Drop a teaspoon of starter into a glass of water. If it floats, it has enough trapped gas to leaven bread. If it sinks, it doesn't — or it's been knocked flat by stirring. That's the float test, and it answers exactly one question: does this starter hold gas right now? It is not a peak detector, it is not a strength gauge, and it does not work on rye or on bulk dough. Used inside those limits it's a fast, useful gut check. Used outside them it lies to you, confidently.
Here's the mechanism, the three situations where it fails, and what to watch instead.
What the float test measures
A starter near peak is laced with CO2 from yeast metabolism, held inside a wheat-gluten film. Gas plus gluten makes the sample less dense than water, so a small piece floats. Once the starter crashes past peak, the network gives out, the bubbles escape, the paste slumps, and it sinks because it's now denser than water.
So a float is direct evidence of one thing: gas retention. That correlates with active yeast in a wheat starter, which is why the test works at all. It says nothing about acidity, bacterial balance, or whether you've hit the one-hour window where leavening power is highest. A sample can float across a wide span of activity, from "rising hard" to "just tipped over the top." That's the gap the test can't see into.
Where the float test fails
Three scenarios cause almost every wrong call. Read the right column as the instruction, not the float result.
| Scenario | What the float test says | What's actually true | What to do instead |
|---|---|---|---|
| Cold starter straight from the fridge | Floats — looks ready | Cold thickened it and trapped old gas; yeast is still sluggish and won't leaven on schedule | Feed it, let it come to room temperature, and rebuild over one or two cycles before baking — see reviving a sluggish starter |
| High-hydration wheat starter at true peak | Sinks — looks not ready | It's peaked and strong, but the slack, sticky sample smears off the spoon and traps no air going in | Trust the dome and the rise, not the float; bake now before it slides past peak |
| Crashed or 100% rye starter | Floats — looks ready | Rye's pentosan gel holds gas long after the yeast has faded, so even dead-flat rye floats | Don't float-test rye at all; judge it by volume rise and smell |
The pattern: the float reflects gas retention, and gas retention drifts away from yeast activity at the cold end, the wet end, and on rye. Each column tells you to look at something more reliable.
Why rye breaks the test completely
Wheat traps gas in gluten. Rye barely has functional gluten — its structure comes from pentosans, long-chain carbohydrates that swell into a sticky gel and hold bubbles mechanically, regardless of whether the yeast is still working. A rye starter can be hours post-peak, acidic, and out of leavening power, and a spoonful will still bob happily on the water because the gel itself is buoyant.
So on rye the float test isn't merely unreliable, it's measuring the wrong substance. Judge a rye starter by how far it rose against a mark, the small even bubbles across the top, and a clean fruity-sour smell. If you're working with rye regularly, the readiness signals and feeding cadence are covered in the rye sourdough rules.
Why it doesn't work on bulk dough either
People sometimes pinch off a bit of mixed dough mid-bulk and float it to check fermentation. Don't. Salted, developed dough has a different gas structure and a tighter gluten film than a fresh starter sample, and pinching it deflates exactly the gas you're trying to measure. A floating dough scrap tells you almost nothing about whether bulk is finished. The float test is a starter check, full stop — for the dough itself, watch volume rise, the jiggle, and the surface bubbles, the way the bulk fermentation guide describes.
The signals that beat a float test
A float is a binary that snaps on at a fuzzy threshold. Volume is continuous, so it teaches you more. Before you mix, look for at least three of these:
- Rise against a mark. Wrap a rubber band around the jar at feeding height. A wheat starter ready to bake has roughly doubled to tripled. Watching the climb tells you not just if it peaked but where in the curve you are.
- A soft dome. The surface bulges upward. Once it goes flat, then concave, you're past peak and onto the downslope.
- Bubbles throughout. Activity across the top and down the sides of a clear jar, not one lonely pocket.
- Smell. Tangy, yeasty, faintly like beer and yogurt. Sharp vinegar means over-fermented; raw flour-paste means under.
- Texture on the stir. Stretchy and gel-like, not soupy and not stiff. A starter that's gone soupy and stringy has fermented too far.
If four of those agree, you're ready and you can skip the glass of water entirely. If only one or two do, no float result will save the bake.
Timing peak is the real problem
Knowing your starter can peak is easy. Making it peak when you need to mix is the actual skill, and it's arithmetic, not observation. Feed too late and you mix with an underdeveloped levain that drags bulk long past schedule; feed too early and you mix with a sliding, over-acidic starter that ferments fast and sour and weakens the dough.
The lever is the feed ratio, because a bigger dilution simply takes longer to climb. In a 22°C (72°F) kitchen, a 1:2:2 final feed peaks in roughly 5-6 hours; a 1:5:5 stretches to 8-10. The Starter tab on the calculator does this backward for you: enter your mix time, kitchen temperature, and ratio, and it tells you when to feed so peak lands on schedule instead of by luck. For the underlying ratio-versus-timing logic, the starter feeding ratio guide has the full picture.
Common questions
My starter floats but my bread still comes out dense. What gives?
A float only proves gas retention, not enough yeast biomass or correct timing. A cold or stiff starter can float and still be too weak to carry a loaf, and a starter that peaked three hours ago floats while already sliding into over-acidity. Check rise-against-a-mark and dome instead, and make sure peak lines up with your mix time. Persistent density usually traces back to bulk, covered in why your crumb is dense.
Does the float test work on a stiff (low-hydration) starter?
Poorly. A 50-60% hydration starter is dense and gas-trapping by design, so it can float for hours after it has actually crashed. Stiff starters are best read by smell and by how far they domed and then fell, not by floating.
How long after feeding should I do the float test?
Only near expected peak, never right after feeding and never right after stirring — both deflate the sample and give a false sink. Let the starter rise undisturbed to its dome, then pull a gentle, untouched spoonful. Better still, watch it climb to its mark and skip the glass.
Can I float-test commercial yeast or a poolish?
A ripe poolish (flour, water, a pinch of yeast) will float much like a starter, since it's the same gas-in-gluten situation, so as a rough ripeness check it's fine. Straight commercial-yeast doughs ferment so fast that volume and timing are far more practical to watch.
Is there ever a reason to trust a sink over the other signs?
Rarely. A genuine sink on an undisturbed, room-temperature wheat sample that hasn't visibly risen is a real warning. But a sink right after stirring, or from a slack high-hydration sample, or from anything rye, means nothing. When the float disagrees with a good dome and a solid rise, believe the dome.
The float test in a sentence
The float test answers a single narrow question — is there trapped gas in this wheat starter right now — and it's honest only on a room-temperature wheat starter pulled undisturbed near peak. Cold it floats too easily, slack and high-hydration it sinks while perfectly ready, and rye floats no matter what because the pentosan gel carries the gas without the yeast. Mark the jar, watch it dome and rise to a line you can see, and keep the glass of water as a quick confirmation — never as the verdict.