eduKateSG · VOYNICH RESEARCH LIBRARY · TANGENTIAL VOYNICH II
Tangential Voynich | Put the Manuscript in a Railway Control Centre
Routes without stations. Terminals without destinations. Interchanges without names. What does an unread manuscript look like when a railway controller is forbidden to translate it?
← Tangential Voynich: The Wrong-System Protocol · Voynich Research Library
Imagine that the Voynich Manuscript is carried into a modern railway operations control centre.
Nobody in the room is allowed to call a glyph a letter.
Nobody is allowed to call a token a word.
Nobody is allowed to identify a plant, a zodiac sign, a bath, a recipe, a city or a person.
The railway team receives only an unidentified system containing repeated units, boundaries, lines, page regions, visual structures, several proposed scribal populations and a physically complicated codex.
Then we ask them a deliberately unreasonable question:
If this behaved like a railway, what would you measure?
Not: “Is Voynich a railway?”
It is not.
The railway did not exist in this form when the parchment was made. The exercise makes no claim about historical function, language, authorship or provenance.
The railway is an alien measuring instrument.
Its value is that railway control has spent generations learning to think about sequence, capacity, route conflict, terminal behaviour, local delay, network propagation, transfer points, recovery and state. Those concepts generate questions that a botanist, cryptographer or manuscript historian may never naturally ask.
Then, at the end, we remove every railway word.
Whatever survives may belong to the manuscript.
FALSE-WORLD CONTRACT
The Railway Is Not the Answer
- A recurring token is not a station.
- A line of script is not a railway line.
- A paragraph boundary is not a terminus.
- A foldout is not a network map.
- A common token is not an interchange merely because it has many neighbours.
- A transition probability is not a train movement.
- A successful graph analogy does not establish manuscript function.
All railway labels in this article are temporary aliases. Their only job is to generate discriminating predictions. Every surviving result must later be rewritten without them.
This is the first real Tangential Voynich stress test.
What a Railway Controller Sees
A passenger sees a railway as stations and destinations.
An operations controller sees a different object.
The controller sees occupancy, direction, headway, dwell, route setting, conflict, capacity, turnaround, delay propagation, recovery margin, transfer demand, depot availability and state transitions.
That distinction matters.
The Voynich equivalent of the passenger is the reader asking, “What does this word mean?”
The railway-control thought experiment asks something else:
- What can occur next?
- Which transitions are common?
- Which transitions are rare or forbidden?
- Where does state reset?
- Which units connect otherwise separated regions?
- Which local perturbations propagate?
- Where do sequences terminate?
- Do repeated pathways exist?
- Are there multiple operating regimes sharing infrastructure?
Those are not translation questions.
They are architecture questions.
eduKateSG’s existing How MRT Works | It’s Mathematics estate already examines headway, dwell time, braking, power, fleet operations, passenger information and network behaviour as interacting mathematical systems. Tangential Voynich borrows that systems vocabulary temporarily and turns it against an unread codex.
First Translation Ban: Replace Meaning With Movement
When an unknown string appears repeatedly, the ordinary instinct is to ask what the string means.
A railway controller asks where it can go.
For a temporary model, let each recurring form be a node. Let observed adjacency create a directed edge. Let the weight of that edge equal the number or probability of observed transitions.
G = (V, E, W)
Here V is a set of observed units, E is the set of observed transitions, and W records their weights.
This model immediately removes a dangerous question—“what does this form translate to?”—and replaces it with several measurable ones.
- What is the node’s in-degree?
- What is its out-degree?
- Does it connect many communities?
- Does it appear only after particular boundary states?
- Does it have directionally asymmetric neighbours?
- Does its local transition pattern vary by Currier regime or scribal hand?
- Does it remain central when high-frequency effects are normalised?
This may reveal routing structure without ever asserting language.
But there is a trap: high-frequency forms will naturally acquire many neighbours. A busy “station” can be nothing more than a common token.
So degree alone is not enough.
The railway tangent must ask whether a form connects neighbourhoods that would otherwise remain separate. That is closer to bridge centrality than mere frequency.
The Interchange Test
In a railway, an interchange is interesting because it connects route systems.
If a Voynich form is temporarily treated as an “interchange,” the railway analogy must predict more than high frequency.
It should disproportionately bridge otherwise separated neighbourhoods.
One possible neutral statistic is betweenness centrality. For a node \(v\):
Cᴮ(v) = Σ σₛₜ(v) / σₛₜ
where σₛₜ counts shortest paths between nodes \(s\) and \(t\), and σₛₜ(v) counts those passing through \(v\).
The exact metric may not be the final choice; shortest-path assumptions can be inappropriate for text. But the railway world has generated a better question:
Are there recurring forms whose structural importance comes from connecting distinct local populations rather than simply appearing often?
