Overnight fermentation: idli and dosa

A bowl of idli batter left on a kitchen counter overnight does something that looks like bread dough rising and is not. Bread rises because one organism, yeast, eats sugar and produces carbon dioxide. Idli batter rises because two different populations take turns, and the first one changes the conditions that the second one needs. Understanding that relay explains every instruction in the recipe, including the ones that sound like folklore.

Who does what, and when

The first phase belongs to lactic acid bacteria, chiefly Leuconostoc mesenteroides, which is heterofermentative: it converts sugars into lactic acid, carbon dioxide, ethanol and a polysaccharide called dextran. Three of those four matter. The acid drops the pH from around 6 to around 4.5 over six to eight hours. The carbon dioxide begins to leaven. The dextran is a viscous polysaccharide that thickens the batter and helps it retain the gas it produces — it is the reason a properly fermented batter looks ropey and sticky rather than watery.

Around pH 4.5 the bacteria inhibit themselves; the acid they made is now hostile to them. This is when the yeasts — Saccharomyces cerevisiae and various Candida species, present on the grain and in the air — take over. They are acid-tolerant and they continue producing carbon dioxide, and they also produce a little ethanol, which contributes to the aroma. The batter keeps rising for a few more hours, then peaks and begins to fall.

Two stages, two different organisms, one after the other. If you skip the first — by acidifying the batter with lemon juice or citric acid, a shortcut people try — you get sourness and no structure, because the dextran was never made and the gas was never generated at the right time.

The ratio is a food supply

Rice and urad dal are not interchangeable. Rice is starch: it gelatinises when cooked and provides the bulk and the sweetness. Urad supplies protein, oligosaccharides and, critically, the mucilage that both feeds the bacteria and holds the foam. A batter with too little urad will ferment, smell right and collapse on the plate; a batter with too much is dense, beany and heavy.

The usual range is three to four parts rice to one part urad by volume. Parboiled rice is common in Tamil Nadu and gives a softer, whiter idli because the starch has already been partially gelatinised and absorbs water differently. Raw rice gives a crisper dosa. The same two ingredients, milled and fermented the same way, produce a soft steamed cake or a thin crisp depending on the rice and the water.

Why baking soda cannot rescue a dead batter

This is the question that separates a cook who understands the process from one who has memorised it. Baking soda, sodium bicarbonate, produces carbon dioxide in two situations: when it meets an acid, and when it is heated. In a batter that has not fermented there is no acid, so nothing happens at room temperature. In a batter that has fermented, there is plenty of acid, so the soda neutralises it — which removes the sourness, removes the preservative effect, and leaves the batter flatter and yellower, because the alkaline conditions degrade pigments. You have traded flavour and shelf life for a brief puff.

The real failure modes are physical, not chemical. A room below 25 °C slows the bacteria to the point where nothing happens overnight. Chlorinated tap water kills the inoculum; so does very hot grinding, because a wet grinder run too fast can push the batter past 45 °C. Salt added at the start draws water out of the cells osmotically and slows the whole succession — which is why it goes in at the end, just before cooking. None of these are fixable after the fact.

Grinding is a process, not a step

A traditional stone wet grinder turns slowly, shears the soaked grain gently, and aerates the batter as it goes. A high-speed blender does the opposite: it heats the batter and whips in large air bubbles that collapse later. The difference is visible in the finished idli — a stone-ground batter gives a fine, even crumb; a blended one gives a coarse, slightly gummy texture with large holes.

Which explains the float test. A drop of properly fermented batter placed in a glass of water floats. This is not a superstition and it is not about sourness; it is a density measurement. The batter floats when it has retained enough gas to be less dense than water, which is exactly the property you need it to have. It is a cheap instrument for a physical property, and it works.

One ecology, many geometries

The same fermented batter, handled differently, becomes half a dozen foods. Idli is steamed in a shallow mould: the steam sets the starch around the gas bubbles before they can escape. Dosa is spread thin on a hot griddle: the water flashes off, the bubbles are destroyed, and what remains is a crisp sheet. Uttapam is the same batter poured thick with vegetables on top. Appam, in Kerala, swaps in coconut and toddy or yeast and is fermented in a different vessel to give a bowl-shaped pancake with a lace edge.

Dhokla, in Gujarat, is the same idea with a different substrate — chickpea flour instead of rice and urad — and a different organism profile. It is worth noting because it shows the technique is not South Indian; it is what happens wherever a starchy batter is left warm overnight in a hot climate. The subcontinent has dozens of these ferments, from dosa to selroti to fermented bamboo shoot, and they share a logic rather than a recipe.

The origin question, briefly

Where idli came from is genuinely unsettled. There is a widely repeated theory that it derives from an Indonesian fermented cake, brought by traders or by cooks in the medieval period; there are Kannada references from around the tenth century to a steamed cake; and there is a thirteenth-century record that some read as describing something similar. The evidence is thin in every direction and the argument is often conducted with more confidence than the sources support.

What is not in dispute is that the food requires two separate technologies to meet: a fermentation tradition and a steaming tradition. Steaming is the older and more widespread of the two in Asia. Once a culture has both a warm climate and a steamer, this food is nearly inevitable, which is why several regions have arrived at something like it independently.

How to tell whether it worked

Smell first: it should be sour and clean, like yoghurt, not sharp like vinegar and not sweet like bread. Look at the surface: fine bubbles, a slight sheen, and a batter that pulls up in strings when you lift a spoon. Then the float test. Then cook one idli as a trial before committing the whole batch — a batter that has over-fermented will smell of acetone and taste thin, and no amount of steaming will bring it back.

And do not be discouraged by a failed winter batch. The organisms are doing exactly what they do; the room is simply the wrong temperature. The fix is a warm cupboard or an oven with the light on, not an ingredient. This is a biological process with a narrow window, and the whole recipe is a set of instructions for staying inside it.

This is entry 05 of Food in About India. The pillar's argument — grain and fat, not spice blends — is set out on the pillar page.