Thanks Ed, that makes it a good deal clearer.

The thought behind my original question was, in effect: Where a primary key contains multiple fields, can I use it (or its index) to quickly access those records where the first component field is the same? The answer, fairly clearly from what you and HanShi have said, is Yes, so there is no point in having an extra index for just that field.

Re the clustered vs. non-clustered aspect though, because (a) I will want to quickly pull out all the records relating to a given document at one time; and (b) this will happen quickest when the records are in the correct physical order; and (c) given your answers, I don't expect to need any more indexes; and finally (d) the table, once populated, will rarely be updated... I am inclined to go for a clustered index. Does this make sense?

Feel free to shoot holes in any mistakes in my reasoning.

MarkD