(I'm going to take a step back and give an introduction to Barfield's thought as I see it. I should have taken the hint because he starts each of his classic books with these examples, and takes a bit of time to develop them till he takes off. I'm struggling with the attempt to accurately represent his thought and argument here, so I figure I'll try with this as an intro. I'll go back to part 3 of this series after this.)
There are two things you need to think about with every sentence as you read something by Owen Barfield: words and rainbows.
Words are where he starts. Poetic Diction (reviewed here) was his first book and was basically an expansion of his Honors Project, and this was about what it sounds like: the words chosen for poetry. But because Barfield believes that all words contain something of the history of their original creation in them, talking about words is talking about history, so that talking about where they came from is talking about where they are going. So on the one hand, it's just words, but on the other hand, everything is about words. This resonates with what we know about us being built from words of DNA, and our ability to look into the DNA to discern the history of the DNA words. Exciting stuff for a scientist.
Barfield observed that we originally had fewer words with broader meanings that would overlap and poetically reverbate. But as time passes, lives are lived and knowledge is increased, the meaning of the words becomes more specific, "fossilized" or "calcified", and the multiple meanings are lost while more specific "scientific" meanings are gained. Not that there's anything wrong with that! The way forward according to Barfield is not to limit our vocabularies and somehow forget the specific meanings, but rather to put our words back together in new and poetic ways. This may sound like sitting around writing poems, but it isn't. Verlyn Flieger makes a strong case in Splintered Light (reviewed here) that J.R.R. Tolkien wrote Middle Earth (starting from words/languages not geography, and including various prominent songs and poems in it!) as his way of putting fragmented things back together, just like his friend Barfield suggested. It is possible that one reason people have found Middle Earth so inspiring is because Tolkien meticulously and creatively captured Barfield's ideas in a story, possibly the only way it can be fully captured.
The other thing you need to constantly keep in mind is rainbows. Barfield's other "great" book, Saving the Appearances (review here) starts with the sentence "Look at a rainbow." and then asks, "Is it really there?"
The rainbow does not exist except as a combination of light, rain, and the observer. It moves with the observer, but it's not a hallucination, it is a true observation of the outside world. You really see it -- it is really there. (Interestingly enough, a recent Scientific American article talked about the rainbow halos seen around shadows, called "glories", and these too are tied to the observer. There was no question in that article of whether the "glory" is really there.)
Barfield's key move is to say that if a friend there with you also sees the rainbow, you talk about it as a real thing: "Hey, look at that rainbow!". It is really there as a "collective representation" that you both see and process in your minds and talk about with your words. Then he extends this to look at a tree nearby. This, too, he notes, is a collective representation. We can understand that from previous experiments that the tree is made of cells and the cells have photosynthetic pigments in the leaves and that can explain all sorts of things about the tree, why it changes color in the fall, etc. But fundamentally, what we're talking about when we talk about the particles that make up the tree is a model or hypothesis of the things we can't see, collective representations on top of collective representations.
When I type "rainbow" or "tree" on my computer, and when you read those words, we can be pretty sure we're talking about the same thing and are using the same "collective representation" for it, and likewise for the scientific models of atoms and photosynthesis that work very well to describe the tree or why the rainbow only shows up when the sun is behind you. They are true and consistent collective representations, invoked by the words.
Barfield's basic point is that the observer is part of the whole scheme for the rainbow and for the tree. And his apparent conflict with science comes in when science tries to remove the observer from the picture and Barfield says "you can't do that!". You are part of the system, whether you're observing a rainbow or a tree, and whether you're observing it at the macro- or micro-scopic level. You're making observations, putting them together, and talking about them with words, or "collective representations" as he likes to call them because he wants to group thoughts and words together.
(Students of 20th-century physics may realize that the involvement of the observer with the experiment is actually one of the fundamental conclusions of 20th-century physics! Barfield's ideas actually cohere well with some physics experiments.)
So as I write about Barfield, remember the words, and remember the rainbow. These are all human observations described with human words.
Now we may be ready to start part 1.