In response to comments by Mark and Jonathan in the comment box to my last post:
I think of phanopoeia, melopoeia, and logopoeia as concepts that invoke three different forms of image, where "image" is used to define any minimally complex mental structuration: visual, aural, and/or intellectual. The visual is mostly self-explanatory, though more on this in a second. The aural is pretty obvious too: "figures of sound," etc. The intellectual I think of as combining elements of both the visual and the aural, but more importantly, bringing "secondary" reflectiveness into the mix. Before I expand on this, however, let me say a little more about how I think syntax relates to the three terms.
I don't think something counts as an image until it is somehow made complex, composed of two or more parts. The phanopoetic image is the closest to an exception here, since we think of the visual image as capable of being self-contained, unitary: an apple, say. But even an apple is composed of different surfaces: skin, stem, leaf, gradations of light and texture, etc. What is at question is whether, for example, a single pure color constitutes an image--but I'll let that pass for now.
Similarly, a sound-image is never a single tone; it is always some combination of sounds. As you mention, Mark, rhythm is the salient factor. "Squirrel" alone is mostly a visual image, but "furry squirrel" creates a sound image because of its vocalic and trochaic repetitions. A single sound such as "I" or "a" may have complex connotations, but this is more a function of logopoeia than of melopoeia.
And in fact, the logopoetic can generally be seen as starting with some extension of the phano- and/or melopoetic. "Furry squirrel" becomes logopoetic when we think of the coincidence between the repeated er sound and the fact that squirrels are indeed furry, as though the words were consciously cooperating with each other to form the most appropriate possible pattern of alignments and resonances.
For all of these three types of image, then, arrangement or syntax is the enabling condition. If someone wants to quibble and say that an apple is not an arrangement but a unified perceptual fact, that's fine, as the phanopoetic, like the melopoetic, only really becomes interesting when it starts to verge on the logopoetic at any rate. When one considers an apple at the logopoetic level, it is anything but an isolable unit: it calls up worms, the Garden of Eden, Snow White, gifts for teacher, and any number of associative links.
Addendum: a quick search reveals to me that I talked about some of these same concerns in a blog post earlier this year, which I read now as though it had been written by someone else, so spotty my memory has become.