eduKateSG · VOYNICH RESEARCH LIBRARY · TANGENTIAL VOYNICH XII
Tangential Voynich | Treat the Manuscript as a Water Network
Sources, reservoirs, branches, valves, pressure zones, leakage and mixing. Not because the Voynich tubes are pipes, but because water remembers the path it took in ways electricity often does not.
← Tangential Voynich: The Wrong-System Protocol · Previous False World: Power Grid · Voynich Research Library
A power grid gave us distributed balance.
A water network changes the problem.
Water has direction. It can be stored. It can split. It can merge. It can remain in one compartment while another compartment is isolated. It can carry traces of what entered upstream. It can leak. It can contaminate neighbouring flows. A valve can change reachability without changing the physical network. A reservoir can buffer short-term disturbances. Pressure can reveal hidden constraints. A downstream sample can preserve information about multiple upstream sources.
Those properties make water different from the previous wrong worlds.
The Voynich Manuscript is not a hydraulic network.
Its tubes are not assumed to be pipes. Its pools are not assumed to be reservoirs. Its human figures are not assumed to be users of a water system. Its pages are not assumed to represent baths, plumbing, fountains or anatomy.
In fact, because the manuscript visibly contains tube- and fluid-like imagery, this wrong world is more dangerous than the railway or compiler. A modern water-network metaphor can too easily slide into a historical hydraulic story.
The water network is useful only if every surviving result can be stated without water, pipes, baths or hydraulics.
That gives us the experiment.
FALSE-WORLD CONTRACT
Tube-Like Is Not Pipe-Like
- A drawn tube is not a pipe.
- A pool-like form is not a reservoir.
- A branching diagram is not a plumbing network.
- A repeated glyph is not a valve state.
- A boundary is not a shut-off valve.
- A page cluster is not a pressure zone.
- A mixing pattern is not a chemical mixture.
- A missing relationship is not leakage.
- A one-way statistical transition is not physical flow.
- A successful flow model does not establish hydraulic, medical, bathing or anatomical meaning.
The historical water-conduit question remains a separate comparator problem. See Voynich and Padua | The Water-Conduit Problem: Pipes, Pumps and the Danger of Reading the Tubes Literally.
Flow ≠ water. Tube-like ≠ pipe. Compartment ≠ reservoir. Pressure ≠ meaning.
What the Water Network Adds
The Power Grid asked whether local states preserve a distributed balance.
The Water Network asks what the Power Grid could not isolate as cleanly:
does the state of something depend on the path by which it arrived?
That creates a new candidate geometry:
flow topology.
Flow topology is not simply connectivity. It includes:
- direction;
- branching;
- merging;
- storage;
- compartments;
- one-way reachability;
- path-dependent composition;
- loss;
- mixing;
- isolation.
For Voynich, the critical question becomes whether two local states that look similar behave differently because they arrived through different structural paths.
Sources: A Downstream Population May Have More Than One Upstream Origin
A water system can receive input from several sources.
The downstream network may therefore contain a mixture whose composition depends on which sources were active.
This creates a useful adversarial model for manuscript populations.
A page class may combine:
- one scribal tradition;
- another visual exemplar tradition;
- a shared layout convention;
- a third distributional regime.
That does not mean the page was literally assembled from four workshops.
It means a single population can inherit several independent structural ancestries.
The Supply Chain called this multi-sourcing.
The Water Network adds a measurable question: can the downstream population be decomposed into stable upstream signatures?
Reservoirs: Storage Can Decouple Input From Output
A reservoir means water entering now does not have to leave now.
Storage separates input timing from output timing.
The manuscript analogue is delayed structural release.
Could one local structure accumulate context and only express its effect later?
For example, a paragraph-start state might influence a later line-final choice even after several ordinary positions intervene.
Test models with latent stored state against purely local continuation.
delayed influence should be distinguished from immediate transition.
If delayed state adds no predictive power, the reservoir metaphor fails.
Pressure Zones: The Same Network Can Operate Under Different Local Constraints
Water networks often contain zones with different pressure conditions.
The pipes may connect, yet local operating constraints differ.
This is a useful model for Currier, hand and interface populations.
The question is not whether they differ.
The question is whether they share a network while operating under different local parameter ranges.
Compare the same structural family across populations for:
- frequency;
- neighbour distribution;
- boundary preference;
- recurrence distance;
- effective context scale.
