"If you want to break down counts to many dimensions you need to represent occurrences according to those dimensiions." I agree, and this is the approach we are taking (multiple lots with their own biocuration properties, see below).
To follow up with Rich's observations on sex (and things like "stage" etc). In our software we have introduced the concept of a Biocuration class. Individuals in this class are biologicaly derived concepts *that are used to organize collections*. As Rich noted you have to make a bit of a leap from asserting that an individual classified under the biocuration class "female" is in fact female. In our software, by default, we do not make this leap, but we are considering extending the system to allow users to add domain/range constraints that would assert additional classifications, e.g. if a specimen is classified under the biocuration class "female(s)", then that specimen can be classified as a
http://www.ebi.ac.uk/efo/EFO_0001265.
We feel that the biocuration concept more closely approximates the day-to-day intent/usage of classifiers like sex, stage, "furs", "bones" etc. as organizational concepts used to manage collections, rather than a specific biological assertion about the specimen(s).
Cheers,
Matt