That question survives railway removal.
If the answer is yes, those forms deserve separate analysis under linguistic, generative and historical models.
The Terminal Test
A terminal is not merely the final station on a map.
Operationally, terminals are special because vehicles may turn around, crews may change, schedules recover, routes reset, and constraints governing departure can differ from constraints governing ordinary intermediate movement.
That makes terminal thinking useful for Voynich line and paragraph boundaries.
Suppose line endings and beginnings are temporary “turnaround zones.” What should happen?
- The distribution immediately after a boundary should differ from interior positions.
- Some forms should disproportionately occupy terminal or launch positions.
- Cross-boundary dependence may differ from within-line dependence.
- The magnitude of the reset may vary between paragraph boundaries and ordinary line breaks.
- Some manuscript populations may show stronger reset behaviour than others.
Those predictions are directly compatible with the existing Voynich boundary research lane. They do not add a railway meaning. They add a control-system question.
In neutral language, the test becomes:
Does local continuation probability undergo a measurable discontinuity at defined manuscript boundaries?
If yes, the railway metaphor has produced transferable yield.
If no, the terminal analogy dies.
Headway: Stop Looking Only at What Appears
Railway capacity is not determined merely by how many trains exist. Timing between trains matters.
Headway is the time separation between successive services. Small headways raise capacity but reduce recovery margin; irregular headways can amplify crowding and delay.
Voynich has no literal trains, but the concept forces us to notice spacing and recurrence intervals.
Instead of asking only how frequently a form appears, ask how its appearances are spaced.
- Does it recur at nearly regular intervals?
- Does it burst locally and then disappear?
- Does recurrence distance change by page region?
- Do related forms alternate?
- Does one family appear to suppress another at short distance?
- Do long gaps carry different information from short gaps?
This connects naturally to intermittency and long-range dependence already studied in Voynichese. The railway tangent adds the concept of spacing as operating state rather than frequency as a static count.
A common form with irregular recurrence is structurally different from a common form with periodic recurrence.
Frequency hides that difference.
Headway thinking exposes it.
For the railway mathematics itself, see How MRT Headway Works Using Mathematics.
Dwell Time: The Boundary Can Dominate the Network
A train spends only part of its journey moving. At stations, doors open, passengers exchange, doors close and safe departure must be established. A few extra seconds of dwell at a busy node can propagate into following services.
The Tangential question is not whether Voynich tokens “dwell.” It is whether some boundary regions consume more structural freedom than others.
For example, a boundary might impose:
- a narrower next-symbol distribution;
- stronger preference for one family;
- more stereotyped opening sequences;
- longer or shorter apparent tokens;
- different ambiguity rates;
- different relation to neighbouring labels or illustrations.
In a railway, a station can constrain network throughput more strongly than open track.
In an unknown symbolic system, a boundary may similarly carry more information than the interior sequence.
That possibility is especially interesting because recent 2026 work argues that some of the manuscript’s order may concentrate at token edges and separator regimes rather than in token-to-token identity succession.
The railway did not create that evidence.
It gives us another reason to take boundary infrastructure seriously.
Research context: Rozanova & Temerev 2026 · railway analogue: How MRT Station Dwell Time Works Using Mathematics.
Multiple Lines Sharing One Network
A metro system may contain several service lines using related engineering standards while occupying different routes. Some infrastructure may be shared; some is separate. A passenger can move from one line to another without implying that the lines are the same system.
This creates a useful alien model for the manuscript’s overlapping populations.
Currier A and Currier B differ statistically. Proposed scribal hands overlap with—but do not simply equal—visual classes and codicological units. The frozen eduKate segmentation matrix likewise resisted collapse into one universal section map.
The railway analogy asks:
- Could multiple local rule regimes share part of one symbol inventory?
- Could some forms behave like network-wide infrastructure while others are line-specific?
- Could a physical bifolio carry one production regime while its text distribution belongs to another statistical community?
- Where are the transfer zones among those regimes?
This is not a claim that Currier A and B are “railway lines.”
The neutral return is more useful:
Model the manuscript as overlapping rule populations sharing some infrastructure rather than forcing every classification into one partition.
That survives metaphor removal and connects directly to the existing hand × bifolio × section work.
Related: Hand × Bifolio × Section.
Route Conflict and Forbidden Transitions
Railway signalling is built around more than allowing movement. It must also prevent incompatible movements.
That produces a subtle Tangential question:
What does the manuscript avoid?
Frequency studies often emphasise what occurs. Constraint systems are equally defined by what does not occur where it could have.
