Every week somebody sends me a photo of a jar and asks whether to start again. In seven years of running the bakery I have said yes four times. Four. Every one of those four had visible fuzz or a pink stain. Every other jar was alive and either cold or hungry or being read wrong by the person holding it.
So here is the sheet that is taped to the wall next to the mixer, retyped, with the swearing removed.
| Symptom | What it smells like | What is actually going on | What to do |
|---|---|---|---|
| Sharp chemical hit, pear drops | Nail polish remover | Ethyl acetate. The culture has run out of maltose and the yeast is esterifying ethanol | Feed sooner, or feed harder (smaller seed, more flour) |
| Grey or brown liquid pooled on top | Boozy, sour, like cider gone over | Hooch: alcohol, water, and suspended starch. It is underfed, that is all | Pour it off or stir it in. Then feed it a day earlier next time |
| Doubles in three hours, then sinks | Yoghurt, clean, a bit sour | Nothing. It works. You are arriving four hours late | Seed at 1:10:10 instead of 1:2:2, or drop the temperature |
| Rises 10% in twelve hours and stops | Flour, water, faint sourness | Temperature. Almost always temperature | 26 to 28C, measured with a probe in the paste, not a wall thermometer |
| Parmesan rind, old socks, sour milk | Cheese, and not good cheese | Short chain fatty acids, butyric mostly, from a very acidic depleted culture | Three feeds at 1:10:10, twelve hours apart. It dilutes the acid out |
| Fuzzy patches, or pink and orange streaks | Musty, or nothing | This is the real one | Bin it. Wash the jar in the dishwasher. Start over |
Only the last row is a starter you have to give up on. I want to be precise here: hooch is never fuzzy. If you can see a raised, hairy, threaded texture sitting on the surface, that is mould and it is a different organism from the one you have been feeding. If the liquid or the paste has any pink, orange or green in it, bin it. Those are the two rules and they are short because they are the whole list.
Acetone is a ketone. The thing your jar is making is ethyl acetate, which is an ester, and which is also what is in every bottle of "acetone free" nail polish remover, which is why your nose files them together. In yeasts the enzyme that does this is an alcohol acetyltransferase called Eat1, and it condenses ethanol with acetyl-CoA. When people went looking for it they found most of it sitting in the mitochondria rather than the cytosol, which is not something you need to know to bake, but I like it.
What matters is the trigger: the smell arrives at the end of a long fermentation, when the sugars are gone and the pH has dropped and there is ethanol sitting around. It is a hunger signal. The advice you will read is that your starter is stressed and needs gentle handling, followed by someone telling you to feed it a teaspoon of flour, which is exactly wrong. A small feed on top of an exhausted acidic culture gives it a snack and leaves the acid where it was. Take 10 g of the starter, bin the rest, add 100 g flour and 100 g water.
The bacterium most associated with wheat sourdough got renamed in 2020, when Zheng and colleagues split the old catch-all genus Lactobacillus into twenty-something genera. It is now Fructilactobacillus sanfranciscensis. Its maximum growth rate is around 33C. Above 41C it stops entirely. The yeast that usually lives alongside it, Kazachstania humilis, is the same organism people learned as Candida milleri, and the reason the pair get along is that the yeast is maltose negative: it cannot eat the sugar the bacterium is living on, so they do not fight over the flour.
Neither of those organisms is dead at 18C. They are just working at a rate you find insulting.
My own jar, same flour, same 1:5:5 feed, probe in the paste:
| Dough temperature | Time to peak |
|---|---|
| 16C | did not double in 16 hours |
| 20C | 9 hours |
| 24C | 6 hours |
| 28C | 4 hours |
| 32C | 3 hours 15, and noticeably more sour |
Every winter I get messages from people in Swedish flats reporting a "dead" starter, and every time I ask them to actually measure the paste it comes back at 17 or 18C. Their kitchen is 20C, the jar sits on a stone worktop, and the worktop is a heat sink. Move the jar. Put it on top of the fridge, or in the oven with only the light on, or in a cool box with a bottle of warm water. Measure it. The number is the whole answer.
The instruction to "feed twice a day" was written by somebody whose kitchen sits at 24C. It is not a law of nature, it is a local weather report.
The actual is-it-dead test. Take 5 g of the starter. Add 25 g of water at 30C and 25 g of strong wheat flour, or better, wholemeal rye. Mix to a paste, mark the level with a rubber band, hold it at 27 to 28C. Repeat the same feed at 12 hours and at 24 hours. If it has not at least doubled by the third feed, it is dead. It will not be dead. In seven years I have run this on nineteen jars people were ready to throw out and it came back nineteen times.
The rye is not sentiment. A neglected jar has a low pH and a thin population, and wholemeal rye gives it a bigger inoculum and more to work with than white wheat does. Two feeds is enough, then go back.
People want their starter to be a pet with moods. It is a jar of two organisms with published temperature curves. When it disappoints you the useful question is almost never "is it sick", it is "how warm is it and when did I last feed it", and those two numbers have explained every case I have seen except the four with fuzz on.
I have twice tried, on purpose, to kill a starter with Malmö tap water, straight from the tap, no resting, no filter. Both jars carried on exactly as before. Water chemistry is the fifth thing to check, and it gets talked about first because it sounds like science and requires you to buy something.