Thanks Jeff. I would call that manually allocating by author, which is what I did for our review.
I had problems though (which I think led to triple/quadruple allocation for screening).
The default method of ordering references (on logging in) appears to be alphabetical in our review, but it isn't quite (and not all the blank author references appear at the top). I therefore needed to ensure that:
1) all ~4000 references were displayed (rather than the default 700 records), then
2) ensure that they are correctly sorted alphabetically (rather than almost aplhabetically), then
3) find the exact place on the long list where I allocated up to (e.g. if the first batch for reviewers 1 and 2 was authors A-D, the second batch for reviewers 3 and 4 will be E-J).
I realise this is less of a problem if all references are being allocated to only two reviewers, but nevertheless I found this process to be quite time-consuming (and, if I forgot to sort aplhabetically, and used the default almost-alphabetical sorting, it could lead to quadruple allocation).
Bearing in mind the advantages I mentioned (of allocating by author) I wonder whether this could be a more worthwhile system (if it's feasible, of course) than randomly allocating?
Thanks
Mark