and degenerative
processes in the organ of Corti after neonatal hypothyroidism. Hear. Res. 11,
203e218.
Webster, D.B., 1985. The spiral ganglion and cochlear nuclei of deafness mice. Hear.
Res. 18, 19e27.
West, M.J., 1999. Stereological methods for estimating the total number of neurons
and synapses: issues of precision and bias. Trends Neurosci. 22, 51e61.
West, M.J., 2002. Design-based stereological methods for counting neurons. Prog.
Brain Res. 135, 43e51.
Whitlon, D.S., Ketels, K.V., Coulson, M.T., Williams, T., Grover, M., Edpao, W.,
Richter, C.P., 2006. Survival and morphology of auditory neurons in dissociated
cultures of newborn mouse spiral ganglion. Neuroscience 138, 653e662.