For every common form or glyph family, construct a possibility set from its broader distribution. Then ask whether particular local combinations are absent more often than chance, transcription or page composition would predict.
This is analogous to route locking: not everything physically adjacent is operationally permitted.
In neutral language, the research question becomes:
Are there statistically stable exclusion constraints among local forms after controlling for frequency, position and manuscript population?
A strong yes would be useful under many competing models: grammar, cipher state, formal notation, copy-generation procedures or production templates.
The railway interpretation is unnecessary once the question exists.
Delay Propagation: Does a Local State Echo Forward?
Railway delay is rarely local. A late train can alter following headways, platform crowding, junction order and turnaround recovery. One disturbed state can propagate.
The Voynich analogue asks whether unusual local states have measurable downstream effects.
Suppose a rare form occurs. Does the distribution of the next one, two, three or five positions differ from a matched baseline? If a line begins unusually, does the effect disappear immediately or persist? Does the decay length vary across manuscript populations?
One might model a perturbation response:
R(k) = divergence at distance k from a defined local event.
If \(R(k)\) collapses at \(k=1\), the state is extremely local. If it persists, the manuscript carries longer structural memory.
This can be tested without deciding whether the underlying mechanism is linguistic, cryptographic or generative.
The railway world has simply changed our temporal scale.
Recovery Margin and Redundancy
Robust transport systems contain slack.
A timetable with no recovery margin may appear maximally efficient until the first disruption. Redundancy and spare capacity can make a system less compact but more resilient.
Voynichese is often described as repetitive or redundant. The railway tangent asks a different question:
Could some repetition function structurally as tolerance rather than semantic duplication?
We must be careful. This is not a claim that the manuscript has error-correcting redundancy. It is a prompt to test whether repeated families make local sequences robust to substitution, omission or segmentation differences.
For example:
- If one apparent unit is removed, does a local family classification remain stable?
- If uncertain separators are collapsed, do the same higher-order structures survive?
- Do related variants occupy similar neighbourhoods?
- Can a sequence tolerate local graphical variation without changing its broader structural class?
If yes, the system may possess representational robustness.
Again, the mechanism remains open.
The Depot Problem: Where Does the System Go When It Is Not Running?
Passengers rarely see depots as part of the railway’s intellectual structure. Operations staff cannot ignore them. A railway’s visible service depends on off-line storage, inspection, preparation, maintenance and re-entry.
This creates a useful question for manuscripts: are there regions that participate in the system without behaving like ordinary running text?
Labels, marginal forms, diagram text, star-marked short entries, foldout inscriptions and other interface populations may have different operational roles from paragraph prose.
Calling them “depots” would be nonsense.
But depot thinking forces a useful distinction between network-participating and sequence-participating.
A label may belong to the same symbolic ecosystem without participating in the same transition grammar as running text.
This creates a direct test: compare symbol and unit distributions across running text, labels, diagram inscriptions and other interface classes while preserving transcription uncertainty.
The railway analogy dies.
The population distinction survives.
Related Voynich lane: Labels Versus Running Text. Railway systems counterpart: How MRT Depots and Fleet Operations Work Using Mathematics.
The Passenger-Information Problem: A Correct Signal Can Still Fail Its Receiver
A railway is not operated only by physical movement. Information must arrive at the right receiver at the right time. A correct message delivered after the train has departed is operationally wrong.
This brings Tangential Voynich back to the receiver problem.
A manuscript’s information may be perfectly structured for a trained fifteenth-century receiver and almost useless to us because the reading convention, technical context, oral instruction, exemplar lineage or professional vocabulary has disappeared.
The question becomes:
Which parts of the manuscript are self-describing, and which appear to require an external receiver model?
A diagram whose labels cannot be interpreted independently may require a decoder supplied by training. A repeated visual grammar may function as retrieval for an expert rather than explanation for a novice. A densely constrained script may be efficient only for a community that already knows its expansion rules.
The railway does not tell us which of these is true.
It reminds us that successful information systems are receiver-dependent.
Related railway article: How MRT Communications and Passenger Information Work Using Mathematics.
The Foldout as Network View—Then Destroy the Idea
A railway controller needs a view larger than the passenger’s immediate position. Network-wide state matters.
The Voynich foldouts tempt us to make exactly that leap: perhaps the larger surface is a global map or system overview.
This is a good Tangential hypothesis because it is also easy to overclaim.
If a foldout functions like a network overview, it should exhibit specific structural relationships absent from ordinary pages:
- more explicit connectivity;
- stable correspondence among repeated regions;
- higher-order relationships among components otherwise shown locally;
- systematic labels associated with nodes or sectors;
- topological constraints that survive rotation or scale.
If these properties are not present, “network overview” should be rejected.
