Every couple weeks or so, we send out a little roundup of new recipes, techniques, and tutorials that we’ve recently posted on the site. Occasionally we announce exclusive giveaways to newsletter subscribers as well. We won’t spam you with ads or share or sell your email address. Every email we send has a 1-click unsubscribe link if you decide it’s not for you.
Sourdough Microbiomes and Bread Flavor
Like many sourdough bakers, I’m intrigued by the debate about what determines a sourdough starter’s microbial composition. Is it the food source, the environment or terroir, storage temperature, frequency of feeding, baker’s hands, or all of the above? And once a starter is established, does a starter’s microbiome shift if you change any of these factors? And do any of these microbial goings on affect the appearance or flavor of bread baked with them?
Bakers tend to agree that starter hydration and storage temperature can encourage some microbes to flourish over others; e.g. the use of low hydration pasta madre to make heavily enriched panettone. It’s also commonly accepted that whole grain flour will make a starter’s microbes multiply faster than refined flour. But debate continues about whether an established starter changes if moved to a new location or if fed a new flour or wheat variety.
I recently learned more about the work of the Puratos World Heritage Sourdough Library in Belgium. This library houses over 130 sourdoughs from more than 23 countries, and refreshes the starters with their original flours, using the original methods. (The storage temperatures may differ from the original customs.) In 2018, the sourdough library did a research study which demonstrated that sourdough starters remained unique even when fed identical flour for a short period of time. In the study, 15 bakers were sent new flour to use for 10 feedings. The 15 bakers then traveled to the library and baked breads with their starters. The bakers detected different flavors in the breads despite the period of identical feedings.
Learning about this research and also reading the research of @benito into rye starter pH and buffering ability in this forum thread inspired me to do this experiment where I created a new sourdough starter from rye flour and compared it to an all purpose (AP) sourdough starter that I converted to rye flour. I wanted to see if the starters would behave the same because they were created in the same kitchen and now being fed the same flour, or if the biomes would be different and persist in being different over time.
Pain de campagne using bread flour, home-milled rouge de bordeaux wheat, and two different home-milled rye flour sourdough starters. Baked six weeks into the experiment, these loaves look similar but taste different.
EXPERIMENT
In this experiment, I compared the behavior of a rye flour starter I built from scratch with a rye flour starter I converted from my five-year old starter [originally created with all purpose flour] which has been fed many different wheat flours over the years. I baked with these two starters at the four-week and six-week marks and during that time I discarded and fed the starters every few days (with one week-long break) with the same home-milled, whole grain rye flour and I stored them in the refrigerator when they weren’t actively doubling after a feed.
Usually when I do an experiment, my results are in the ballpark of what I had predicted, but this experiment surprised me. After six weeks, the two starters had been discard-fed between 15 and 20 times, and they’re still different colors and have different aromas. Moreover, they make breads that taste different, even though I’ve synced the starters completely in terms of ripening speed, dough leavening strength at warm and cold temperatures, and final loaf appearance.
The rye from “Scratch” starter is grayer in color, fruitier in aroma, and when mature, it feels thicker when stirred. It also makes a bread that most of my taste testers preferred. The rye “Convert” from all purpose flour starter is pinker in color, has a sharper, slightly acetone aroma, and when mature, it feels looser when stirred. It makes a bread that my taste testers enjoyed but found to be less flavorful. (See the conclusion below for more taste testing details.)
Spoiler 1: These results indicate that the sourdough starter microbiome may be significantly determined by the population of microbes in the initial flour used to create the starter and that this microbiome can remain stable even in the presence of a consistently different feeding flour.
Spoiler 2: A third test bake (see FURTHER RESEARCH below) indicates that with persistent feedings using a new flour and method, the starter microbiome may eventually shift away from its original composition and flavors.
