It would seem that people who teach reading a lot, over many years, would be in the best position to see what works and what doesn't, in that they operate in a kind of laboratory, and can see, in the long term if not the short term, the success of their actions. Much of what scientific research proves is based on what teachers suspect in the first place, through teaching, but in fact real life puts barriers between teachers and researchers, so it's not that simple.
As a teacher I have always found it necessary to both keep up on the research and carry on rigorous in-class research on what works and what doesn't at least for my own students. My own classroom habits have shown me that it's not hard to distinguish the students who actually work through a reading from the ones who don't; and, it's rare for students who work hard on a reading to not want to tell you how they feel about it. So, a teacher who pays close attention to readers over the years can improve reading choice, tailor it to students' needs, and get readings to fit readers better with time. From this experience I judge the current state of research.
Extended Reading controversy
Krashen once said that all standardized tests, such as the TOEFL, should be abolished and replaced with programs in which students are simply given time to read, and good books to devour. The philosophy of extended reading rests on the fluency one gets simply from reading a lot of anything, starting with a good novel, and is based on the premise that since most good first-language readers became good readers by reading and getting pleasure from a number of novels, second language readers should be able to do the same. This is because fluency and confidence are necessary building blocks; the pleasure that comes from novels is a reward; and reading is really really fun if you get a good novel and a soft chair. Extended reading programs have had varying success worldwide, but the followers of the system basically devote countless hours to researching variations of Krashen's advice and hoping to turn the world's classrooms into reading libraries with soft chairs.
Here are my problems with the extended reading camp. We have tried to make students read novels and noticed that there is no such thing as "pleasure reading" in a second language. Without tests, etc., students are more likely going to fall asleep in a soft chair, than read a novel. Since novels are deliberately written for enjoyment, the more serious textbook readers consider novel-reading a painful diversion that doesn't even take them where they want to go with their reading; they feel an enormous burden (sometimes 10 pages/day), and don't ever get the pleasure out of it that they would if, for example, they were a young schoolchild.
From these programs and from experience I can say they are right about some aspects: first, volume, pure volume of reading, is the crucial variable. Students can and do make themselves read large quantities, since that's what they need, and teachers can help and reward them by noticing their hard work and passing them to a higher level (this can't be done without tests, though, since some students can pretend to read or read poorly and still look a lot like they are reading; such students skew the results and make all efforts appear to be fruitless). But, extended reading would work better if all learners were young, with plenty of time on their hands, and getting pleasure was the main reward they would expect from any given venture; they aren't, and it's not. Second, careful attention to pleasure students do get from any reading is not wasted; whether it is a textbook, news article, or low-level newspaper.
Context clues controversy
A few years ago a friend of mine got fed up with the "context clues" camp and the exercises they put into reading texts, and pointed out that if students don't know much vocabulary, they can't very well apply that skill to what they read. He knew, instinctively, that it's a skill all readers know and use all the time, but mostly when they know 90% or more of the words they are reading; it doesn't work when they are facing more than 10% of unfamiliar words. He suspected that that limit was quantifiable, and went about finding it; he set about proving his theory, and soon did.
I learned several things from this controversy, somewhat by accident. First, using context clues is something that any sentient reader does naturally; we don't have to teach people how to do it. Second, calling it "guessing" is offensively incorrect and is likely to be misunderstood by the vast majority of learners; if it's truly guessing, why bother? Third, there are some "clues" that are genuinely new to learners, such as using text organization to predict topic sentence and support; but, successful use of these skills relies more on being able to do these things as one reads, in other words, not so much being aware of the skill, but being able to practice using it while decoding language simultaneously. Thus I concluded that practice, practice, practice was a better strategy than doing context clues exercises. It is worthwhile to point out that one can use context clues, but students already know that; it's often more productive to point out where they could have done it, but didn't because they were too busy translating.
Folse, K. (2004). "Myths about Teaching and Learning Second Language
Vocabulary: What Recent Research Says." TESL Reporter 37 (2), 1-13.
Folse, K. (2004). "Limitation of Context Clues in Real-World Comprehension:
A Case Study" Nexus: A Journal for Teachers in Development 7, 1.
She is a friend of mine who set out to do research on the component skills of reading. She figured that if the testing companies (ETS, etc.) quantified and measured these skills separately, they ought to at least play out as statistically discrete. But they didn't. The essence of her research showed that, at least as far as discrete skills go, reading in its essence is an intermix of skills relying on grasp of vocabulary, such that as vocabulary grows, the skills begin intermixing rapidly, but never developing discretely. People therefore don't tend to be imbalanced in the skills; nor do they show any particular favoring of one skill over another.
(SCALAR 2007) http://www.studentgroups.ucla.edu/scalar/scalar4/SCALAR4abstracts.html
I used to have much more, but have only this to share. Given the importance of vocabulary, what do we do to help our students remember it, and have it available for use whenever they wish? Much has been written; it has been studied, and here's a place to start.
Llinares, G., Leiva, B., Cartaya, N., & St. Louis, R. (2008, September). Acquisition of L2 vocabulary for effective reading: Testing teachers’ classroom practice. The Reading Matrix, 8, 2 http://readingmatrix.com/journal.html