----- Original Message ----- > Sorry, Dave, you have the example wrong! > > The results would be: > 80 total > 43 distinct > 40 unique > > Your juxtaposition of the meaning of 'distinct' and 'unique' is probably > more logical, but 'distinct' was being used already in the current sense. :-) You are correct! Which also sorts out Allan's question :-) _________end of Original Message _____________________ This explanation seems to imply that "distinct" should always be larger than "unique" And in most cases this is indeed the case - see http://freebmd.rootsweb.com/db-stats.html I'm now struggling to think of an explanation for the exceptions - eg. marriages 1839, 1841, 1848, 1850, 1852; and births 1840. It seems to imply that the normal code (ie the code that produces the "distinct" count) sometimes over-combines, ie. it combines entries which look similar but are in fact different. Presumably if Jo Bloggs and his widowed dad, also Jo Bloggs, both decide to get married on the same day in the same church and appear on the same page in the marriage register, this should create two identical index entries. If these two entries are then transcribed by two different people, it would be virtually impossible for the code to distinguish these transcriptions from a double-keying of the same record. But if that's the explanation, how does the new "unique-count" code pick it up? Andrew Gough ___________________________________________________________________________________________________ Visit the web site of the Financial Times at http://www.ft.com