METHODS
Weeks 1 & 2
Creating a “Convert” to Rye Starter (Tap or roll your mouse on the photos to see the captions)
I fed 6g of all purpose flour starter with 30g of whole grain rye flour and 30g of water (6:30:30 or 1:5:5). After it doubled in 8.5 hours, I refrigerated it overnight. The next day, I discarded down to 6g and again fed 30g each whole grain rye flour and water. This starter, which now contained 0.3g AP flour, doubled in 7 hours. I then refrigerated it for a couple of days. This FAQ explains the math behind the disappearance of the all purpose flour in just two feedings. The amount of AP flour in the Convert starter after multiple feeds is comparable to what might float through the air and land on either starter when I have the starters out on my work space.
Creating a From “Scratch” Rye Starter
I mixed 40g whole grain rye flour and 40g water. After 18 hours, the paste still looked dormant so I stirred it vigorously. At 30 hours, the starter had expanded by about 75% and seemed dormant again. I added more rye flour and water (40g each).The starter then doubled by 42 hours from the initial mixing.
I decided to bake a loaf of bread with this expanded very young starter, and it turned out well. I suspect this initial burst of leavening power was due to the fact that I used home-milled rye flour, which is arguably a more alive than refined flour.
The Scratch starter continued to evolve, however, and was very weak for the next 2-3 feeds. I believe I may have overfed it by discarding and refrigerating too soon. The starter needed more time and feedings at room temperature without discarding for the pH to drop and the final balanced starter microbiome to get established. This took until about day 11, with less but still some refrigeration and discarding.
I didn’t photograph the progress of each feeding, but in the photos below, you can see that during these first couple of weeks, the Convert starter (purple rubber band) and original AP starter (green rubber band) always expanded faster than the Scratch starter (blue rubber band).
Weeks 3 & 4
During these weeks, the Scratch starter began to gain in strength and it became apparent that the two starters were different in color. When the Scratch starter began to exceed the Convert in vigor, I tried to sync the starters up in terms of strength by playing with temperature and feeding ratios.
As the two sourdough starters got closer to syncing up, I decided it was time to do a test bake. I used Breadtopia’s Pain de Campagne recipe, but instead of AP starter and rye flour, I used the Convert and Scratch rye starters. I also didn’t autolyse or laminate any of the doughs, which is part of the process I had originally written into that recipe. For all of the test doughs, I mixed all-in and did two rounds of stretching and folding, and two rounds of coil folding.
Week 4 Test Bake
Week 4 Bake Results
The texture of the Scratch and Convert loaves was similar, but taste testers (two family members) noted in blind taste tests that the Scratch starter bread had “more wheat flavor” and “less sourness.” This was an interesting result on its own, but also unexpected because the Scratch dough spent 7 hours longer in the refrigerator trying to catch up fermentation-wise to the Convert dough. All other things equal, my experience is that more time in the cold usually confers more of a sour flavor, but not in this case.
Weeks 5 & 6
During week 5, I fed the starters once, let them mature, refrigerated them and went away for a week. Upon returning, I discarded both starters down to 30g and fed them 1:1:1 (30g rye flour and 30g water). Below is a photo of the discards, all of which I used in a delicious “just got home from vacation” instant yeast bread. You can see that the starters continue to be two different shades of brown even in different lighting situations.
I refrigerated the two starters at the point you see them in the photo above, and a few days later, I brought them to room temperature and used them in two doughs. This was my week-six test bake, again two pain de campagne breads, with the same formula as the prior test bake, except this time I used rouge de bordeaux as the whole grain red wheat flour.
Week 6 Test Bake
Week 6 Bake Results
Similar to the week-four test bake, the Scratch and Convert loaves of the week-six test bake looked about the same, though this time the similarity was achieved with a virtually identical fermentation duration, as opposed to syncing up the appearance of the dough with differences in bulk fermentation and final proofing time as I did in week 4.
