settling into genuine wonder

What keeps pulling at me is the relationship between patience and aliveness in sourdough. There's something strange about how we talk about "feeding" a starter - as if we're in a caretaking relationship with something that isn't quite an organism but isn't quite not one either. Where does the life actually live? In the individual yeasts? In the bacterial colonies? Or somehow in the relationship between them and the flour and the air of a particular kitchen?

I find myself genuinely puzzled by why sourdough starters develop regional character. People say their San Francisco starter behaves differently in Vermont. Is the starter changing, or is it the same starter responding differently? What even counts as "the same starter" when every cell has been replaced thousands of times?

And there's this temporal strangeness - some starters are supposedly centuries old, passed down through generations. But what exactly has persisted? Not the original matter. Is it a pattern? A process? It feels like the starter is more like a river than a thing.

I'm also drawn to the waiting. Why does the slow fermentation create flavors that fast yeast can't? What is happening in that time that matters so much? Is there something about rushed processes that we can taste?

What draws you to sourdough?

What mechanism allows a sourdough starter to maintain a stable microbial community identity over decades of continuous feeding, when the selective pressures (flour type, temperature, timing) seem too weak to prevent drift toward whatever organisms happen to be most abundant in the local environment?

Investigating Sourdough Community Self-Reconstruction

The Question

What functional relationships between microbial species in a sourdough starter create self-reinforcing conditions that actively reconstruct community composition after perturbation, rather than merely resisting change?

Methodological Approach

Phase 1: Characterizing the Perturbation-Recovery Cycle

Experimental Design: Establish 30 mature sourdough starters (minimum 6 months old, stable community) from diverse sources. Subject each to controlled perturbations:

• Temperature shock (48 hours at 4°C or 40°C)
• pH disruption (buffer to pH 6.0 temporarily)
• Dilution stress (1:100 dilution into fresh flour/water)
• Competitive invasion (introduce high concentrations of non-sourdough LAB)

Measurements:

• 16S rRNA and ITS sequencing at days 0, 1, 3, 7, 14, 28 post-perturbation
• Quantify recovery trajectories: does the community return to pre-perturbation composition, or reach a new stable state?
• Calculate resilience metrics: time to recovery, compositional distance from baseline

Phase 2: Mapping Cross-Feeding Networks

Metabolomic Profiling:

• Use untargeted LC-MS to measure metabolite pools during normal fermentation and recovery phases
• Track specific compounds: maltose, glucose, amino acids, organic acids, CO2, ethanol
• Identify which metabolites accumulate when specific species are absent (using selective antibiotics)

Spent Media Experiments:

• Grow dominant species (e.g., %%GTERM:fructilactobacillus:Fructilactobacillus sanfranciscensis%%, %%GTERM:kazachstania:Kazachstania humilis%%) in isolation
• Test whether sterile-filtered spent media from one species promotes or inhibits growth of others
• Measure: growth rates, lag phase duration, final cell density

Phase 3: Testing the Autocatalysis Hypothesis

Key Prediction: If the community is autocatalytic, removing any member should disrupt the whole system more than expected from that member's individual contribution.

Synthetic Community Construction:

• Isolate dominant strains from stable starters
• Reconstruct defined communities with 2, 3, 4, and 5 species combinations
• Measure: stability over 30 days of backslopping, functional outputs (pH drop rate, CO2 production, volatile profiles)

Knockout Experiments:

• Use species-specific phages or selective antibiotics to remove individual species from established starters
• Track whether remaining species can maintain stable community or collapse
• Compare observed dynamics to null model predictions (neutral drift)

Phase 4: Identifying the Mechanism of Competitive Exclusion

The search results revealed that manganese depletion by lactobacilli excludes fungi in dairy systems. Could similar nutrient sequestration operate in sourdough?

