When a tissue volume is sectioned, cells or other objects are cut into
segments by the sectioning process. The Abercrombie and empirical met
hods count object segments in histological sections and then apply a c
orrection formula to convert the segment count to object number There
has been considerable recent controversy over whether these methods sh
ould be abandoned (in favor of the disector). Although both methods ap
pear unbiased as thought experiments on paper, regardless of variation
in object size, shape, or orientation, in practice two problems are i
nherent in the segment-counting approach: the practical problem of los
t caps and a conceptual flaw that becomes apparent only when a need fo
r unbiased estimation of certain factors in the correction formulae is
seriously addressed. The Abercrombie method is inevitably biased by l
ost caps, whereas in the empirical method, this potential bias can be
avoided. In the Abercrombie formula, the relevant factor to be estimat
ed (aside from section thickness) is H, mean object height in the axis
perpendicular to the section plane, and in the empirical method, it i
s the ratio of segment number to object number. In both methods, the f
actor in question should be estimated from an unbiased sample of the t
otal population of objects. But unbiased selection of a statistically
adequate number of objects for this estimation constitutes an unbiased
, statistically adequate count. Once this is done, there is no reason
to complete the steps for estimation of the factor; the count in this
specimen is finished. It is shown that the empirical method's serial s
ection procedure can be used to estimate H. This estimate is more sens
itive to lost caps than the Abercrombie equation, but, when the formul
a for H is substituted into the Abercrombie equation, the lost caps er
ror disappears. However, this approach is useless, as making this subs
titution transforms the Abercrombie equation into the empirical method
equation. (C) 1998 Wiley-Liss, Inc.