Pitfalls and Solutions


  • Not planning to use the appropriate section thickness.
    • Some unbiased stereological probes require thin optical or physical sections, some require thick sections and small depth of field and z resolution. Please see the individual probes for details. You should know how thick to cut your sections before you section the tissue.
  • Not planning for isotropy.
    • The probe or the tissue or some combination of the two must be isotropic; no interaction of the probe with the feature can be favored. Depending on the probe, you may be required to use isotropic or vertical sections or you may be able to use preferred orientations when sectioning. Please see the individual probes for details.
  • Failure to label the tissue all the way through the section.
    • Beware of ‘dead zones’ in the middle of the section where the labeling did not reach. This often occurs if you are using an antibody that doesn’t fully penetrate the tissue sections.





  • Failing to use systematic and random sampling.
    • Section through the entire region; using a set interval but with a starting section picked by chance.
      • Avoid using one section per region, for instance through the middle. It is very unlikely that a precision this low will work (see the MBF Bioscience webinar: What is Bias, for more on precision vs. bias concerns, especially minute 1:09).
    • Use a set spacing and a random starting spot within a section also. The number of sites will vary depending on the size of the section.
  • Counting too many events or not enough events overall.
    • If the means between or among groups are in your opinion, far apart, but the statistical test does not show significance, you have to ask ‘is this because not enough sampling was done?’. On the other hand, you don’t want to do extra sampling and waste time and resources (however, please see The oversampling approach).
  • Counting too many events per counting site.
    • The goal is not to know a lot about every place you sample but to spread out the sampling, “do more less well” (Gundersen and Osterby, 1981).
  • Using an objective lens with too low of a magnification and/or numerical aperture.
    • For probes that require thick sections and small focal planes, you must use an objective lens with a numerical aperture high enough to resolve many focal planes through the section. For example, the optical fractionator probe is typically done with sections that are between 15 to 30 microns thick using an objective with a numerical aperture of 1.3 or 1.4. Please see the individual probes for details.
  • Not using a focus encoder when estimating number or length in three dimensions.
  • Not using a unique point to count a particle, such as a cell or a synapse.
    • When using the optical or physical disector, a consistent, unique point must be used to identify a particle, this is often the leading edge of the particle as it comes into focus. Whatever criteria you choose to use, you should  decide on it before starting the counting, and use it consistently to determine whether to count the particle or not.
  • Blindly trusting automatic counting techniques.
    • Nearly all automatic counting software is not accurate (Schmitz, et al., 2014). The fact that it is commercially availabile, or or in wide use is not proof that these techniques work.



  • Analyzing density only.
    • Avoid the volume/reference trap: if you use the Nv-Vref method, be sure to obtain the reference volume and multiply it by the numerical density to avoid being fooled into, for instance, thinking number went up when actually volume went down.
    • Always report the numerator and denominator and not just the density. For example, don’t report only the density of cells or the length of fibers per volume but instead graph the explicit number or length vs. the volume.



  • Not defining the anatomical region and stereology parameters in your manuscript.
    • For the sake of reproducibility and clarity there should be a good explanation of how you defined the region where the sampling is taking place and what parameters you used for the unbiased stereology.


Gundersen, H.J.G., and R. Osterby, 1981, Optimizing Sampling Efficiency of Stereological Studies in Biology: or ‘Do more less well!’. J. of Microscopy, Vol. 121, pp. 65-73.

Schmitz, C, Eastwood, B.S., Tappan, S.J., Glaser, J.R., Peterson D.A., and P.R. Hof, 2014, Current Automated 3D Cell Detection Methods are not a Suitable Replacement for Manual Stereologic Cell Counting. Fron Neuroanat. 8, article 27, pp. 1-13.




Sponsored by MBF Bioscience
developers of Stereo Investigator, the world’s most cited stereology system