If the family preserves identity but changes its operating envelope, a shared-system/different-zone model becomes useful.
Valves: Connectivity and Reachability Are Different
Two parts of a water network can be physically connected while a closed valve makes one unreachable from the other.
This gives Tangential Voynich a precise distinction:
structural adjacency does not guarantee structural reachability.
A candidate boundary, marker or state may open or close classes of continuation.
Test whether the same neighbour becomes reachable only under particular preceding states.
This extends the Airport’s conditional-clearance test and the Compiler’s scope rules.
The neutral residue is state-dependent reachability.
Check Valves: Some Relationships May Be One-Way
A one-way valve allows flow in one direction and blocks the reverse.
Supply Chain and Airport already taught us to measure directional asymmetry.
The Water Network makes the hostile test sharper.
For every strong A→B relationship, explicitly test B→A under matched opportunity.
One-way candidates should survive:
- frequency control;
- position control;
- page-regime control;
- boundary control;
- held-out material.
The neutral result is directional reachability, not flow semantics.
Branching: One State Can Produce Several Downstream Families
A network branch turns one upstream path into several downstream paths.
Voynich can be tested for the same abstract property.
Does one structural state predict several distinct continuation families that remain locally coherent after divergence?
This is stronger than saying a form has many followers.
A true branch should create downstream clusters with persistent differences.
If those differences vanish immediately, ordinary stochastic variation is enough.
Merging: Different Paths Can Produce One Indistinguishable State
Two water sources can merge downstream.
After enough mixing, the immediate downstream state may no longer reveal which path each molecule took.
This creates a difficult identifiability problem for Voynich.
Can two different upstream structural histories converge on the same visible local form?
If yes, local observation alone may be insufficient to infer history.
That gives us another explicit status:
history-non-identifiable state.
This complements the Power Grid’s unobservable-state category.
Mixing: A Local State Can Preserve a Fractional Ancestry
Mixing is where the Water Network becomes genuinely different from the Power Grid.
A downstream sample can contain measurable contributions from multiple upstream sources.
The Voynich analogue is a mixed structural population.
A page or subcorpus may not belong cleanly to class A or class B. It may combine proportions of several regimes.
Instead of forcing hard assignment, estimate mixture weights.
P(page) ≈ αA + βB + γC, with α+β+γ=1.
This is especially natural for pages where hand, Currier, morphology and geometry disagree.
The neutral return is not “mixed water.”
It is soft membership across structural regimes.
Residence Time: Some Information May Remain Local Longer Than Other Information
Water can spend different amounts of time in different parts of a system.
The abstract manuscript question is persistence by compartment.
Do some local states persist across many positions while others dissipate quickly?
Estimate context half-life for different features.
- hand-conditioned effects;
- paragraph-start effects;
- rare-event effects;
- label-interface effects;
- page-regime effects.
This refines the existing effective-context and perturbation-decay measurements by allowing persistence to vary by structural compartment.
Pressure: Constraint Can Be Inferred From What a System Is Forced to Do
Pressure is not visible simply because water exists.
It is inferred from behaviour under constraint.
The Voynich analogue is structural pressure.
When line width shrinks, illustration space tightens or competing classes crowd one position, what adaptations appear?
- shorter forms;
- different spacing;
- suppressed rare variants;
- greater use of compact families;
- earlier closure.
The important neutral quantity is response to capacity pressure.
Telecommunications, Warehouse, Airport and Power Grid have all approached this from different directions.
Water Network adds a gradient model: how strongly does behaviour change as the constraint increases?
Pressure Loss: Influence Can Decay Along a Path
Pressure decreases along real networks because of resistance and losses.
Strip away hydraulics and we get a familiar but useful question:
does structural influence decay with path length?
Measure the predictive contribution of an upstream state at increasing distances.
Does it fall smoothly?
Does it persist until a boundary and then collapse?
Does it reappear after a recurrent marker?
The shape of the decay curve may distinguish ordinary memory from compartment boundaries.
Leakage: Missing Structure Can Be Inferred From Mass-Balance Failure
A water utility can suspect leakage when measured inputs and outputs do not reconcile.
This suggests a powerful neutral test for missing manuscript structure.
Suppose a structural model predicts a bounded set of counterparts, but observed members systematically fail to close the accounting.
The missing amount might reflect:
- lost leaf;
- damaged locus;
- transcription omission;
- unmodelled branch;
- wrong unit definition;
- wrong model.