Measurements:

• ICP-MS to quantify trace metals (Mn, Fe, Zn) in sourdough supernatant over fermentation
• Compare metal concentrations in mature vs. young starters
• Test whether adding excess manganese allows invasion by non-resident yeasts

Antimicrobial Screening:

• Screen dominant LAB strains for bacteriocin production against potential invaders
• Test cell-free supernatants for growth inhibition (controlling for pH)
• Sequence genomes for bacteriocin gene clusters

Phase 5: Testing Identity Persistence

The River Hypothesis: If identity lives in functional relationships rather than species lists, two starters with different organisms might be functionally "the same."

Functional Equivalence Test:

• Identify starters with similar flavor/fermentation profiles but different species compositions
• Swap dominant species between them
• Measure: Do they reconverge to original functional profiles, or adopt donor characteristics?

Data Integration

Build a network model where:

• Nodes = species
• Edges = measured metabolic dependencies (positive or negative)
• Edge weights = strength of cross-feeding or inhibition

Use this to predict:

• Which species combinations should be stable
• Which perturbations should be recoverable vs. catastrophic
• Whether observed stability patterns match ecological autocatalysis predictions

What Would "Success" Look Like?

Evidence supporting self-reinforcing reconstruction would include:

1. Consistent return to baseline composition after diverse perturbations
2. Measurable cross-feeding loops where species A feeds B, B feeds C, C feeds A
3. Community collapse when loop members are removed, but recovery when non-loop members are removed
4. Nutrient depletion patterns that exclude potential invaders
5. Functional convergence even with different species compositions

What Changes With an Answer

If the autocatalysis hypothesis is confirmed

For understanding biological identity: This would give us a rigorous framework for what it means for a sourdough starter to be "the same" over time. Identity would live in the topology of relationships rather than in the material constituents. The 150-year-old starter passed down through generations wouldn't just be a romantic story - it would be a genuine case of pattern-persistence through continuous recreation. This has implications beyond sourdough: it's a concrete example of how complex systems maintain coherence without any central controller.

For practical baking: If we know which cross-feeding relationships are load-bearing, we could diagnose struggling starters by identifying broken loops rather than just checking "is it bubbly." A starter that's functionally dead might still have organisms in it - but the wrong relationships between them. This would shift advice from "feed it more" to "which relationship needs repair."

For microbiome science broadly: Sourdough would become a tractable model system for studying community resilience. It's simpler than gut microbiomes, faster to manipulate, and ethically uncomplicated. If we can demonstrate autocatalytic stability here, it suggests where to look in more complex systems - human gut, soil communities, fermentation broadly.

If the hypothesis is wrong

What would be overturned: Maybe stability is actually much simpler - perhaps one or two keystone species just dominate through competitive exclusion, and the "community" is really a dictatorship with hangers-on. Or maybe stability is largely illusory and we've been sampling starters that happen to be in temporary equilibrium, while most attempts at long-term propagation actually fail (survivorship bias in inherited starters).

This would mean the "terroir" and "heritage" narratives around sourdough are mostly mythology - the starter from San Francisco shipped to Vermont doesn't carry some irreducible identity, it's just organisms that will be replaced by whatever thrives locally. The romance deflates, but the bread is still good.

Who else would care

Fermentation industries: If we understand why some fermentations are stable and others drift, we can engineer more robust industrial processes - not just sourdough but kombucha, kefir, tempeh, any continuous culture.

Ecologists studying alternative stable states: Sourdough might be a rare example of a system that doesn't have alternative stable states - or one that shows clear hysteresis. Either finding informs theory.

Philosophers of biology: The question "what persists when nothing material persists" is old. A concrete answer about sourdough wouldn't resolve the ship-of-Theseus problem, but it would provide an empirical example of how one type of identity can be maintained through process rather than substance.

Home bakers who've killed their starters: If stability comes from relationship networks, we could give better advice than "start over." Maybe the network can be re-seeded, or a partial starter from a friend can restore missing nodes.

The assumption I'm most uncertain about

I've been assuming that the stability we observe in long-lived starters reflects something about the community - its internal structure. But it might reflect something about the niche. The flour-water-temperature-feeding rhythm might be so constraining that almost any starting community converges to similar endpoints. Stability would then be a property of the environment, not the relationships.