For this test bake, I had six people do blind taste testing. One tester didn’t detect any flavor differences, and the other five preferred the bread leavened with the Scratch rye starter built from rye flour. They described the Scratch bread as having “deeper flavors,” more “huskiness and pepperiness,” and a “better sour” than the Convert rye starter, which had been transformed from an all purpose flour starter. The difference also came through in the aroma of the bread for three of the six testers.
Rye starter built from scratch on left; Rye starter converted from AP starter on the right
PRELIMINARY CONCLUSIONS
For the six-week time period that has passed thus far, this experiment suggests that the microbial composition of these two sourdough starters is not the same despite their being fed the same flour. The sourdough starter built from scratch with rye flour and fed 15-20 times during the test period is grayer in color, mixes with a thicker glue-feel, smells fruitier and less like acetone, and bakes a bread that five of six taste testers preferred. The sourdough starter converted from my original all purpose flour to rye flour and discard-fed 15-20 times is pinker in color, mixes with a thinner liquid-feel, smells less fruity and more like acetone, and bakes a more “one dimensional” and sour bread.
Continuing the experiment for a longer time frame could show an evolution and convergence of the starters or a continued stability of difference. More time and baking is needed to determine this.
It would also be interesting to do a test bake comparing breads made with the Convert starter and with the original all purpose flour starter it grew from (making sure the flour types in each final dough were identical, of course). This could demonstrate whether the Convert has a similar microbial composition as the AP starter or its own microbial composition, not identical to its parent starter or to the Scratch starter.
FURTHER RESEARCH
Week 12 Test Bake (after 10 additional 1:1:1 starter feedings)
In the six weeks since my last test bake, I fed the experimental starters 10 more times (30 feeds in total) with a 1:1:1 ratio of starter-to-flour-to-water. About halfway through this time period, the two sourdough starters still looked and smelled different (25 or so feedings).
Scratch on left more gray; Convert on right more pink and acetone-smelling
By week 12 (30 feeds), the starters looked and smelled more similar, though they still had a different mixing-feel. The Convert starter continued to mix with a looser texture despite being the same hydration as the Scratch starter.
Scratch starter on left; Convert starter on right
At this point, I decided to make two whole grain sandwich breads using sprouted spelt flour and the Scratch and Convert sourdough starters. The doughs fermented at the same rate, staggered by an hour due to timing of starter readiness. (Convert was faster this time.)
I used identical shaping motions for both loaves, but the Convert starter dough was looser, and you can see from the photos below that it filled the loaf pan differently, more like the dough had been poured in. When the loaves were baked up though, the Convert starter bread unexpectedly had more of a dome.
Here is the formula for the two breads:
Here is the method for preparing and baking the breads:
Mix the flour, water, and ripe starter (starter more than doubled).
Let sit or “fermentolyse” for 1 hour, and then add the salt and oil, mixing until smooth.
Perform one round of stretching and folding at the 1.5 hour mark.
Allow the dough to expand by about 75%. (In a summer-temp kitchen, this took about 5 hours from the initial mixing.)
Shape the dough into a tube, place it in an oiled loaf pan, cover, and let the loaf double in size (2.5 – 3 hours).
Brush with water and bake at 375F for 20 minutes and 350F for an additional 25 minutes.
Though there remained some differences in consistency/feel between the Scratch and Convert starters and doughs, taste testers for this round of baking were unable to pick out flavor differences in the breads. I gave repeated blind taste tests, and the testers were so uncertain and inconsistent with themselves about the flavors that I decided to improve the testing strategy and use a double blind testing approach.
Each tester got three pieces of bread (two from one loaf and one from the other). They didn’t know which breads were which, and they were asked to identify which one was the different bread. No one was able to guess this correctly, including me.
Week 19 Test Bake (back to Pain de Campagne breads)
By week 19, seven weeks since the last test bake, I had fed the experimental starters 10 more times (40 feeds in total) with a mix of 1:1:1 and 1:3:3 ratios of starter-to-flour-to-water, higher ratio if I was aiming for overnight ripening.
