Saying "When" to Literature

As an unpublished researcher, I tend to wonder what proportion of citations appearing in publications in life sciences journals (including bioinformatics and computational biology journals) have authors typically read thoroughly? All? A quarter?

On a related note, how does one know how much of a paper to read? More importantly, how does one know how much effort to expend comprehending a publication? The law of diminishing returns holds in the pursuit of reading scientific literature as much as any other, but when has one really taken due diligence in reading a publication?

My continued difficulty reading scientific literature motivated these questions. I suspect I may have a reading disability, as even when reading for pleasure, I consume material at around half to a third of the speed of my bookish acquaintances. The possibility of this aside, however, I typically find the writing in scientific publications dense and abstruse. Some of this could come from attempting to read literature at the crossroads of biology, computer science, and statistics, subjects of which I possess some smattering but no authoritative comprehension. When I get stuck on a particular section, that usually means I've reached my limit of understanding that section, as the field of bioinformatics is so vast that even within my own research group, we tend to not read the same literature, preventing discussion and collaborative learning.

Often when reading a publication, I lament the inability to cull something immediately and practically useful from the time spent reading it the same way I could when reading a book on a programming language, a blog post, or technical document. I wonder if perhaps I have wound up in the wrong field of work (engineer's mind in a scientist's world) or have not chosen an appropriate topic of study.

I'd love to hear some anecdotal evidence and suggestions from researchers and other students out there in this crazy pursuit we call science.