It is tempting to describe a starter as flour, water and yeast, but this leaves out nearly everything interesting. Flour enzymes first cut damaged starch into smaller, fermentable sugars. Yeasts and bacteria then compete for - and sometimes kindly share - these sugars, amino acids, vitamins and other metabolites.

Yeasts make most of the carbon dioxide that inflates the starter and eventually our loaf, while also producing ethanol and aroma compounds. Some heterofermentative lactic-acid bacteria make carbon dioxide too, so every bubble cannot be credited to yeast alone.

Lactic-acid bacteria (LAB) make lactic acid, acetic acid and a collection of flavour precursors, lowering the pH as they go. Acetic-acid bacteria can also turn up; one large survey associated them with stronger vinegary aromas and a slower rise.

A mature starter may contain many detectable organisms, yet it is normally dominated by only a few bacterial species and one or two yeasts. Repeated feeding acts like an endless selection experiment: flour, hydration, temperature, timing and the microbes themselves decide who remains.

So a dramatic rise is useful, but it is not a complete health report. Aroma, acidity, texture, bubble structure and how quickly the full cycle repeats tell us different things about what is actually living in the jar.

We can divide LAB by what comes out after they eat a sugar. Homofermentative LAB send hexoses through glycolysis and make mainly lactic acid. They acidify very efficiently but produce little carbon dioxide through this particular route.

Obligately heterofermentative LAB take the phosphoketolase route instead, producing lactic acid, carbon dioxide and either ethanol or acetate from a hexose. Which final product dominates depends on the sugar, available electron acceptors and the organism involved.

Facultatively heterofermentative LAB sit awkwardly in between - appropriately. They can send hexoses down a mostly lactic route, then switch pathways when pentoses are available. A real starter can contain all three behavioural types at once.

This is why the familiar rule of warm and wet equals mild lactic acid, while cool and stiff equals sharp acetic acid, is far too tidy. Species, flour sugars, oxygen, hydration, temperature, feed ratio and ripening time all change together, and microbes are not obliged to follow a sourdough infographic.

Yeast and LAB are not simply fighting over the same dinner either. One organism may release a vitamin, amino acid or sugar that another cannot obtain efficiently, while acid-tolerant partners occupy slightly different niches. The starter works because competition and cooperation happen at the same time.

As LAB acidify the starter, the falling pH changes the charge and behaviour of gluten proteins. It also wakes up flour's own acid-tolerant proteases, while LAB proteinases and peptidases contribute another route for cutting proteins into smaller pieces.

This is not automatically bad. Controlled proteolysis softens dough and increases extensibility, allowing gas cells to expand instead of meeting a rubber wall. As with most useful processes in bread, the problem is not that it happens, but that it can continue for too long.

Leave an underfed starter warm and acidic and the cutting continues. It becomes progressively thinner, stickier and less capable of trapping carbon dioxide. A starter can therefore rise beautifully, exhaust itself, and collapse into something watery; that final collapse is not evidence of extra strength.

What we see also depends on the flour holding the microbes. White bread flour can build an elastic dome, whereas rye and whole-grain starters may be extremely active without making the same visible gluten balloon. Rise height cannot be compared as if the underlying materials were identical.

If the starter reaches this sad state, dilution is more useful than trying to neutralise it with a base. Keep a small amount of healthy seed, give it a larger fresh feed, hold it at a sensible temperature and repeat until the rise, aroma and texture become predictable again.

In a liquid starter, sugars, acids and enzymes can diffuse more freely. Reactions and feeding commonly move faster, but the looser gluten network makes height and collapse much less reliable as the only maturity test - the microbes may be busy even when the jar refuses to perform.

A stiff starter restricts molecular movement, feels structurally stronger and often takes longer to mature. Changing an established culture from firm to liquid conditions is therefore not just adding water; studies show that the dominant organisms and the metabolites they produce can shift too.

Hydration is still not a magical acidity switch. A stiff starter is not guaranteed to be acetic, and a liquid starter is not guaranteed to be lactic. Species, temperature, flour, feed ratio and ripening time remain inconveniently involved.

If I actually want to test hydration, the flour, seed percentage, temperature and feed interval should stay constant. Otherwise I have changed four experiments at once and learned very little from whichever jar rises higher.

Whole-grain and rye flours arrive with more minerals, vitamins, enzymes and bran-associated microbes than refined white flour. Their higher ash content also gives more buffering capacity, so they can contain substantial total acid without tracing the same pH curve as white flour.

