There are plenty of recipes out there for sourdough starters, and there's a mountain of literature that you can read to understand how they work. Almost all of these resources try to answer one central question: Is there an optimal way to make a starter? And if so, how?
The answers aren't simple. There are many variables at play, from the type of liquid used and hydration levels to fermenting temperatures—not to mention all the uncertainty introduced by the unpredictable world of wild yeasts and bacteria that make each and every starter unique. You could spend a lifetime testing all the variables, and you still wouldn’t arrive at any clear-cut conclusions.
But that doesn't mean we shouldn't try to shed some light on the matter. I decided to focus on one of the first decisions a person has to make when setting out to create their own sourdough starter: What flour should you use?
If you don’t feel like reading on, here's the one-line conclusion: There’s no right answer. Many flours work well. But it’s worth noting that different flours can have distinct effects on the development of a starter. And if different flours can lead to different starters, then different starters can lead to differences in baking.
My plan was to make several starters using different flours or blends of flours. But first I needed some expert guidance in the process. I'm not an inexperienced sourdough baker—I've been doing it for the past eight years (though I’ve only truly become comfortable with it in the past two or three). My current starter, Ryan, has lived with me since my time as a restaurant cook in New York several years ago; I grew him from just a spoonful of my pastry chef’s old starter. Ryan is a hungry and active culture, and he's been responsible for some delicious breads over the years. But Ryan’s path to maturity was paved with trial and error. I wasted dozens of bags of flour feeding him, and sometimes lapsed into periods of neglect, careless feeding, and improper storage.
Just like a parent raising their second (or third, or fourth, or fifth) child, I was determined to do it right this time. So I reached out to Kristen Dennis, a former scientist and sourdough home baker based in Chicago, who has over 200,000 followers on her Full Proof Baking Instagram account. Her mesmerizing breads are consistently masterful specimens—tall, with blistered exteriors and outrageously open crumbs. “I’ve gotten so many people who just started their starter this week,” she told me a few weeks ago. “It’s on day four, day five, day six, they’re saying, Oh my god, there’s no substantial activity. What’s wrong? That’s fine, that’s totally normal."
Kristen keeps a number of guides and video tutorials online—ranging from starter maintenance to high hydration loaves—and she's remarkably consistent in providing a level of explicit detail rarely found elsewhere. Making the transition from her life as a life sciences PhD was natural. “My whole background is in taking notes and writing—details, details...changing one variable at a time.”
Kristen helped me devise a methodical testing structure, and provided advice throughout my testing process. Here's what we settled on.
Primary Testing Variable: Choice of flours
When it comes to the flour you use in a sourdough starter, there are several options to choose from, ranging from wheat flour to cereal grains. Here are the flours and flour mixes I chose for testing, which reflect some of the most commonly available options in stores.
I settled on five samples, each fed exclusively a single flour type: 100% all-purpose flour; 100% whole wheat flour; 100% rye flour; a 50/50 blend of whole wheat flour and all-purpose flour; and a 50/50 blend of rye flour and all-purpose flour.
- Unbleached All-Purpose Flour: All-purpose flour is admittedly a broad category, but it’s a popular option for sourdough starters due to its wide availability. “I know Trevor Wilson [another master baker and respected sourdough philosopher] uses an all-white flour starter,” says Kristen. “King Arthur Flour sells a fresh starter and it’s made using their all-purpose flour, too.” The protein content of all-purpose flour varies between 9% and 12%, depending on brand, and while all-purpose flour can be bleached or unbleached, I decided against using bleached flour because the bleaching process depletes the potential population of microbes in flour, statistically making inoculation slower and more difficult—meaning a starter likely wouldn’t develop as fast. So for the purposes of my test, I chose King Arthur's all-purpose flour, a relatively high-protein (11.7%), unbleached white flour.
