Query Bias

When JOG used to post here regularly he talked quite a bit about query bias. It also came up recently on the TTM mailing list. It also gets a vague nod on the wikipedia page on database normalisation.

Does the term have any formal, well defined or shared meaning?

I think I understand the intuition, that in some sense an 'unbiased' data model minimises some sort of complexity metric aggregated across some set of queries. (I think that 'some set' might actually be replaced by 'all' for some people).

But complexity can mean lots of things:

   How simple is the query to write?
   How simple is the query to execute?
   How simple is the query to understand?

What does it mean to the folk here?