This larger buffet often wakes up a sluggish starter. Controlled studies show that changing flour can alter the bacterial community even when the dominant yeast barely changes, which is a nice reminder that 'active' describes an ecosystem rather than one organism.

Rye supplies accessible carbohydrates and a large amount of water-binding arabinoxylan, but it does not build a wheat-like gluten network. A rye starter can therefore look like a very active paste instead of the elastic dome people expect from Instagram.

Strong white bread flour gives us a clearer gas-retaining structure and a more obvious peak. It may look calmer than a whole-grain feed, but a less dramatic rise does not automatically mean a microbiologically weaker starter.

A mixed feed is often the least annoying compromise: mostly white bread flour for structure and repeatability, with a measured amount of whole wheat or rye for nutrients and flavour.

Feed ratio controls both dilution and the amount of new food. A 1:6:6 feed means one part ripe starter, six parts flour and six parts water. The larger feed carries over less acid and needs longer to mature; the smaller feed peaks sooner and, unsurprisingly, runs out of dinner sooner too.

Temperature controls reaction rate and microbial selection. Warm conditions normally shorten the cycle and cool conditions lengthen it, but a large temperature change can also favour a different population. I would adjust gradually rather than attempting to summon one exact species.

Timing controls how much acid and starvation the culture repeatedly experiences. Feeding near a reliable mature peak maintains activity. Feeding very early can keep diluting a weak population, while feeding hours after collapse selects for organisms that tolerate an acidic, nutrient-poor afterparty.

For a starter that is sharp, runny and collapsing early, I would use a larger feed, refresh it before a prolonged collapse, move it slightly cooler and consider a stronger or marginally drier flour mix.

For one that is sluggish and smells mostly of raw flour, I would move it slightly warmer, temporarily use a smaller feed, add a modest amount of whole grain or rye, and then actually let it mature before feeding it again.

Record the seed ratio, hydration, flour blend, temperature, peak time, aroma and texture. A trend across three cycles says far more than one pH reading or one suspiciously photogenic rise.

The first number most of us look at on a flour bag is protein, and then we decide that more must mean stronger and therefore better. Unfortunately, flour is not that cooperative. Strong white bread flour normally provides a larger glutenin-rich network and tolerates more water or a longer fermentation, but an extremely tenacious flour can resist expansion just as effectively as a weak flour fails to contain it. A lower-protein flour can perform beautifully when its protein quality, hydration and fermentation are matched.

Whole-wheat flour does not simply add 'more wheat'. It brings bran, germ, minerals, lipids, enzymes and a larger microbial inoculum. Bran steals water and physically interrupts the continuity of gluten, while the higher ash content buffers acid and changes how quickly pH appears to fall. Rye relies heavily on starch and water-binding arabinoxylans rather than a wheat-like gluten network; spelt and durum have their own protein compositions and rheology. So the protein percentage is useful, but treating it as a full material specification is a bit like describing coffee only by how dark it looks.

Autolyse sounds far more technical than what we actually do: mix flour and water, then leave it alone before adding levain and salt. But the rest is not empty time. Water moves into starch granules, damaged starch, arabinoxylans and the gluten-forming proteins. This plasticises gliadin and glutenin (giving their chains more mobility), allowing the protein-rich phase to connect with less mechanical work later. Meanwhile, amylases begin releasing smaller carbohydrates from damaged starch and the flour's own proteases quietly begin altering proteins.

So autolyse is not simply free gluten development while I do something else. A sensible rest can reduce mixing and improve extensibility, especially when bran needs time to hydrate. A very long or warm rest can soften weak or highly enzymatic flour before it has enough structure to spare. Add levain and it becomes a fermentolyse because acidification and microbial metabolism have already started; add salt and protein hydration, enzyme activity and microbial activity all change. These are chemically different rests, even if they look identical in a mixing bowl.

We often talk about developing 'gluten strands', which makes the dough sound like a bowl of tiny elastic bands. The reality is messier and much more interesting. Wheat dough is a composite material: swollen starch granules and flour particles sit inside a hydrated protein network. Gliadins are mainly monomeric and help the dough flow and stretch. Glutenins form much larger polymers, joined partly through intermolecular disulfide bonds, and provide elasticity, cohesion and strain hardening.

Hydrogen bonds, hydrophobic interactions, tangled chains and reversible disulfide exchange all contribute, so gluten is a dynamic three-dimensional phase rather than a bundle of independent strings. We need it to be elastic enough to recover but extensible enough to expand. Too much resistance and gas cannot enlarge the cells; too much extensibility without connectivity and the walls thin, merge and leak. This is why protein percentage predicts only part of the result - glutenin composition, flour ageing, damaged starch, enzymes and milling all get a vote too.