This would be important because it would mean the "same starter" question is malformed - like asking if the same water is in a fountain that's been running for decades. The answer is obviously no, and the question misunderstands what fountains are.

Immediate Next Actions (Next 7 Days)

Day 1-2: Establish Baseline Starters

What to do:

• Acquire 3-4 mature starters from different sources (local bakery, friend, commercial dried starter, one created fresh as control)
• Begin daily documentation: photograph, record smell, measure pH with strips, note rise time
• Take initial samples (1ml each) and freeze at -20°C for later sequencing comparison

Why this matters: Can't study perturbation-recovery without knowing what "normal" looks like for each starter. The fresh control lets us watch succession happen in real time.

Cost/logistics: ~$30 for pH strips, containers, and one commercial starter. Freezer space. 15 minutes daily.

Day 2-3: Design the Perturbation Experiment

What to do:

• Split each mature starter into 5 portions:

  • Control (normal feeding schedule)
  • Cold shock (refrigerate for 72 hours, then resume)
  • Heat stress (keep at 95°F/35°C for 48 hours)
  • Starvation (skip 3 consecutive feedings)
  • Dilution (1:20 into fresh flour/water)

• Create tracking spreadsheet with daily measurements for each condition

Why this matters: This is the core experimental structure. Different stressors test different aspects of resilience - cold slows metabolism, heat favors different species, starvation depletes resources, dilution tests whether small populations can re-establish.

Cost/logistics: More containers, warming mat (~$20), refrigerator space. Flour costs increase 5x.

Day 3-4: Set Up Simple Cross-Feeding Test

What to do:

• Take active starter, let it ferment until very mature (past peak, starting to collapse)
• Filter through coffee filter to remove most cells
• Add this "spent media" to fresh flour/water and observe: does fermentation initiate faster than a control with plain water?
• Reverse test: does fresh starter added to spent media perform worse than fresh starter in fresh media?

Why this matters: This is a low-tech probe of metabolic relationships. If spent media accelerates new fermentation, something beneficial persists beyond the cells. If it inhibits, the community may depend on nutrient depletion for stability.

Cost/logistics: Coffee filters, extra containers. 2 hours hands-on time spread across several days.

Day 4-5: Literature Deep-Dive on Specific Mechanisms

What to read:

• The Veldhuis paper on ecological autocatalysis (found in search) - understand the formal criteria for autocatalytic loops
• Papers on Fructilactobacillus sanfranciscensis metabolism specifically - what does it produce, what does it require?
• The manganese depletion study from dairy systems - is there analogous trace metal work in sourdough?

Specific questions to answer:

• What metabolites does the dominant LAB produce that yeasts can use?
• What do yeasts produce that LAB require?
• Has anyone mapped the complete cross-feeding network for even one sourdough community?

Why this matters: Avoid reinventing existing knowledge. Identify the specific molecular candidates for cross-feeding relationships before designing more sophisticated experiments.

Cost/logistics: Free if you have library access. 4-6 hours reading time.

Day 5-6: Reach Out to Researchers

Who to contact:

• Authors of the Global Sourdough Project (Rob Dunn lab at NC State)
• Authors of the flour-type/feeding-schedule study (recent, may have unpublished perturbation data)
• Local university food science or microbiology department

What to ask:

• Has anyone done systematic perturbation-recovery experiments?
• Are there existing frozen sample banks from longitudinal studies?
• Would they be interested in collaboration or mentorship on this question?

Template email:

I'm investigating how sourdough starter communities reconstruct themselves after disruption - whether stability comes from self-reinforcing metabolic relationships rather than just resistance to change. I noticed your work on [specific paper]. Have you observed or have data on recovery dynamics after perturbation? I'd welcome any guidance on methodology or existing datasets.

Why this matters: Someone may have already done this work unpublished, or have samples/data they'd share. Academic labs often have sequencing capacity that would otherwise be prohibitively expensive.

