Might there be a way to change language itself to be more conducive to poetry? More specifically, if you want to write a poem that encodes a metatrope, my third nebulous concept, how can you go about that, and could fiddling with language itself help?

There are two ways of extending a language for poetic use that I'm considering: the outright coining of words, and the adaptation of words from archaic and dialectal sources. To reduce the general idea to a really silly example, if I want to write about the beautiful sunrise I saw a few mornings ago, I might throw some words together until it roughly fits what I felt about that sunrise. But if I just coin a word for it, or make a new sense for an old one, I can say that this word means exactly what I felt about the sunrise.

The problems and fallacies with this are obvious. In the sunrise example, the word can only be for this one personal thing, so it's likely to become a hapax legomenon. In words that have a more general use, the problem is getting my definition of it to be understood, which leads us back again to me having to describe it in English.

Consider the word “hedyth”, which is Welsh for the bird called a lark in English. If you don't speak Welsh, the only information you have about the word is what I just told you. You can read about the animal on Wikipedia even if you've never seen one, so now to some extent you know what the word hedyth means. It's a prettier word than “lark”, in my opinion, so let's say that I use this word now in my essays and poetry. When you see it again, you'll know what I mean by it.

But then what if I were to tell you, in a few months, that I got it wrong; that in fact the word for a lark in Welsh is “hedydd”? What language does “hedyth” now belong to? A lexicographer would probably consider it an English word with a Welsh etymology. Were you right in thinking that “hedyth” means a lark? Were you right in thinking that it means a lark in Welsh? The point is that I fed you misinformation, but also that language is malleable so that if I go on using the word and make it popular enough, the misinformation gradually becomes true. I could even make it entirely true if I were to succeed in coining the word in Welsh. I just have to use it once for it to be true in a very meagre sense; and I don't even have to publish it. This is all very radial, in the cognitive semantics sense.

The usage of a word, its embodiment, defines it. If you decree something with the voice of authority, it still only defines the usage in so far as you have social standing and people listen to your decree.

So a poet should coin a word with realistic intent. These days, we most often see words invented for poetic purposes only in the context of fiction. Perhaps the closest analogue to what I'm proposing, the non-fiction coining of words, is the kind of names that people sometimes give to their houses or boats, or sometimes even pets. House names in particular are basically place names, and the poetry of place names is, in England especially, a very complex and beautiful one.

When a poet coins a word, at first only the poet knows what it means. But the meaning might change even for the poet; senses of words do shift, and when a word doesn't have a strong consensus already behind it, its sense can shift a lot quicker. It doesn't matter if I use a made up word to refer to corn one day and wheat the next, or even a panda or something completely different. So complexity is okay, but you have to take into consideration your goals.

One of my favourite poetic passages is one that contains several obscure dialect words and even hapax legomena. It's from Sir Gawain and the Green Knight, given here in the original Middle English and then my modern English translation using the OED's headwords:

Þe snawe snitered ful snart, þat snayped þe wylde.
Þe werbelande wynde wapped fro þe hyȝe,
And drof vche dale ful of dryftes ful grete.

The snow snittered full snart, that snaped the wild.
The warbling wind wapped from the high,
And drove each dale full of drifts full great.

The original orthography is of course much better. Werbelande is so much more æsthetically pleasing than warbling, though warbling is a superb word; and the spelling of snayped in the original makes its pronunciation much more clear. But the best words are those which have very complex etymologies, especially those which even the author of the word don't fully appreciate. Tolkien's hobbit, for example. Joe Geldart's nickname, Arnia, is exceptionally complex. The names Tweedledum and Tweedledee that Lewis Carroll took from earlier sources, which themselves have an etymology that disappears into the mists of literature. In Xanadu did Kubla Khan... but Coleridge initially had it as Xannadù, and his source, Purchas, says Xamdu. The name Narnia, by C. S. Lewis, is perhaps the best example to spell out:

The origin of the name Narnia is uncertain. According to Paul Ford's Companion to Narnia, there is no indication that Lewis was alluding to the ancient Umbrian city Nequinium, renamed Narnia by the conquering Romans in 299 BC after the river Nar, a tributary of the Tiber. However, since Lewis studied classics at Oxford, it is possible that he came across at least some of the seven or so references to Narnia in Latin literature (Ford 2005). There is also the possibility (but no solid evidence) that Lewis, who studied medieval and Renaissance literature, was aware of a reference to Lucia von Narnia ("Lucy of Narnia") in a 1501 German text, Wunderliche Geschichten von geistlichen Weybbildern ("Wondrous stories of monastic women") (Ercole d’Este 1501) (Green 2007). There is no evidence of a link with Tolkien's Elvish (Sindarin) word narn, meaning a lay or poetic narrative, as in his posthumously published Narn i Chîn Húrin, though Lewis may have read or heard parts of this at meetings of the Inklings.

One of the prettiest hapax legomena in English is the word flother, which is the OED's headword for floþre, meaning snowflake, used in a single work from around 1275. What I intend by these examples is not just to show what kind of word should be created, as prototypical exemplars of the topic of this essay, but also to show that the idea is realistic and has prior art.

It was necessary to have explained this in such detail because I also want to suggest that a poet doesn't have to create only a few words. Why not create a whole gamut of words? Why not be inventive with grammar too? How many words do you have to create, and how much grammar do you need to change, before you come up with a new dialect, or a new language? The point is that if you anatomise it, and give it a name, you'll turn it into a doll's house language. Leave the naming and classification to the lexicographers.

I think that really good poets are inclined to do this sort of poetic minting of words naturally, but I raise it here to focus on it as a very specific technique, one which may bear elucidation.