Why Paint Floats Instead of Sinking in Paper Marbling

Nothing about marbling paint makes it want to float. It is pigment and polymer carried in water, and both of those are denser than the water carrying them — put that paint on plain water and it mostly disperses and settles, which is why acrylics need a thickened bath to float at all. Sprinkled onto a marbling tray, the same paint sits on top and opens into a flat disc several inches across.

The tell is drop size. Fine droplets flicked off a whisk float; the same paint squeezed from an eyedropper makes fat drops that punch through and sink. If buoyancy were doing the work, a large drop would float exactly as well as a small one. It doesn't, which means the paint is not really floating — it is being held up.

Three things do the holding, and all three are adjustable: the thickness of the bath, the surface tension at the boundary between paint and bath, and a surfactant you add to the paint on purpose. Get them into balance and the colour spreads and stays. Miss on any one and the paint sinks, beads up, or washes off the paper afterwards — which is why a beginner paper marbling setup is less a shopping list than a set of conditions to hit.

The bath is called size, and the common description of it as a gel is wrong in a way that matters. Marblers use lambda carrageenan, the one form of the seaweed extract that does not gel. Kappa and iota carrageenan set into solids in the presence of potassium or calcium ions; lambda carries more sulfate groups, never forms the helices the other two do, and so thickens water without ever setting. The result is a viscous liquid about the consistency of syrup.

That distinction is the whole trick. A gel would hold the paint beautifully and then refuse to let a comb through. A viscous liquid resists slowly — fast enough that a small drop cannot sink through it before the paper goes down, loose enough that a comb parts it and it closes again behind. Lambda carrageenan is also shear-thinning: it turns runnier exactly where a tool is moving through it, and stays thick everywhere else.

Depth barely matters. Half an inch of size is enough to marble on and many marblers use less, because the floating film only ever interacts with the top of the bath. Water chemistry matters enormously. Hard, heavily mineralised water stops carrageenan thickening properly, as dissolved calcium and magnesium interfere with it; marblers in hard-water areas mix their size with distilled water or add sodium hexametaphosphate to trap the metal ions. A tray mixed with the wrong tap water gives pale, sunken, disappointing prints that look exactly like bad technique. It is the same kind of invisible chemical precondition that decides why blueprints came out blue rather than any other colour.

What happens when you take the size away

Japanese suminagashi runs the same experiment with the thickener removed: plain water, no size at all. Sumi ink floats on bare water held only by surface tension, and a brush loaded with a surfactant, touched to the ink, pushes it outward into concentric rings. It works, but the margins are brutally thin — press a brush too far in, the surface breaks, and the ink sinks or mixes in. Thickened size buys the forgiveness that lets marblers comb elaborate, stable patterns instead of only rings.

Thickened water alone will hold a drop of paint, but it holds it as a drop. To make colour open out across the surface you have to lower the paint's surface tension below the bath's, and that is the job of gall — historically ox gall, now usually a synthetic surfactant sold under the same name. It goes into the paint, never into the tray.

Gall is powerful out of all proportion to how little you use. Marblers add it one or two drops at a time to a small working jar, test on the surface, and add another only if needed. Too little and the drop keeps its surface tension, stays a bead and sinks. Too much and the colour spreads so far it goes pale and crowds everything else off the tray — and that cannot be undone, because nothing added to paint slows its spread back down. Some colours want a couple of drops, some want none, and the behaviour shifts with temperature and humidity, so marblers test each colour on the size at the start of a session instead of trusting last week's settings.

Surfactant balance is also how marblers make one colour dominate: a little extra gall in one jar and that colour pushes the others aside as it spreads. The property that looks like a defect is the control surface. Film photographers meet the same inversion when they learn why film negatives look orange — the odd-looking feature is the deliberate mechanism, not a fault.

Dragging a comb through a tray of wet paint ought to mix it. It doesn't, because there is very little left to mix. Each drop has already spread into a film far thinner than the drop it came from, floating as a discrete sheet pinned at the boundary between paint and size. The comb draws those sheets into new shapes instead of stirring them together: the size is viscous enough that the films get pushed around rather than folded into one another. Marblers keep the bath still and avoid abrupt movements for the same reason — sudden motion disturbs the pattern and raises bubbles.

Sequence does the rest. Every new round of drops lands on a surface already covered, and the fresh paint spreads by pushing the existing films outward and compressing them. Colours laid down early end up squeezed into thin dark veins between the later ones, which is why marblers put black down first — it outlines everything added after it and gives the sheet its depth. A tray that prints washed out often simply has too little paint on it, and the films were never crowded enough to concentrate.

Two conditions break this. Size that has spoiled loses viscosity, and the signature is unmistakable: colour shoots across the whole tray instead of settling into a disc. Size straight out of the fridge is the opposite problem — cold size lets colour spread less, then bleeds and drips after printing, so marblers let the tray reach room temperature before starting. Both faults are read off the surface in seconds, much as a cyanotype print tells you about its exposure before the wash is finished.

At this point the pattern exists only as a floating film. Laying paper onto it transfers the colour at once — the sheet lifts away carrying the design, and whatever little is left behind gets skimmed off before the next print. But lay down untreated paper and the pattern comes off again the moment you rinse it. Marbling is a transfer process, and unlike a beginner linocut setup, where pressure drives ink into the sheet, nothing here is pressing anything. The bond has to be chemical.

That is what alum does. Paper and fabric are sponged or soaked in a solution of aluminium sulphate and dried before marbling, and the alum left in the fibres is what the pigment fixes to. Skip it and almost all the colour rinses away. Miss a patch and you see exactly where: floating paint will not wet an uncoated area, so it retreats into droplets and leaves a hole in the design matching the gap in the coating. Alum does not keep either — treated paper left exposed to air for more than a few days oxidises, loses its alum, and has to be done again.

The counterintuitive part is that too much alum fails the same way as none at all. Overdose the sheet and the paint bonds to the alum rather than through it to the fibre, and the colour lifts off in sheets. That is the complaint practitioners hear most from people whose prints look perfect in the tray and ruin in the rinse, and the fix runs in the direction nobody expects: use less. The fabric recommendation that circulated for years, two tablespoons per gallon, was eventually cut to two teaspoons by the people who published it, and marblers working thin silk or polyester go down to half a teaspoon, because low-absorbency material needs barely any. Paper recipes run much stronger than fabric ones for the same reason. It is the same narrow-window problem as matching cyanotype exposure to sunlight or a UV lamp: the variable has a right value, and both sides of it fail.