The larger page is not automatically a map. The existing Voynich programme already protects this boundary.
The Tangential contribution is to sharpen what a genuine overview interface would need to do.
Related: The Foldout Interface Problem: Why a Bigger Page Is Not Automatically a Map.
Reverse Test: Hide Singapore’s MRT Labels
Now reverse the experiment.
Take a railway system we understand.
Remove names, line colours and geographic labels.
Keep only anonymised topology, direction, service intervals, transfer counts, terminal behaviour, dwell distributions and disturbance propagation.
Could our Voynich methods reconstruct useful truths?
- Could clustering recover route families?
- Could centrality identify interchange-like nodes?
- Could edge asymmetry identify direction?
- Could boundary statistics identify terminals?
- Could perturbation analysis identify bottlenecks?
- Could population analysis separate service regimes?
If the answer is no, those methods should be treated cautiously on Voynich.
If yes, we have calibration: the method can recover known functional structure from unlabeled data.
This may be the most valuable research return of the railway experiment. We gain a known world in which the truth exists, deliberately hide it, and see whether Voynich-style inference can recover it.
An unknown manuscript should not be the first place we discover whether our method works.
What Would Make the Railway World Fail?
We should write the failure conditions before becoming fond of the analogy.
| Railway projection | Failure condition | Possible neutral residue |
|---|---|---|
| Interchange | Centrality disappears after frequency control or is not stable across samples. | None, or ordinary high-frequency behaviour. |
| Terminal/reset | Boundary distributions do not differ robustly from matched interior controls. | Boundary class not operationally distinct. |
| Route family | Communities depend strongly on arbitrary tokenisation or page selection. | Segmentation sensitivity becomes the result. |
| Headway | Recurrence intervals carry no information beyond frequency and local clustering. | Frequency sufficient; spacing model discarded. |
| Delay propagation | Local perturbations do not change downstream distributions beyond immediate adjacency. | State appears strongly local. |
| Network overview foldout | No distinctive higher-order connectivity or correspondence appears. | Large format does not imply global interface. |
A good Tangential article should be happy to publish those failures.
The railway is allowed to lose.
The First Railway Experiment Pack
The article now generates a concrete experimental programme rather than a metaphor.
- Build directed adjacency graphs under at least two transcription and segmentation regimes.
- Frequency-normalise centrality and identify candidate bridge forms.
- Compare boundary versus interior transition distributions for line, paragraph and interface boundaries.
- Measure recurrence intervals for common and medium-frequency families, not only raw counts.
- Estimate perturbation decay after defined rare or boundary events.
- Compare running-text and label-like populations as separate interface classes.
- Test community structure across Currier, hand and codicological axes rather than forcing one partition.
- Run the reverse calibration on an anonymised known railway network.
- Freeze a held-out Voynich set before tuning thresholds.
- Remove all railway terminology before any result is promoted.
This is the difference between an entertaining analogy and a research lane.
Metaphor Removal
Now remove the railway.
Remove stations.
Remove tracks.
Remove trains, headways, depots and control centres.
What questions survive?
- Are some recurrent forms structurally important because they bridge otherwise distinct neighbourhoods?
- Do defined boundaries produce measurable state discontinuities?
- Does recurrence spacing contain information not captured by frequency?
- Do local unusual states have downstream statistical effects?
- Do separate manuscript populations share some structural infrastructure while retaining local rules?
- Are some interfaces part of the same symbolic system but governed by different transition distributions?
- Can the analytical method recover known structure when labels are hidden in a control world?
Every one of those questions can be asked without a railway.
That means the first Tangential experiment has already produced transfer yield.
It has not solved Voynich.
It has changed what we know how to ask.
World Return
The Voynich Manuscript is not a railway.
That sentence is not a disappointment.
It is what makes the experiment safe.
A railway controller walks into the manuscript carrying an obsession with movement, capacity, boundaries, route conflict and recovery. Those obsessions are foreign. They distort the object.
Then we use that distortion deliberately.
It makes us inspect adjacency instead of meaning.
It makes us look at what is forbidden instead of only what is frequent.
It makes us treat boundaries as possible state changes.
It makes us ask whether common forms are merely common or structurally bridging.
It makes recurrence distance visible.
It makes us calibrate our analytical methods on a known network before trusting them on an unknown one.
Then the controller leaves.
The railway disappears.
The measurements remain.
This is Tangential Voynich working exactly as intended: take the manuscript somewhere it does not belong, then return with a question that does.
Tangential Voynich: The Wrong-System Protocol → · Voynich Research Library → · How MRT Works | It’s Mathematics →
Next Tangent: put the manuscript inside a computer operating system and ask what changes when “words” become states, processes, calls, memory and permissions.