- Unbleached Bread Flour: According to Kristen, bread flour is a great option for starters. Whether it’s in a starter culture or in a bread, “the more protein you have, the more rise you will have,” says Kristen. She uses a blend of bread flour to maintain her starter, Ozzy. With a protein content ranging from 12% to 14%, bread flour has higher gluten potential than all-purpose flour. Greater gluten potential means increased capacity to withstand the expansion of carbon dioxide as a starter matures, trapping gas more easily. Despite this feature, I opted not to use bread flour, in light of its limited availability for the average consumer.*
- Whole Wheat Flour: Some bakers swear by using whole wheat flour in their starters. This flour includes the whole grain of wheat, including the bran, endosperm, and germ. Since it is minimally processed when you compare it with white wheat flours, whole wheat flour is said to be packed with nutrients and potential microbes, providing faster fermentation for a starter. On the other hand, the inclusion of bran can inhibit gluten development. The tiny bits of bran are sharp and act like razor blades, mechanically disrupting the bonds formed in a gluten matrix. Impeded gluten development means your bread might not rise as high, retain its shape, or effectively trap gases. Bran is also high in fiber, which makes whole wheat flour more water absorbent than processed white flour (resulting in stiffer bread doughs).
- Rye Flour: Rye flour is another popular option for sourdough starters. Naturally rich in amylases, nutrients, and microbes, this cereal flour enables a starter to quickly convert sugars for faster, more efficient fermentation. Compared to whole wheat flour, rye flour is said to be the most nutrient- and amylase-dense option for a sourdough starter. Overall, it has a lower gluten protein content than wheat flour, which means it produces slack, sticky, and dense doughs. Rye flour also contains pentosans, five-carbon sugars that absorb water and can also make doughs sticky and gummy; some research suggests that added pentosans stimulates the growth of lactobacilii and the production of acetic acid, presumably resulting in a loaf with a more sour flavor. Rye is high in gliadin but low in glutenin, which also contributes to the viscous and sticky behavior of rye-flour doughs. It’s important to note that these textural qualities are not deal-breakers in bread baking (100% Danish rye bread is proof that rye flour produces good bread). And in the right hands, rye flour’s vast fermentative potential is a powerful tool that can result in tall, flavorful, and aesthetically pleasing sourdough bread.
- Flour Blends: Of course, you don’t have to stick to one flour when making a starter. Many bakers employ blends of flours to produce starters that match their preferences for activity and flavor. Kristen employs a specific mix of rye and bread flours. After tinkering with varying percentages of rye flour, she found a point of diminishing returns. “My favorite mix is a 90 percent white [bread] flour and only 10 percent rye. That smaller amount of rye doesn’t really make it sour, but it does give it that boost of activity. For me, that’s my perfect balance.” For her, it’s also a question of cost and convenience. “Rye is a little harder to find, and more expensive.” If using a flour blend, it’s helpful to pre-mix a large batch of it to make your life easier when it comes time to refresh your starter.
*Added gluten protein does not necessarily mean that a starter has greater leavening power in the strictest definition (the ability to produce carbon dioxide through fermentation). Even though a starter made with bread flour might rise higher in the jar than a starter made from all-purpose flour, both starters could have similar levels of microbial activity and fermentative power. It would be technically incorrect to say that the starter made with bread flour is more “active” solely because of the height difference.
Other Considerations During Testing
In order to compare samples effectively, I needed to keep all other variables constant. Here are several other essential considerations, briefly explained.
For my testing, I stuck with 100% hydration starters, which means cultures made up of equal parts flour and water. Keep in mind that starters can range from "liquid" (100%) to considerably stiff (a 50% hydration Italian lievito madre is a good example). 100% hydration starters are the most common type.
They're easy to maintain, easy to mix into a dough, and more straightforward to adapt into dough calculations for bread. Drier starters require more maintenance, are harder to mix into doughs, and the math is a little trickier. Meanwhile, wetter starters eat through their supply of starches very quickly, and so they require more frequent feeding and attention.
I also had to decide on a feeding ratio. Bakers use the convention X:Y:Z in reference to feeding, where X is the amount of starter, Y is the amount of flour, and Z is the amount of water. Typical ratios range from 1:1:1 all the way to 1:10:10. The higher the ratio, the more food you are providing to your starter, but the more you are diluting the population of microbes. “The idea is you always want to feed it right at peak. You increase the ratio to dilute out your inoculation a little bit,” says Kristen. “So instead of 1:1:1, maybe next time you try 1:2:2. And this should, in theory, slow it down a little bit.”