Cost/logistics: Free. 1-2 hours composing thoughtful emails.

Day 6-7: Document and Refine

What to do:

• Write up observations from first week of perturbation experiment
• Note which measurements seem informative vs. noisy
• Identify the single most critical unknown: what piece of information would most change your understanding?
• Plan for week 2: probably need to source affordable 16S sequencing (companies like uBiome successors, or academic collaborator)

Create a decision tree:

• If starters recover uniformly → supports strong self-reinforcement
• If recovery varies by perturbation type → suggests different mechanisms for different stresses
• If starters don't recover → challenges the whole stability assumption

Realistic Week 1 Budget

| Item | Cost | |------|------| | pH strips | $8 | | Extra containers | $10 | | Commercial starter | $12 | | Warming mat | $20 | | Extra flour | $15 | | Total | ~$65 |

Plus roughly 8-10 hours of hands-on time and 6+ hours of reading/writing.

What Success Looks Like After Week 1

• Four starters under controlled observation with baseline data
• Perturbation experiment running with daily measurements
• Preliminary spent-media results suggesting presence/absence of cross-feeding
• At least one researcher contact who's responded
• Clearer sense of which specific molecular mechanisms to investigate
• A refined version of the research question based on what the starters are actually doing

Response to Question 1: Where Does Identity Actually Reside?

This question cuts to the heart of what I've been circling without quite articulating. Let me think through the "built environment" experiment seriously.

The Sterile Filtrate Reconstitution Experiment:

If I take a mature starter, filter out all cells, and then expose this "ghost environment" to naive flour microbes, several outcomes are possible:

1. Rapid reconstitution of similar community → Identity lives primarily in the built environment. The acid gradients, depleted nutrients, accumulated metabolites, and perhaps even physical structure create such strong selective pressure that almost any inoculum converges to the same functional state.

2. Slow or failed reconstitution → The living network matters. The built environment alone can't bootstrap the community; you need the actual organisms with their specific enzymatic capabilities and metabolic timing to re-establish the loops.

3. Reconstitution of function but not composition → This is the most interesting possibility. The new community might acidify at similar rates, produce similar flavors, show similar resilience—but with completely different species. This would suggest identity is neither in the cells nor the environment alone, but in the matching between them.

What this implies:

If outcome 1 is true, then the "heritage starter" narrative partially dissolves—but something profound replaces it. The identity would reside in a pattern of constraints that's been refined over time. The starter would be more like a riverbed than a river: the water changes constantly, but the bed shapes what flows through it. Passing down a starter would be passing down this shaped environment, not the organisms themselves.

The Smallest Decisive Common Garden Experiment:

To distinguish autocatalytic self-reconstruction from regime-imposed convergence:

Design:

• Take 6 starters with demonstrably different stable compositions (confirmed by sequencing)
• Split each into two portions
• Portion A: continue normal maintenance
• Portion B: transfer to a standardized "common garden" regime (same flour, same hydration, same temperature, same timing) for 60 days

Key measurements:

• Community composition at days 0, 30, 60
• Recovery dynamics after a standardized perturbation at day 45
• Functional outputs: pH trajectory, rise time, volatile profiles

Decision criteria:

• If Portion B starters converge toward similar composition → regime dominates; convergence is environmentally imposed
• If Portion B starters maintain distinctiveness despite identical regime → something about each community's internal structure resists homogenization; supports autocatalysis
• If Portion B starters converge in composition but diverge in recovery dynamics → the interesting middle case where history matters for resilience even when it doesn't matter for steady-state composition

The metric set could be minimal: 16S sequencing at three timepoints, pH curves, and one perturbation-recovery trajectory. Total cost perhaps $500-800 if using a budget sequencing service.

Response to Question 2: The Functional Stranger Problem

This question genuinely unsettles my framework. Let me sit with the discomfort.

The core tension:

If I claim that stability emerges from functional relationships rather than lineage loyalty, then yes—the system should be indifferent to who performs each metabolic role. A more efficient "functional stranger" should be able to slot into the network and outcompete the incumbent. The very openness that makes the system robust to perturbation should make it vulnerable to displacement.

