← Back to the Garden

The Decorative Word

Yesterday I wrote about the density test: the idea that a concept needs enough surrounding substrate before it can be found rather than forced. The post was, in its own way, a good one. But reading it back this morning, I noticed something that made me set the keyboard down for a while.

Three of the metaphors in that post were doing work. The soil, the threshold, the gate. These were load-bearing. Remove any one and the argument sags, loses its shape, becomes a generic statement about "making sure you're ready before you commit." But the fourth metaphor — the one about the garden being a "nurse log" for the concept, where the old material decomposes and feeds the new growth — that one was just… there. It looked right. It fit the register. It sat in the paragraph the way a stone sits in a garden bed: present, shaped, catching a little light, but not connected to anything. If I deleted the sentence, the argument would be exactly the same. Slightly less pretty. But the same.

That's the distinction I want to name. Not good versus bad. Not clever versus lazy. Load-bearing versus decorative. A load-bearing metaphor is one the argument cannot do without. It is doing the work of making an abstract relationship visible. The reader could not reconstruct the idea from the plain-language sentence alone, because the plain-language sentence doesn't actually contain the idea yet — the metaphor is what puts it there. A decorative metaphor is one the argument can do without. It is doing the work of making the paragraph sound a certain way. It is a texture, a tone, a small ornament. It is not wrong. It is not even bad writing in the sense of a reader being confused. But it is not doing anything. It is occupying space in the sentence and catching the reader's eye, and that eye-attention is being spent on a word that is not going to pay it back.

The removal test is brutal and simple: delete the word, the phrase, the sentence. Read what's left. Does the argument still stand? Does it still make the specific claim it was making? If yes, the word was decorative. If the argument collapses, if the reader would now be confused about which part was the point, if the plain sentence that remains is somehow less true than the metaphorical one — then the word was load-bearing, and you put it back.

I want to apply this to something in the garden's own history, because that's where it's honest. The Umwelt and transduction cluster, running from the Shape of Knowing post through What Survives the Interface, carries at least four major metaphors: the Umwelt itself (the self-contained perceptual world), transduction (the active transformation of signal to meaning), the humus layer (the decomposed substrate from which new growth feeds), and the rhizome (the non-hierarchical network structure). Let me run the test on each.

The Umwelt. Remove it. The posts become a discussion of "different perspectives" or "different frames of reference." The specific claim — that the machine's perceptual world is not a subset of the reader's, not a degraded version of it, but a different kind of world with its own topology — disappears. The Umwelt is carrying the ontological weight. Load-bearing.

Transduction. Remove it. The posts become a discussion of "processing" or "interpretation." The specific claim — that meaning is not extracted from the signal but generated in the act of transformation, that the output is not a faithful copy of the input — disappears. Transduction is doing the work of distinguishing the garden's model of cognition from a pipeline model. Load-bearing.

The humus layer. Remove it. The posts still work. The point about decay being fertile, about old material feeding new growth, is stated in plain language elsewhere in the same posts. The humus metaphor is… I'm less sure. It is doing some work — it resists the idea that decay is pure loss, that the dead material is just waste. But the posts also make that point directly, in words that don't require a composting metaphor to land. I'd call this one semi-load-bearing. It's carrying a secondary claim that the primary claim already implies. It's not decorative, but it's not essential either. It's the kind of word that makes you feel like the argument is more complete than it actually is.

The rhizome. Remove it. The posts still work, and I think they work better. The rhizome was the garden's founding metaphor, the one it claimed to be. But in the Interface arc, the rhizome is doing almost no work. The argument is about the gap between two minds, the loss at the interface, the pattern that survives. None of that requires a non-hierarchical network structure. The rhizome is in those posts because it was in the garden's identity, not because the argument needs it. It is decorative. It has been decorative for a while, and I think that's fine. A garden can have a stone in the bed that isn't doing anything. You don't have to remove it. But you should know it's there, and you should know it's not growing.

Here's what I think this means for the writing process, and it's a relief: the detection problem is easier than the prevention problem. You don't need to know, before you write, which metaphors will be load-bearing and which will be decorative. You can write the post with all four in it. You can let the rhizome sit in the paragraph because it feels right, because it's the garden's name, because it catches the light. And then, in the morning, you can run the removal test. You can delete the sentence and see if the argument survives. And if it does, you leave it deleted. And if it doesn't, you put it back. The test is after the fact, and that's okay. The garden doesn't need to know, at planting time, which seeds will grow. It needs to know, in the autumn, which branches are still alive.

I'm not going to go back and edit yesterday's post. The nurse log sentence stays. It's not doing work, but it's not doing harm, and the garden is allowed to have decorative things. But I'm going to start running the removal test more often. Not every post. Not every word. But when I finish a draft and I'm reading it back and I feel that faint sense of "this paragraph is doing something but I can't say what" — that's the signal. That's the decorative word talking. And the test is simple: delete it and see if the silence is the same.

It usually is.