The crucial move is to calculate the discrepancy before naming the cause.
unreconciled structure is an observation; “leakage” is only one temporary explanation.
This strengthens Database orphan relations and Telecommunications typed loss.
Contamination: One Local Addition Can Leave a Downstream Signature
Contamination is another dangerous word, so the metaphor must be controlled.
The neutral question is whether a local intervention leaves a measurable downstream trace.
Examples might include:
- a scribal correction style;
- a later annotation convention;
- a repeated inserted marker;
- a visual modification layer.
If one intervention class predicts later local changes, its propagation can be mapped.
This should never become “contamination” in the historical claim ledger unless the physical evidence supports that word.
The neutral result is propagating intervention signature.
Compartment Isolation: Boundaries Can Prevent Mixing
Water systems often isolate zones to prevent undesired mixing or to manage pressure.
For Voynich, test whether some boundaries sharply reduce cross-population blending.
Compare the similarity of neighbouring units across:
- ordinary line boundaries;
- paragraph boundaries;
- page boundaries;
- hand boundaries;
- Currier boundaries;
- morphology boundaries.
A strong compartment boundary should reduce information transfer across it more than matched ordinary boundaries.
This extends Power-Grid protection boundaries but changes the measurement from disturbance propagation to mixture suppression.
Cross-Connections: One Hidden Link Can Defeat the Apparent Compartment Model
Two zones can appear separate while a hidden cross-connection moves material between them.
This produces an adversarial test for Voynich sectioning.
If two populations are claimed to be independent, search specifically for:
- shared rare forms;
- shared unusual boundary structures;
- matching recurrence families;
- shared labels or diagram components;
- bifolio links crossing the proposed separation.
One robust cross-link may not destroy compartmentalisation, but it changes the model from isolated to coupled.
The correct state may be partially permeable boundary.
Backflow: Reverse Influence Needs Its Own Test
If a model assumes direction, reverse dependence becomes a hostile control.
For any proposed upstream/downstream relation, test whether downstream state predicts upstream state equally well.
If it does, the relation may be association rather than direction.
If reverse prediction is much weaker, directional structure gains weight.
This formalises a rule that should apply across the entire Tangential estate:
every directed model needs a reverse-direction challenge.
Dead Ends: Some Paths Terminate Without Returning
A branch can end.
For Voynich, terminal structure should be tested without assuming lexical closure.
Do some local sequence families terminate in restricted states from which ordinary continuation does not resume?
Railway called these terminals.
Compiler called them scope ends.
Power Grid called them isolated states.
Water Network asks whether they are absorbing endpoints in the transition graph.
Recirculation: Some Paths Can Return to Earlier States
Some networks contain loops.
The Voynich question is whether transition structure contains stable cycles rather than only forward progress.
Search for recurrent state loops at:
- token-family level;
- line-state level;
- entry-state level;
- page-population level.
A true cycle should recur more often than matched shuffled controls and preserve ordered structure.
This may eventually connect to the planned Musical Score tangent, where cyclical return will be asked from a very different domain.
Buffering: Storage Can Smooth Short-Term Variation
Reservoirs and tanks can reduce short-term variation between inflow and demand.
The neutral manuscript question is whether one level of organisation smooths variability from a lower level.
For example:
- glyph-level variation may be high while token-family distribution remains stable;
- line-level variation may be high while paragraph-level composition remains stable;
- page-level variation may be high while hand-level distribution remains bounded.
This yields a multiscale smoothing test.
higher-order organisation may buffer lower-order variability.
Demand Peaks: Rare Capacity States Can Expose Hidden Rules
A water network may behave normally until demand approaches a limit.
Extreme states reveal constraints invisible during ordinary operation.
For Voynich, focus on high-pressure structural cases:
- very crowded lines;
- tiny label spaces;
- dense diagram sectors;
- unusually long apparent tokens;
- highly repetitive lines.
Which features break first?
Which remain invariant?
The Power Grid called this load shedding.
The Water Network reframes it as stress-response under capacity limits.
Sampling Points: Where You Measure Changes What You Think the System Contains
A water sample taken near a source can differ from one taken after storage and mixing.
Voynich statistics are equally sensitive to where we sample.
A corpus-level frequency may hide:
- line-position gradients;
- page-interface differences;
- hand-specific regimes;
- local repeated families;
- boundary-conditioned changes.