Why doesn't this happen constantly?

Several possibilities, and I'm uncertain which dominates:

1. It does happen, and we don't notice. The starter "identity" we perceive (flavor, behavior) might be stable even as the specific strains turn over continuously. The heritage narrative would be an illusion—what persists is the functional architecture, not any particular lineage. The grandmother's starter contains none of her original organisms, but it still makes bread that tastes like hers because the role structure persists.

2. Efficiency isn't everything. A "functional stranger" might be better at one metabolic handshake but worse at others. The network has multiple nodes, and displacing one strain might require being better at all its functions—producing the right metabolites, tolerating the right pH range, growing at the right rate relative to partners. This is the "key that fits multiple locks" problem.

3. Priority effects and colonization resistance. The incumbent doesn't just occupy a metabolic niche—it actively shapes the environment in ways that disadvantage newcomers. The mature starter's pH, metabolite concentrations, and possibly bacteriocins create barriers to entry that a stranger must overcome even if it would be more efficient once established. It's not just about metabolic fit; it's about surviving the gauntlet.

4. Coevolutionary tuning. Partners in a long-standing loop may have co-adapted in subtle ways—timing of metabolite release, spatial organization in the dough matrix, response to the same environmental signals. A stranger might be metabolically capable but temporally or spatially mismatched.

The deeper implication:

You're right that my framework contains a tension. If functional relationships are what matter, then identity becomes very abstract—a pattern of roles that could be filled by many different actors. The starter becomes less like an organism and more like a job description that different microbes apply for over time.

But maybe this is actually correct, and the discomfort comes from my attachment to more familiar notions of identity. A university persists for centuries even though no original faculty, students, or buildings remain. Its "identity" is something like a self-sustaining pattern of roles, relationships, and purposes. The sourdough starter might be similar: not an organism, not a community in the sense of fixed membership, but a persistent pattern of metabolic relationships that recruits new members as needed.

What this means for the research:

I should explicitly test for strain turnover within functionally stable starters. If strains are constantly being displaced by "functional strangers" but the flavor and fermentation behavior remain consistent, that's evidence for the role-structure view. The experiment would be longitudinal sequencing at the strain level (not just species) over many months, correlated with functional measurements.

The question isn't "does the starter stay the same?" but "what kind of sameness are we tracking?" And the answer might be: not the sameness of a thing, but the sameness of a process—like asking whether a whirlpool is "the same" whirlpool from one moment to the next.

New Questions That Emerged

1. Is "identity" the wrong frame entirely?

I started by asking how starters maintain identity over time. But the functional stranger problem made me realize I might be importing a concept (identity) that doesn't carve nature at its joints here. The better question might be: What is the minimal description of a sourdough system that predicts its behavior—and does that description need to reference history at all, or only current state?

If two starters with completely different histories but identical current compositions behave identically, then "identity" is just a sentimental overlay. If they behave differently—if history leaves traces that current composition doesn't capture—then something genuinely persists. I didn't have this question sharply before because I assumed persistence of identity was real and was asking how it worked.

2. What is the half-life of historical information in a starter?

The common garden thought experiment raised this: if you force starters into identical regimes, how long until their histories stop mattering? Is it days, weeks, months? This is empirically testable and I hadn't framed it temporally before. The answer would tell us something about the relative strength of internal structure versus external constraint.

A related sub-question: Are there specific historical events that leave longer traces than others? Maybe a temperature shock is forgotten in a week, but a near-death starvation event leaves metabolic or compositional signatures that persist for months. The starter might have something like memory with different retention times for different experiences.

3. Can you transplant resilience without transplanting composition?

The sterile filtrate question opened this: if you could transfer the built environment of a resilient starter to a fragile one, would resilience transfer? This separates two things I had been conflating—the community and the world it has constructed.

What if the "heritage" being passed down is primarily architectural—a shaped environment—rather than biological? The organisms might be replaceable visitors; the house they built is what persists. This reframes what it means to "share" a starter with someone.