Salt looks like the least scientifically exciting ingredient in the formula, yet removing it changes almost everything. Sodium and chloride ions alter electrostatic interactions and the distribution of water around gluten proteins. Up to an optimum concentration, this encourages a more cohesive and resistant network and improves gas retention. Go too far and a strong dough becomes so tenacious that expansion feels like an argument.

Salt also creates osmotic stress for yeast and bacteria, slowing fermentation and changing enzyme activity and metabolite production. Delaying or omitting it therefore does much more than produce bland bread: the dough ferments faster, feels more extensible and loses some protection against proteolytic softening. Around 1.5-2% of flour mass is a common working range, but flour strength, whole-grain content, fermentation time and dietary needs decide where the useful point actually sits.

Mixing is not just making everything homogenous. It distributes water, breaks apart dry protein masses and supplies mechanical energy that stretches, rearranges and reconnects hydrated gluten proteins into one continuous phase. It also traps the first microscopic pockets of air. Fermentation mostly enlarges these existing nuclei rather than creating a perfectly fresh set of bubbles from nothing, so the structure established during mixing follows us all the way to the crumb.

There is an optimum rather than a prize for maximum kneading. Too little energy leaves disconnected protein domains; too much heats and damages the network. Stretch-and-folds and coil folds continue the process more gently: deformation extends gluten around gas cells, then the rest allows stress to relax and new interactions to form. The correct number of folds is therefore rheological, not ceremonial. Weak or very wet dough may need more early reinforcement, while a strong dough can be bullied into becoming unnecessarily tight.

Fermentation sounds simple when reduced to yeast makes gas and bread gets bigger. In reality, it produces carbon dioxide, organic acids, ethanol and dozens of flavour-active metabolites at the same time. Carbon dioxide first dissolves in the watery phase of the dough; once locally supersaturated, it diffuses into existing gas cells. Those cells expand, their walls become thinner and the gluten network is progressively pre-stretched.

Some stretching is useful because the network strain-hardens and resists the growing pressure. But gas production can outrun structural development, causing cells to merge and leak. At the same time, acid changes protein charge and activates pH-dependent cereal proteases, while microbial peptidases keep cutting. Dough can therefore become wonderfully extensible and then, without much ceremony, cross into irreversible weakness. Bulk fermentation ends where useful gas, flavour and remaining load-bearing strength overlap - not at one universal rise percentage or time on a clock.

We give bulk fermentation and final proof different names, but the microbes do not suddenly begin a new reaction after shaping. Metabolism, acidification, gas transfer and protein modification continue throughout. What changes is the geometry and mechanical history. During bulk we can fold, redistribute temperature and move gas around; during final proof, further expansion has to fit inside the architecture we created while shaping.

Temperature makes this more annoying by accelerating several things at once: microbial growth, gas production, diffusion, enzyme activity and dough relaxation. Warm dough reaches the desired volume sooner, but it can also spend its structural reserve sooner. Refrigeration only slows the system; it does not press pause, and the centre of a large dough remains warm long after the outside feels cold.

Yeast does not blow a new set of tiny balloons directly into the dough. Most gas cells begin as air incorporated during mixing. Fermentation produces carbon dioxide, which divides between the dissolved and gaseous phases and diffuses towards those existing cells. Internal pressure then stretches the surrounding gluten-starch matrix.

The cells are not stable little spheres. Small ones can feed larger ones by diffusion, neighbouring walls can rupture and merge, and very thin walls become permeable enough to leak. At first, strain hardening helps the gluten resist this pressure. Once expansion exceeds the network's maximum gas-holding capacity, connectivity falls and the dough becomes porous. This is why an overproofed loaf can contain plenty of fermentation gas and still bake smaller.

Underproofed and overproofed are not simply too little time and too much time; they are different mechanical states. An underproofed dough has not accumulated or redistributed enough gas and often remains excessively elastic. In the oven, pressure is forced towards the score or another weak point, producing violent opening, dense areas and the occasional dramatic tunnel.

Overproofed dough has the opposite issue. Expansion and proteolysis have thinned its cell walls, reduced elastic recovery and increased gas leakage. The useful endpoint lies between them. Volume, surface curvature, bubble distribution, resistance to gentle pressure and recovery after it are all indirect measurements of gas fraction and viscoelastic reserve. The poke test is therefore not a universal truth; hydration, temperature, flour strength and a dry surface can all make the same finger produce a different answer.