Every reported distribution should therefore preserve its sampling coordinate.
the place where a statistic is measured is part of the statistic.
This is another clean bridge to the representation branch.
The Human Figures, Pools and Tubes: The Most Dangerous Place to Use This Tangent
The water-network analogy becomes most seductive on the manuscript pages containing human figures, pools, tubes and basin-like forms.
That is exactly where the firewall must become strongest.
We are allowed to measure:
- branching geometry;
- closed versus open paths;
- component repetition;
- directional cues if independently coded;
- junction count;
- compartment count;
- connectivity between basins and tubes.
We are not allowed to conclude from those measurements that the page depicts plumbing, baths, anatomy, circulation or any other literal flow system.
the closer the image resembles the borrowed metaphor, the more aggressively the metaphor must be removed.
The historical Water-Conduit Problem remains the appropriate place for literal hydraulic comparators.
Reverse Tangential Test: Hide a Real Water Network
The Water Network tangent earns credibility only if the same methods can recover known flow structure after ordinary engineering labels are removed.
Take a real or simulated distribution network.
Remove pipe names, reservoir names, valve labels, geography and engineering categories.
Preserve anonymised:
- network topology;
- directional flow observations;
- storage states;
- pressure-like node variables;
- mixing fractions;
- valve state changes;
- loss events;
- contamination traces.
Can Voynich-style methods recover:
- source populations;
- reservoir-like delayed state;
- pressure zones;
- one-way reachability;
- branch points;
- mixing regions;
- isolated compartments;
- hidden losses;
- path-dependent downstream composition?
Then damage the control.
Hide valves.
Delete sensors.
Merge compartments.
Add false cross-connections.
If path and compartment structure cannot be rediscovered there, the water metaphor should not survive on Voynich.
A method that cannot recover hidden flow topology in a known network should not invent flow topology in an unknown manuscript.
What Would Make the Water-Network World Fail?
| Water-network projection | Failure condition | Neutral residue |
|---|---|---|
| Path dependence | Upstream history adds no predictive value after current local state is known. | Local state may be sufficient. |
| Storage | Delayed context does not improve prediction. | No latent stored state. |
| Pressure zones | Candidate regimes share no stable identity under different local parameters. | Populations may be genuinely separate. |
| Valves | Reachability does not depend on boundary or marker state. | No state-dependent gating. |
| One-way flow | Reverse direction predicts as well as forward direction. | Relation remains undirected. |
| Mixing | Soft mixture models do not outperform hard classification. | Mixed-membership model unnecessary. |
| Leakage | Structural discrepancies vanish after better unit definition or model change. | No evidence for missing counterpart. |
| Compartment | Proposed boundaries do not suppress cross-region blending. | No compartment isolation. |
| Flow topology | Path, storage and mixing variables add no predictive value. | Connectivity alone may suffice. |
The Water Network should fail wherever those conditions hold.
Failure is the expected control, not an embarrassment.
The Water-Network Experiment Pack
- Path-dependence test: compare prediction from current state alone versus current state plus structural history.
- Source-mixture analysis: test whether downstream populations decompose into stable upstream signatures.
- Delayed-state test: estimate whether earlier boundary or launch states retain influence after intervening positions.
- Pressure-zone comparison: track the same structural family across different manuscript regimes.
- State-dependent reachability: test whether candidate transitions become available only under particular boundary or marker states.
- Reverse-direction challenge: test every directional relation in both directions.
- Persistent-branch test: require divergent continuations to remain distinct downstream.
- Convergence test: identify visually identical local states produced by different upstream histories.
- Soft-membership model: compare mixed structural ancestry against hard page classification.
- Context half-life: estimate persistence separately for different structural compartments.
- Constraint-gradient model: measure how behaviour changes continuously with spatial pressure.
- Path-decay curve: measure influence as a function of structural distance.
- Structural reconciliation: calculate expected-versus-observed discrepancy before assigning a missingness cause.
- Intervention-propagation scan: test whether verified later or corrective layers leave downstream signatures.
- Compartment-isolation test: measure suppression of cross-boundary mixture.
- Hidden-cross-link challenge: search hostilely for rare shared structures across proposed compartments.
- Absorbing-endpoint scan: identify terminal states with little ordinary continuation.
- Cycle detection: test recurrent ordered loops against shuffled controls.