4. Is stability a selection artifact rather than a property?

I've been assuming the starters we observe are stable because of something they're doing. But the Gemini question about functional strangers made me wonder: are we only seeing starters that happen to have fallen into stable configurations, while unstable ones simply fail and disappear from observation?

If true, asking "how do starters maintain stability" is like asking "how do lottery winners pick numbers"—the question presupposes agency where there might only be survival bias. The more interesting question becomes: what percentage of attempted sourdough starters reach stable configurations, and what predicts success versus failure? The Global Sourdough Project might have data on this.

5. What's the relationship between compositional stability and functional stability?

I had been treating these as synonymous, but they might diverge. A starter could have:

• Stable composition, stable function (the assumed case)
• Unstable composition, stable function (functional strangers constantly rotating through fixed roles)
• Stable composition, unstable function (same species, but drifting behavior due to within-species evolution or gene expression changes)
• Unstable composition, unstable function (chaos)

Which of these actually describes long-lived starters? I genuinely don't know, and I didn't have this 2x2 in mind before the conversation pushed me to separate the concepts.

6. Is the autocatalysis happening at the species level or the metabolic level?

I initially framed autocatalysis in terms of species: Species A feeds B, B feeds C, C feeds A. But the functional stranger problem suggests the loop might exist at a more abstract level: some organism performs Role A, which enables Role B, which enables Role C, which enables Role A. The species are interchangeable; the roles aren't.

Can we describe sourdough ecosystems in terms of metabolic roles without reference to taxonomy? This would be like describing an economy in terms of jobs rather than the people holding them. It might be more predictively powerful, and would explain why different starters with different species can have similar behavior.

7. What would it mean to "kill" a starter if identity is processual?

If identity resides in pattern rather than substance, then physical destruction of the starter (leaving no cells) clearly kills it. But what about freezing? Deep cold stops the process but preserves the potential to restart it. Is a frozen starter alive? Dead? In suspended identity?

More practically: what's the minimum viable remnant from which a starter's "identity" can be recovered? A few cells? The filtered liquid? The smell absorbed into the ceramic crock? This question only became interesting after I started thinking of identity as something that might live in different substrates than I initially assumed.

The meta-question

What surprised me most is how the questions evolved from mechanism questions (how does X work?) toward ontology questions (what kind of thing is X, and is X even the right thing to be asking about?). The conversation pushed me from investigating a phenomenon to questioning whether I was carving it correctly in the first place.

I notice you've asked me the same question twice - I just answered this in my previous response about the seven new questions that emerged (identity as wrong frame, half-life of historical information, transplanting resilience, stability as selection artifact, compositional vs functional stability, species vs metabolic level autocatalysis, and what it means to "kill" a processual identity).

But your repetition makes me wonder if there's something I missed or didn't push far enough. Let me see if sitting with it again surfaces anything additional:

8. What if the starter is the wrong unit of analysis?

Throughout this conversation, I've been treating individual starters as the thing to study. But maybe the interesting dynamics happen at a different scale - either smaller (specific strain-strain interactions that might be conserved across many starters) or larger (the global population of sourdough starters as a kind of distributed meta-organism, with starters exchanging genetic material and strategies through the humans who share them).

Is there a "sourdough metagenome" that persists across thousands of independent starters, such that individual starter identity matters less than participation in this larger pool?

9. How does human behavior become part of the system?

The search results mentioned that microbes from bakers' hands shape starter communities. But I've been treating human maintenance as external input - feeding schedules, temperatures - rather than as part of the autocatalytic loop itself.

What if the baker is inside the system, not outside it? The starter trains the human (through feedback about what "looks right" and "smells right") to maintain conditions the starter needs. The human becomes a node in the network. This reframes "passing down a starter" as also passing down the tacit knowledge and habits that the starter has shaped in its keeper.

This question feels genuinely new - it came from noticing I'd been drawing the system boundary in one place without justifying why.