You reach a junction between horde fights. Two doorway inscriptions use the same groups of signs in a different order. You frame and capture them with their doorways visible, then open the journal. **The whole game pauses: enemies, oxygen and unfinished actions stop, even during combat.**

Pin the records beside each other, zoom in and highlight the swapped groups. Your note might say: “Does the first group describe what this place is for?” Close the journal and follow one way. In the pump hall, a great pump main ends at the machinery while a cable main crosses the room. The signs have given you a reason to look at the architecture differently.

That is the journal’s purpose: carry observations between places, give you time to think and help you choose an experiment. Recognition becomes a prediction you can act on. Later evidence may change the note; its earlier wording remains part of your discovery.

A record is one player-framed still, with its visible context, place, capture time and observed arrangement. Anything outside that frame remains unrecorded. Getting a new angle, watching motion, operating a machine and committing each capture happen while the world runs. The pause lets you examine stored evidence and the frozen view already visible to the astronaut. It cannot finish or batch-complete captures, or reveal a later state.

| Tool | Your action | What is kept |

On request, the recorder can suggest a **visually similar record you already captured**. Suggestions are labelled software-generated and can be wrong. They help bring distant observations together; you decide whether the marks belong together and what they mean. Manual comparison remains available throughout. The journal does not fill a translated dictionary, award correctness or uncover unseen sites.

The [Language Expedition](/Hold%20The%20Flood/Language%20Expedition) uses three existing ways: Pump Hall, Cable Junction and a stair-only vestibule. Here the recorded shapes are represented by readable placeholders:

- Pump Hall’s way: `KETH SORAN VEL`.

In Pump Hall, the pump main ends and the cable main crosses. In Cable Junction, the cable main ends in banked connections while the pump main continues. A small cooling pump also ends there, visibly fed by a local branch. The useful comparison is **which of the two continuous mains ends**, rather than every piece of equipment in the room.

Swapped order now tracks an observable difference. “First position describes the primary purpose; the rest qualify it” becomes a useful working reading. Compound place names remain possible. The short `TAAN` run fits the stair-only case, but one short inscription settles little. The journal keeps the captures and places reached; the grammatical inference remains yours.

If Yeva is aboard, she can identify the crew’s “pump road” and “cable road” from a shown record because she worked this spine. Those attributed names supply place references, not a translation or the positional rule. Either room can be visited first, and repeated physical examples support an unaided reading. A later valid control test may challenge your generalisation; an interrupted observation does not.

The visitors wrote in a different hand. One use of their language places related marks at opposite ends of a transfer. At the junction you record a half written here as `OSH·a`, with another token beside it. Much later, reached by the long route, a far bay has `OSH·b` beside a receiving fitting; its builders’ inscription begins `SORAN`.

The direction comes from other observations. Across two or three previously recorded transfer pairs, a driving device visibly pushes into a line and a vessel at the other end fills. One extra token recurs at driving ends, another at receiving ends. Those actions support a local source/receiver distinction without telling you everything either token means.

Pin the junction and far-bay records. Their additional tokens match the driving and receiving roles respectively. A lookalike suggestion may surface their shared bracket shape; recognizing the roles is your inference. The builders’ runs independently suggest cabling. Together, the two languages point toward a particular connection worth investigating.

There is physical reason to try: beside the junction mark, a conduit enters a human-sized maintenance continuation where you previously turned back on supply. At the far bay, the receiving fitting stands beside another maintenance opening. Material can pass where a person cannot, so neither the paired marks nor those separate openings prove a continuous walkable route or an oxygen saving. You may link the continuation to the far-bay record and attempt it, or use the long route already walked.

Yeva says she never read the visitors’ hand. Her road names cannot supply this inference. Missing `OSH·a` need not end the investigation: another persistent pair can reveal the paired use, and the builders’ runs alone support a weaker route hypothesis. Richer examples of both languages are in the [Two Language Discovery Map](/Hold%20The%20Flood/Two%20Language%20Discovery%20Map).

A voluntary route test attaches one factual result to your prediction:

| Recorded result | What actually happened |

A different terminal can challenge the proposed route without disproving the transfer roles. A blocked passage tells you about present access. It does not establish that the inscription was misunderstood. Even a wrong reading may predict one journey correctly; further comparisons matter.

Controls follow the same principle. A powered, intact control in a valid state, with its relevant component and completed activation, can yield a recorded visible result. Missing power, a removed component or unfinished work leaves absent test conditions or an incomplete attempt, rather than linguistic disproof. You judge what the observation bears on.

Testing costs live time, oxygen and exposure, and may consume the rest of an expedition. You can press on, retreat where practical or return through death with everything still carried. Already consumed oxygen/material stays spent; installed or handed-over objects stay outside the carried set. Temporary combat growth ends at home. Your records, notes, testimony and previous results remain.

Capturing a place again adds a dated record beside the earlier one. After a death shift, both can still inform a language comparison. Their old unbound routes become historical; reaching somewhere different in the new arrangement does not by itself refute a reading. Completed work, objects and persistent joins follow the [current continuity rules](/Hold%20The%20Flood/Death%20Shift%20And%20Rescue%20Continuity). Current neighbours and access need current observation. Neither testimony nor a matching image supplies them.

The [moving scan](/Hold%20The%20Flood/Moving%20Scan%20Expedition) makes the distinction tangible. Its dated captures survive when a changed arrangement invalidates partial or complete instrument samples. The visible return chain rewinds the carriage from its current notch while the game runs; paused comparison cannot rescan or replay it. A living return preserves a valid partial scan for resumption, or a completed bridge for reuse. The journal preserves evidence; the instrument and walked route establish current conditions.