In the initial stages of my experiment, I stuck to a 1:1:1 ratio, which is easy to measure and provides a good amount of food in the early stages of a starter's development. But over time, as my starters grew more active, that ratio changed.
As the active microbes eat up available sugars in a starter, those sugars run out. At that point, it's essential to feed or refresh the starter with more flour, to provide more sugar to drive fermentation. It’s advisable to feed a starter when it reaches its peak: the maximum height it will reach in its container, which is usually about double or triple the original volume. (To track this increase in volume, I wrapped rubber bands around the mason jars that housed my starters.)
Peak indicates a balance between two things: It represents a stage of significant yeast activity, as carbon dioxide accumulates to aerate the glutinous mass; and it indicates a point at which the gluten matrix in your starter is still relatively strong (and, in turn, can lend some structure to a bread). You can also feed when your starter begins to smell more alcoholic and pungent than pleasantly sour—an indicator of an excessive buildup of ethanol and acids as microbes exhaust available sugars. Finally, you can feed when your starter looks runny or soupy. In this last case, your starter has become proteolytic, meaning enzymes and the buildup of acids have broken down gluten into a gooey mess. The starter has fallen, and is said to be “past peak”. The latter two options are not always ideal, because it’s difficult (but not impossible) to use an increasingly acidic, proteolytic, and soupy starter past its peak for baking.
Of course, in the nascent stages of a starter’s life, there is no way to know what peak is. For my tests, I would have to closely monitor the rising and falling of my samples, and adjust both feeding ratios and feeding schedules to maintain peak activity. I started my tests with a 24-hour feeding cycle, eventually switching to a 12-hour period, then finally a six-hour period.
The important point here was to keep feeding times consistent without changing the schedule too quickly. That way, I could observe how my starters adjusted to changes in the schedule over a longer period of time. I only increased feeding frequency when it became apparent my starters were peaking early—a sign they needed to be fed sooner.
Temperature has a profound effect on the development of a starter. (I’ll save that deep-dive for a separate article.) Simply put, the ideal range for fermentation falls between 70 and 80 degrees. No matter what temperature you choose, make sure you can keep it consistent. I chose to keep my starters in a cool (off!) oven with the light on, which maintained a steady temperature between 78 and 80 degrees.
A Note on Sanitation and Cross-Contamination in Testing
Making a starter involves encouraging the growth of wild microbes. But that doesn’t mean I could be cavalier in my sanitary practices. “You have to use somewhat of an aseptic technique,” advises Kristen. “And you have to keep [your starter] clean, and you have to feed it.” That means proper washing of utensils and jars between feeds, making sure tools are sterile and free of other potentially invasive and undesirable microbes or chemicals (soap, chlorine, etc.).
In my case, I wanted to ensure minimal cross-contamination between my five samples, reducing the chances of microbes from one sample making their way into another, which could skew my results. To that end, I transferred my starters to fresh, clean jars at each feeding. I heat-sterilized every used jar between feedings (by running them through the dishwasher), and used five separate, heat-sterilized spatulas assigned to respective starters over the entire testing period.
With all that information in mind, here’s how my testing went, recorded day by day.
Day 1, 9 AM
I mixed 20 grams of each flour or flour blend with 20 grams of water and left the samples alone for 48 hours at a constant temperature of 78°F. Since the initial population of active microbes was effectively zero, this extended time boosted the chances of inoculation.
There’s not much to look at here, but notice the difference in textures below. Rye formed a stiff mixture, and whole wheat flour was slightly less so. All of these samples smelled...like flour. The rye sample had a slightly earthy aroma, while the whole wheat flour sample was nuttier and mustier. The all-purpose flour sample smelled plain, inert, and faintly of dairy.
After 48 hours, I noticed some activity in my starters. The 100% whole wheat sample appeared to rise the most, with the 50% whole wheat sample a close second. All samples showed some level of bubbling, and they all smelled distinct from one another. 100% rye smelled earthy and funky, with slight notes of nail polish remover (acetone).
