A landmark in this catalogue is not a notable building. It is a structure I demonstrably used — recorded in the log at the moment of use, with the phrase "took bearing from" or "recognised" beside it. Decorative monuments that never entered a navigational decision are not here, and there are a great many of those.
One hundred and eighty-four qualified across forty-eight games, which is fewer than four per game. Several games contributed none at all.
Sorted by what made them work
| Property | Count | Share | Typical example |
|---|---|---|---|
| Asymmetric silhouette | 61 | 33% | A tower leaning, a broken spire, a wall with one gap |
| Unique colour or material | 38 | 21% | The one copper roof, the one whitewashed building |
| Sound at range | 27 | 15% | A waterfall, a mill, a bell on a timer |
| Motion | 22 | 12% | A turning wheel, a flag, a rising column of smoke |
| Light at night | 19 | 10% | Lighthouse, brazier, lit window in an empty region |
| Height alone | 17 | 9% | A tall tower otherwise indistinguishable from others |
Height alone comes last, at nine per cent, and that is the finding that made this note worth writing. Tall structures are the standard tool for making a world navigable, and in practice a tall structure that looks like other tall structures is close to useless. Sixty-one of the one hundred and eighty-four worked because their outline could not be confused with anything else, regardless of how tall they were.
The three lighthouses in Fathom Coast are not tall. They are squat, plain and leaning respectively, and the difference between those three profiles is the entire navigational system of a forty-one-hour game.
The revision: sound was undercounted
The first version of this catalogue, in volume two, had no sound category. That was an error of method rather than of judgement — my log recorded visual references carefully and audible ones informally, so twenty-seven entries were sitting in the notes column unclassified.
Rebuilding the catalogue in volume four moved them in, and sound immediately became the third-largest category. It is also the most reliable in poor visibility: in Signal Tundra, where the fog removes every visual reference, three of the four usable landmarks are audible ones.
Superseded, volume four: the original claim that "visual distinctiveness is the only property that matters" was an artefact of how I was taking notes rather than a finding about games.
What disqualifies a structure
Repetition, obviously. A distinctive building placed four times in one region is not a landmark, it is furniture, and the log has eleven cases of a structure being used as a reference and then abandoned when a duplicate appeared. That moment — realising the thing you were steering by exists twice — is more disorienting than having no landmark at all.
Interaction is the subtler disqualifier. Structures that are quest locations are almost never used as bearings, because the game marks them, and a marked structure is a destination rather than a reference. Fourteen of the strongest silhouettes in the catalogue are buildings with no gameplay function whatsoever.
Distance bands
Landmarks divide cleanly by the range at which they work, and a world needs all three bands. Long-range references — visible across a whole region — are the rarest, and the median game in this sample has one. Mid-range references at a kilometre or two are the most common and the least useful, because they are usually only visible from places you can already navigate. Short-range references, the crooked wall at the end of the lane, do the actual work of local orientation.
The games with full recall all have at least two long-range references. Marker Fields has three excellent ones and I used none of them, which is a reminder that this catalogue records what players use rather than what designers provide.
The one property nobody designs for
Twenty-two entries worked because something moved. A mill wheel, a flag in a specific wind direction, smoke from a forge that is only lit in the mornings. Motion is the strongest attention-getter in the list per instance, and it is almost never used deliberately for navigation — in every case here it appears to be atmospheric detail that happened to become a bearing.
If there is one recommendation buried in one hundred and eighty-four entries, it is that: give one structure per region something that moves, and it will be remembered whether or not anybody planned for it.