During these weeks, the starters began to be indistinguishable in terms of mixing-feel after ripening, and sometimes the Convert starter would ripen faster after being fed, and sometimes the Scratch starter would ripen faster after being fed. I always aimed to refrigerate or use them based on their appearance (doubling) rather than similar timing. I suspect the inconsistency of ripening time was because of jar position in the refrigerator, which I tried to control but not always successfully. A starter positioned farther back in my refrigerator would be colder and therefore more dormant at the next feed.
After the prior test bake of whole grain sprouted spelt loaves, I wanted to return to using the pain de campagne recipe because I had worried that any mild microbial flavor differences might be harder to detect in nutty whole grain breads. Here’s the formula I used, which is the same as the week 6 test bake.
Week 19 test bake of pain de campagne loaves; Scratch starter on the left and Convert starter on right
For this bake, the Scratch starter developed faster during the overnight ripening. When I headed to bed, it was higher in the jar, and though the two jars looked the same in the morning, I suspect the Scratch starter was in the midst of deflating while the Convert starter was still expanding. I went ahead and mixed the doughs at the same time anyway and crossed my fingers.
My suspicion seemed to be confirmed by the rest of the process: the Scratch starter dough took longer to ferment, felt less elastic when I was shaping it (as if the gluten was weaker), and it baked up flatter than the Convert starter bread.
Scratch starter loaf on left; Convert starter loaf on right
Flavor-wise however, the loaves were almost indistinguishable. Some taste testers picked out a little more sourness in the Scratch starter loaf; and other taste testers didn’t detect any differences.
On a side note, you can see how far I let the doughs bulk ferment for this bake compared with the other pain de campagne test bakes. These doughs almost reached the 1.5L mark on the buckets, whereas in prior bakes, the doughs barely reached the 1L mark. As a result, these very fermented loaves have smaller ears and less height overall. Yet they’re still wonderful and delicious breads. (When I realized how far I’d let the dough bulk ferment, I skipped the pre-shape and bench rest, shaped the doughs, and immediately refrigerated the proofing baskets).
FINAL CONCLUSIONS
Based on the results from my latest two rounds of baking at 12 and 19 weeks, it appears that over time and under similar conditions, sourdough starters (at least these two starters…) that began as different in character move toward convergence with identical feeding schedules and temperature shifts into and out of the refrigerator.
For the first 6 weeks (20 or so feeds), the made-from-scratch-rye sourdough starter and the converted-to-rye-from-all purpose-flour sourdough starter were different in color, texture, and aroma; and the breads baked had flavor differences that were easy to detect. Even at 9.5 weeks, the starters were still different in appearance (gray vs. pink), but by about 12 weeks (30 feeds), the starters looked similar and produced similar tasting bread. These breads were whole grain sprouted spelt sandwich loaves, whereas the breads from the test bakes of weeks 4 and 6 were pain de campagne loaves.
After the week 12 test bake where taste testers couldn’t detect differences in flavor, I continued to feed the starters for seven more weeks. At week 19 (40 feeds), I did another test bake, going back to the original pain de campagne recipe just in case the nutty whole grain spelt flavor of the previous test masked small differences in microbial flavors.
In this final bake, some taste testers couldn’t detect a difference in the breads and some found the Scratch starter loaf slightly more sour. There was no mention of differences in depth of flavor or pepperiness.
I suspect this extra sourness in the Scratch starter loaf can be explained by the over-maturity of the sourdough starter, which would also explain why the Scratch bread seemed to have slightly degraded gluten, baking up flatter despite expanding to the same level in the fermentation bucket and proofing basket as the Convert starter loaf.
A more controlled sourdough starter build would have been ideal, but despite the breads looking different and having a mild difference in sourness, the differences in flavor character that existed earlier in the experiment continue to be absent.
Sourdough Microbiomes and Bread Flavor