The 100% whole wheat sample smelled buttery and toasty, but not sour. The all-purpose sample smelled strongly of chlorine initially, and later smelled like a rotten gym sock by the end of the day. I suspect this could have been due to infiltration of sulfur bacteria—which give a “rotten egg” smell—or just as likely any other bacteria. Finally, some samples were looser than before: the all-purpose and 50/50 samples were considerably runnier than the whole grain samples. I refreshed each starter by mixing 15 grams each of starter, flour, and water. I discarded the rest.
Day 4, 9 AM
By now, both rye samples appeared taller than the others. All samples showed some evidence of gas production. Note the separated layer of liquid at the top of the all-purpose sample. This liquid is known as “hooch,” a clear to dark-colored, watery substance that consists of the alcohol produced as wild yeast ferments. It’s harmless to the development of a starter, but it is a clear indicator that a starter needs to be refreshed and fed. Here, the presence of “hooch” was encouraging, a sign that yeasts were metabolizing sugar and that the starter was heading in the right direction. Given these promising signs of activity, I decided to switch to a more frequent feeding schedule, feeding every 12 hours.
The rye samples were still showing the most visible growth in the jar. Notice the size difference between bubbles of the three samples on the right. Finer bubbles generally indicate a starter that has passed its peak and become proteolytic, since deteriorated gluten can't hold big gas bubbles as easily. But given the recently adjusted feeding schedule, those finer bubbles could have been an anomaly. I wanted to wait until at least three samples showed signs of peaking over a couple days. I decided to hold steady with the current feeding schedule and ratio, continuing at a ratio of 1:1:1 every 12 hours.
This was an interesting day in terms of aroma. The 100% rye sample smelled like alcohol and wood, a bit like whiskey. Meanwhile, the 100% whole wheat sample “smelled like a baby,” according to my girlfriend. The all-purpose sample smelled sour and reminiscent of buttermilk.
The samples were beginning to equalize and reliably double in volume over a 12-hour period. Overall, the funky, musty aromas were dissipating, giving way to increasingly sour and yogurt-like notes.
The all-purpose sample seemed the runniest and most proteolytic, indicating it might be time to feed a higher ratio of flour and water to the starter to keep the sample from fermenting too quickly. All samples were doubling reliably. Given the consistency in height gain over the last two days following this schedule, this was a good time to increase the feeding ratio to 1:2:2.
At this point, all samples were rising predictably and peaking at around the same height. See the pattern of level streaks on the 50/50 rye culture? Those streaks are evidence that the starter may have peaked and fallen before its next feed. All samples were smelling reliably sour. I fed the samples 1:2:2 in the morning. By night, four samples had peaked and fallen, so I decided to increase the feeding ratio to 1:3:3 on Day 9.
Below were the samples by day 11, with a rubber band to reference the starting points. The starters don’t look as active, which might seem discouraging. But keep in mind: The starters were adjusting to a 12-hour feeding cycle at 1:3:3—a higher dilution which slows down fermentation. All samples were at least doubling in volume, which was promising.
With the exception of 100% rye, the starters were inching closer and closer to tripling in volume within a 12-hour period. Each sample was strong, with no signs of runniness, and no overly acidic or alcoholic aromas. They were all sour, in subtly and pleasantly different ways.
Once again, there was some evidence of early peaking, as seen in the above photo.
Day 14 and onward
Finally, all samples (except for 100% rye) more or less tripled in volume over a 12-hour period. They all smelled sour and robust, with no signs of over-fermentation or overly acidic, proteolytic soupiness. Given the predictable rise and fall over the previous days, the consistency of aromas, and the strength of gluten, I could reasonably say that these starters had matured. I could say with some confidence that they were ready to use for baking—but I would need to test them to be sure. At this point, I continued feeding 1:3:3 every 12 hours to maintain this level of activity.
But why didn’t the 100% rye sample triple in volume? It was much stiffer and more viscous than the other starters. Rye contains far less glutenin and lacks the ability to expand and retain as much gas as wheat flour. Alternatively, my particular brand of flour may have been high in soluble fiber, raising water absorption, which likely made the mixture stiffer and less able to climb the sides of the jar. It may be that the rye flour sample had just as much, if not more of a robust population of yeast and bacteria, but it was hard to say without baking.
Overall, I was able to grow stable cultures in all five cases. Here were some general trends:
- Based on the data, rye-based starters tended to reach sourness and mature more quickly than all other cultures.
