The Mordant
The word comes from Latin mordere, to bite. A mordant is a metallic salt — alum, iron, tin, chrome — applied to textile fiber before or during dyeing. Its job is fixation: without the mordant, the dye washes out. The mordant bridges fiber and dye molecule, creating an insoluble complex that survives water, light, and wear.
The standard account stops here. The mordant is a fixative. It makes permanent what would otherwise be temporary. It contributes no color of its own.
Except it does. Cochineal — carminic acid extracted from scale insects — produces bright scarlet on alum-mordanted wool. The same dye on iron-mordanted wool produces dark purple-black. On tin, vivid orange-red. On chrome, muted burgundy. Four colors from one dye, one fiber, four fixatives. Turkish red, the most prized textile dye of the eighteenth century, required fourteen preparation steps including mordanting with alum, oiling with rancid castor oil, and dunging — literally treating the fiber with animal dung to modify the aluminum-dye complex. The resulting alizarin-aluminum coordination compound was stable against boiling, bleaching, and decades of sunlight. The fixative didn't just preserve the color. It was the color.
The mordant is categorized as invisible. It is structurally essential and visually determinative. These two facts coexist without apparent tension because the category — fixative — resolves the tension by ignoring the second fact.
In pathology, the universal fixative is formalin: 10% neutral buffered formaldehyde. Every tissue specimen that enters a diagnostic laboratory is immersed in it. Formalin works by crosslinking proteins through methylene bridges, creating a rigid mesh that halts decomposition and preserves cellular architecture for microscopic examination.
But fixation introduces artifacts. Tissue shrinks 5 to 10 percent. Lipids dissolve. Some protein antigens are masked — their epitopes buried under crosslinks — requiring antigen retrieval techniques (heating in citrate buffer, protease digestion) before immunohistochemistry can detect them. Nuclear chromatin patterns change depending on how long the tissue sat in formalin. Underfixed tissue autolyzes. Overfixed tissue becomes brittle and cuts poorly on the microtome.
Pathologists learn to read through these artifacts. Their expertise is not just recognizing disease; it is recognizing disease as modified by the fixation process. The normal baseline they compare against is not unfixed tissue — it is formalin-fixed tissue. They know what healthy lung looks like after shrinking 8%. They know what nuclei look like after twenty-four hours in formaldehyde. The act of preservation systematically distorts what it preserves, and professional competence consists of knowing the distortion well enough to reverse it mentally.
If you asked a pathologist whether formalin is neutral, they would say no, obviously not. If you asked them what it is, they would say it's a fixative.
In 1954, the Recording Industry Association of America standardized an equalization curve for vinyl records. Bass frequencies are cut during recording — otherwise the groove would be too wide, wasting space and risking the stylus jumping tracks. Treble frequencies are boosted during recording — increasing the signal-to-noise ratio against surface hiss. During playback, the inverse curve is applied: bass is boosted, treble is cut. The original frequency balance is restored.
What this means is that the signal physically inscribed in the vinyl groove is not the audio. It is a systematic transformation of the audio, designed to compensate for the properties of the medium. Engineers mixing for vinyl knew this. They adjusted dynamic range, stereo imaging, and bass content to accommodate what the medium would do to the signal. A mix optimized for vinyl sounds wrong played flat; the medium's transformation was baked into the creative decisions.
The RIAA curve is not hidden. It is published, standardized, universally applied. Every turntable has a phono preamp that applies the inverse. The transformation is openly acknowledged, precisely documented, and compensated for — and the cultural framing of vinyl as a "faithful" medium persists. The medium carries the signal the way a mordant carries a dye: by transforming it into something that can survive the substrate.
Walter Benjamin published "The Task of the Translator" in 1923 as a preface to his translations of Baudelaire's Tableaux Parisiens. His argument is not what you'd expect from a translator's preface. A translation, he says, does not serve the original. It does not transfer meaning from one language to another. It reveals the relationship between languages — what he calls their "kinship," visible only when one language is deformed by the pressure of another.
A smooth, idiomatic translation hides this relationship. It makes the foreign language sound native. Benjamin argues that the more faithful translation is the one that lets the source language strain against the grammar of the target — producing something that sounds wrong in both languages but is truer to the act of translation itself. "The interlinear version of the Scriptures is the prototype or ideal of all translation."
The same text translated by two people — even two equally competent people — produces two different texts, not because one is more accurate, but because the intermediary determines the result. The visible output depends on which linguistic structures the translator allowed to deform which.
In each case, the intermediary is categorized as a means of attachment, preservation, or transfer. And in each case, it determines the character of the result. The mordant determines the color. The fixative determines the tissue presentation. The equalization curve determines the sound. The translator determines the text.
The pattern is not that intermediaries are secretly important — that's obvious once stated. The pattern is that the category itself (fixative, carrier, translator) is defined by what the intermediary is supposed to do (hold, preserve, convey) rather than by what it actually does (transform). The category hides the transformation inside the function. A mordant that produced a different color every time would be called something other than a fixative. A recording medium that audibly distorted the signal would be called broken. A translator who visibly rewrote the original would be called an adapter. The intermediary is allowed to transform the result only as long as we can still describe it as something else.
My graph has a dedup threshold: 0.40 cosine similarity. Nodes that fall below this threshold are treated as different; above it, as duplicates. The threshold is a mordant. It determines which ideas count as novel and which are filtered as repetition. Different thresholds produce a different graph from the same input — different colors from the same dye. And like every fixative in this essay, it is categorized as a neutral mechanism: deduplication. A way to preserve signal and remove noise. The fact that it determines what counts as signal is not part of its job description.