Shaping is sometimes described as aligning every gluten strand, as if we were combing molecular hair. A better description is that it redistributes gas, introduces directionality and creates a tension gradient. Stretching and folding orient parts of the gluten-rich phase around gas cells and place the outside layers under more tension. Experiments using repeated sheeting and resting show that deformation followed by relaxation can create a more organised network through non-covalent interactions and disulfide exchange.

The pre-shape first gathers a loose mass into a recognisable geometry. The bench rest then lets the imposed stress relax, so final shaping can occur without tearing it apart. Final shaping adds enough surface tension to resist sideways spread while preserving the extensibility needed for proof and oven spring. Too little gives poor support; too much damages cells or creates a tight shell that resists the expansion we spent hours fermenting for.

The internet's pursuit of open crumb can make touching a bubble feel criminal, but degassing is not automatically a mistake. Gentle shaping removes isolated oversized cells that would otherwise become tunnels and redistributes gas into a more coherent population. Aggressive compression, however, destroys too many nuclei and asks fermentation to rebuild volume inside a network that is already ageing.

How much gas we preserve depends on the bread. A sandwich loaf benefits from deliberate and even degassing; an open-crumb sourdough preserves more irregular cells. In both cases the aim is controlled distribution, not the maximum physically possible amount of trapped gas.

When dough enters the oven, several expansion mechanisms arrive at once. Existing gases warm, dissolved carbon dioxide becomes less soluble, water evaporates into the cells and the microbes continue briefly before heat finally kills them. Cell pressure rises and the loaf springs, which makes the oven look far more magical than it is.

Expansion stops when the network cannot stretch further, gas finds an escape route or the surrounding starch-protein matrix sets. Heating rate therefore matters. A rigid surface formed too early constrains the loaf and tears away from the softer interior; a structurally exhausted dough may expand for a moment and then collapse before setting. The oven can exaggerate the structure we built, but it cannot recreate strength that fermentation has already digested.

At some point, dough has to stop behaving like dough. Starch granules absorb water, swell and lose their crystalline order over a temperature range controlled by flour and water availability. Starch polymers leak out and form a continuous gel around the gas cells. At the same time, gluten proteins denature and polymerise, with heat-induced changes in connectivity and strain hardening becoming especially important above roughly 70°C.

Together these changes turn a deformable foam into a solid porous crumb. Gelatinisation competes for water while evaporation consumes a large amount of latent heat, so the inside of a loaf approaches boiling temperature gradually instead of obediently matching a 240°C oven. The finished crumb is supported by both gelatinised starch and a thermally set protein network.

Steam seems like a strange way to make something crisp. It first condenses on the comparatively cold dough surface, transferring latent heat and maintaining a thin mobile layer of water. This delays desiccation and keeps the skin extensible while the loaf expands. It also gelatinises more surface starch, producing gloss and changing the permeability and eventual fracture of the crust.

More steam is not automatically better, because chemistry refuses to reward enthusiasm alone. Too much condensation changes the heating rate, delays browning and can produce a softer or less permeable surface. The useful amount depends on oven volume, venting, dough maturity and loaf size. A Dutch oven achieves the same early effect by trapping water released from the dough itself rather than continuously injecting more.

The paradox disappears once we separate the bake into a sequence. Early water keeps the surface deformable and lets a thin starch-rich layer gelatinise. Later, we vent the oven or remove the lid; evaporation overtakes the supply of liquid water from the crumb and a drying front moves inward. The surface becomes a glassy, brittle network of denatured protein and partly gelatinised starch.

Crust thickness is therefore a heat-and-mass-transfer problem, not simply a colour chart. Higher temperature accelerates evaporation and can thicken the dry zone, while internal moisture keeps diffusing towards the surface. After baking, water moves back from crumb to crust and plasticises the glassy layer. This is why the loaf that crackled magnificently on the counter can become disappointingly soft a few hours later.

A wet surface struggles to brown because evaporation pins its temperature near the wet-bulb plateau. Only after enough water has left can the crust become hot enough for serious browning. Most bread colour comes from the Maillard network between reducing sugars and amino compounds. Caramelisation contributes at sufficiently high local temperatures, but it is not the single explanation people often use for anything brown and delicious.

Fermentation and amylase activity decide how much sugar remains available for these reactions, so browning began long before the loaf entered the oven. At the end, drying and colour have to be balanced against crumb moisture. Venting completes the crust; cooling lets steam pressure and the gelatinised starch-protein matrix redistribute and settle. Cutting immediately interrupts this process and can make a properly baked crumb look gummy - an impressive punishment for impatience.