- After rye, samples with whole wheat flour appeared to stabilize to sour cultures faster than the all-purpose culture.
- Whole wheat-containing cultures also tended to be looser and slightly runnier than all other samples**.
- The all-purpose starter was by far the most pleasant and mild smelling, with attractive dairy and sour cream notes.
**After further research, I learned that the thickness or stiffness of a whole wheat culture varies greatly depending on the variety of wheat. The same concept applies to varieties of rye flour. In my tests, my brand of rye flour (Arrowhead Mills) produced the stiffest starter over time. But in Kristen’s experience, she found her whole wheat starters to be stiffer when compared with rye flour starters.
So far, I'd succeeded in growing five distinct, mature starters from different flours or flour blends. But beyond visual cues and smells, I had no way of determining whether one starter performed better than the other when it came to baking a real loaf. I couldn’t stop there. So I put all my new starters to the true test of quality.
I baked five identical sourdough loaves using each of my developed starters; all loaves consisted of 80% bread flour and 20% sifted whole wheat flour, in addition to the starter (20% of the total flour weight). I won’t get too deep into the details of baking bread here, since that’s material for another post, but to hold variables constant, I closely monitored dough temperature to maintain 74 degrees, and kept the fermentation time consistent across the board. I evaluated exterior crust, interior crumb structure, flavor, and texture.
100% Rye Flour Starter
Crust: Blistered, tall, and round. Dark caramel color.
Crumb Structure: Even and open bubbles, not too much irregularity.
Flavor and Texture: Mild, balanced sour, slightly earthy, more substantial texture; denser than a plain white loaf.
This loaf proofed and baked beautifully. The fermentation was quite active, and the finished proofed dough was light and bubbly. The crumb was relatively even, with large and small bubbles distributed across the entire surface of the loaf. While breads with rye flour are typically gummier and dense, the texture here was sturdy and substantial—but still pleasant. This bread was slightly darker in appearance—both inside and out—than the other samples. The flavor was a rounder, more dairy-forward sour that played nicely with the earthiness of the rye flour.
100% Whole Wheat Starter
Crust: Mildly blistered, less tall loaf. Light caramel color.
Crumb Structure: Wild, molten, and slightly uneven.
Flavor and Texture: Sharper sour, soy, malt, umami; gummy texture.
This loaf was harder to handle and bake. I found the dough to be more slack, stickier, and stretchier (extensible, in baker’s terms) than the other loaves. I suspect that the inclusion of bran from whole grains inhibited gluten development somewhat, making the dough harder to handle. The overall fermentation was not quite as active as the other loaves, resulting in a slightly shorter loaf. Still, it was a good loaf despite these minor imperfections. The crumb was wild and molten like lava, with massive bubbles in some areas and smaller but open bubbles concentrated in the middle. The flavors were delicious: a sharp tang reminiscent of pecorino cheese, hints of soy, malt, and roasted umami. The texture was slightly gummy in comparison to the other loaves, and could have benefited from additional time resting before cutting. (Extended resting is a common technique for loaves higher in whole-grain content: It allows cooked starches to retrograde and set, giving a firmer, less gummy texture.)
50/50 Rye and All-Purpose Flour
Crust: Very blistered, tall, round loaf; tan exterior.
Crumb Structure: Wild, molten, and uneven.
Flavor and Texture: Balanced sour, light tasting, airy.
This loaf was slightly over-proofed, indicating a vigorous and more active starter than the other samples. (Remember: I kept fermentation time and temperature constant for all tests.) This dough rose faster and higher in the designated time frame than all the others—a little too fast, in fact. Despite that minor error, the bread still baked nicely. The dough behaved most closely to doughs I've made using my old starter, Ryan, showing strong fermentation as it rose prior to shaping. The crumb was massively open and molten through 60 percent of the loaf, with smaller but still open bubbles concentrated on the right side—maybe an indicator of slight over-proofing, or a shaping error. The flavor was balanced and mildly sour, the texture light and fluffy.
50/50 Whole Wheat and All-Purpose Flour
Crust: Moderate oval, moderately blistered, mahogany exterior.