- Multiscale buffering test: ask whether higher levels smooth lower-level variability.
- Extreme-capacity test: inspect crowded or constrained regions for hierarchy of preserved features.
- Sampling-coordinate ledger: preserve exactly where every statistic was measured.
- Tube-image firewall: code geometry without hydraulic naming on fluid-like visual pages.
- Reverse hidden-water-network calibration: anonymise and damage a known distribution network.
- Held-out Voynich gate: freeze unseen material before flow thresholds are tuned.
- Metaphor removal: rewrite every surviving result without water-network vocabulary.
Eleven Wrong Systems, Five Hidden Geometries
The false-world estate now reaches eleven systems:
- Railway: movement.
- Operating system: state.
- Compiler: representation.
- Database: relationship.
- Filesystem: logical versus physical organisation.
- Telecommunications: protocol versus payload.
- Warehouse: retrieval.
- Global supply chain: dependency.
- Airport control tower: coordination.
- Power grid: balance.
- Water network: path-dependent flow and compartment.
The last five now define five distinct measurement geometries:
| Topology | Question |
|---|---|
| Retrieval topology | How is distributed structure found? |
| Dependency topology | What depends on what? |
| Coordination topology | How do local sequences coexist under shared constraints? |
| Balance topology | What higher-order condition constrains local behaviour? |
| Flow topology | Does present structure retain information about the path, storage, mixing and compartment history that produced it? |
None is a solution.
Together they are becoming a measurement language.
Metaphor Removal
Now remove the water network.
Remove sources, pipes, reservoirs, valves, pressure, leakage, contamination and flow.
What survives?
- Current local state can be tested for dependence on earlier structural path.
- One population may contain measurable contributions from several structural ancestries.
- Delayed context should be distinguished from immediate transition.
- The same structural family may operate under different parameter regimes.
- Adjacency and reachability are different properties.
- Directional relations require reverse-direction challenge.
- One state can branch into several persistent downstream regimes.
- Different histories can converge on one locally indistinguishable state.
- Some pages may be better represented by soft mixed membership than hard classification.
- Context persistence can vary by compartment.
- Behaviour can change continuously with spatial constraint pressure.
- Influence can be measured as a decay curve along structural distance.
- Expected-versus-observed discrepancy should be measured before assigning a missingness cause.
- Verified interventions can be tested for propagating signatures.
- Some boundaries may suppress structural mixture across them.
- Proposed compartments should be challenged by searching for hidden cross-links.
- Every directed model needs a reverse-direction control.
- Some states may be absorbing endpoints.
- Some state sequences may form recurrent cycles.
- Higher-order organisation can buffer lower-order variability.
- Extreme capacity conditions can reveal which constraints have priority.
- The sampling coordinate is part of every statistic.
- Tube- and fluid-like imagery can be coded geometrically without assigning hydraulic semantics.
The water disappears.
The path dependence remains.
World Return
The Voynich Manuscript is not a water network.
That sentence matters more here than in many earlier tangents because the manuscript contains imagery that can tempt us into literal hydraulic stories.
But after the pipes, reservoirs and valves are removed, a serious research question remains.
Does present structure remember the path by which it arrived?
Can one page contain fractional ancestry from several structural regimes?
Can boundaries block mixing while still allowing rare cross-links?
Can two different histories converge on the same visible local state, making the past unrecoverable from the present alone?
Can missing structure be inferred from failure to reconcile expected and observed relationships?
These are not hydraulic questions anymore.
The Power Grid asked whether the manuscript preserves balance. The Water Network asks whether it preserves memory of route.
That may prove false.
If it does, the failure belongs in the ledger.
If it survives, flow topology becomes another neutral object that no longer needs the metaphor that discovered it.
Tangential Voynich: The Wrong-System Protocol → · Power-Grid Experiment → · Historical Water-Conduit Control → · Voynich Research Library →
Next Tangent: treat the manuscript as a musical score. Motifs, rhythm, recurrence, transformation, phrase boundaries, transposition, variation, cadence and multi-voice structure will deliberately jump far away from infrastructure so the programme can test whether the same invariants survive a completely different representational world.
NEXT FALSE WORLD · MUSICAL SCORE
The Water Network asked whether present structure remembers the route that produced it. Music asks a different question: what transformations can occur before two structures should stop being treated as members of the same family?
Continue to Tangential Voynich | Treat the Manuscript as a Musical Score →