Crumb Structure: Slight molten and mostly even, slight tan interior.
Flavor and Texture: Toasted soy, malt, nutty, tangy; moderately soft.
This loaf proofed and baked reasonably well. It wasn’t the tallest loaf, but the fermentation was steady. The dough wasn't over-proofed by the time of shaping, resulting in a consistent and round shape. The dough was extensible but strong throughout its fermentation. The crumb was slightly molten and generally even, with a darker tan interior. The texture was soft but substantial enough for a sandwich. The flavor was outstanding: sour, full of toasted malt flavors, and reminiscent of soy sauce—my favorite of the bunch.
100% All-Purpose Flour
Crust: Light tan, moderately blistered, thin crackling crust.
Crumb Structure: Textbook open web, molten, uniform; creamy, pale white.
Flavor and Texture: Mild, faintly sour, very clean flavor; fluffy.
This loaf baked nearly perfectly; it was tall and round, sporting a pronounced raised point of crust (or ear). The fermentation was steady, and the finished dough easy to handle with minimal sticking. The crumb was pale white, featuring wild and molten bubbles that were evenly distributed throughout. This is the kind of crumb most people gush about on social media—the so-called ‘IG Crumb.' I found this loaf soft, fluffy, mild in sourness but lacking the subtle umami flavors of the whole wheat samples.
What Are The Important Takeaways?
You Can Make a Starter Using a Variety of Flours
Based on my tests, it’s possible to build a functional starter using a range of wheat- or cereal-grain based flours. No matter what flour you choose, over time you will produce a culture that is sour, bubbly, and—most importantly—one that can raise bread.
Different Starters Will Behave and Bake Differently
Depending on what flour you use in your starter, there appear to be differences in time to maturity, flavor, activity, and performance in baking.
- For faster maturity and vigorous activity, starters with rye flour seemed to be the best option—presumably because of rye's high amylase and nutrient content.
- Whole wheat flour provided the most interesting flavors and aromas in my baked loaves, but did not produce the tallest loaf or most visually vigorous fermentation. The inclusion of bran in a starter is likely an inhibitor in the gluten development of a baked loaf. The 100% whole wheat starter also appeared to ferment at the slowest rate.
- The 100% all-purpose flour starter matured the slowest and had mild sour flavors, but added the most strength to dough for a tall bake and attractive crumb structure. As Kristen explains, bread formulas use starter in amounts as high as 20 percent of the initial weight of flour. When using a 100% hydration liquid starter, that means that 10 percent of the flour in the final dough comes from your starter. For this reason, she recommends that your starter should have strong gluten potential. “The more your [starter] is broken down, the more you’re just adding watery glop to your [dough],” she says. One way to ensure a structurally strong starter is to use some proportion of high-protein flour (above 11%) such as King Arthur all-purpose flour, or, if you can find it, high-gluten bread flour.
In light of these differences in starters, Kristen recommends adjusting dough formulas to account for the flours in your starter. Since starters can contribute a significant portion of a bread's final flour content, it helps to think about ways to compensate for that flour in a complete bread recipe. Doing so takes experience and some math. For example, if you use a 100% whole wheat starter, you may opt for a higher percentage of high-protein all-purpose flour or bread flour in the dough formula. Additional higher-protein flour will provide the necessary gluten structure that the whole wheat flour tends to inhibit.
Cultivating a starter isn’t hard. I suppose this testing was a long-winded way of saying that there’s no real “right” answer—just guidelines to follow, concepts to understand, and choices you have to make along the way. I’m not going to sit in the ivory tower of recipe development and tell you, "This is the best way to make a starter." To say anything like that would be disingenuous, however infuriating that may be. Take some flour—pretty much any flour—and mix in some water. Under the right conditions, given enough time, a starter just lives. Ultimately, despite all the detailed, explicit tutorials she’s provided for thousands of bakers, Kristen ponders, “Everyone wants the quick, 'Tell me what to do, and I’ll do it.' That’s not how it works...and this is part of the highly subjective nature of sourdough starter.”
So slow down, respect the process, use your senses, and just keep going. In the inimitable words of Jeff Goldblum in Jurassic Park, “Life, uh...finds a way.”